How many clinicaltrials.gov studies with an LLM-based Chatbot as the intervention/treatment will be recruiting on November 1, 2026?

- [2026-10-09] Attempts to directly load the resolution-source URL (clinicaltrials.gov/search?...intr=LLM-based%20Chatbot&aggFilters=status:rec) and the ClinicalTrials.gov API returned no accessible content in this research session, so the live, current count of "Recruiting" studies tagged with the exact intervention/treatment name "LLM-based Chatbot" could not be directly verified today. (https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec)
- [2026-10-09] Individual study records that use this specific intervention tag show mixed, mostly non-recruiting statuses: NCT06914492 ("Quit to Win" 2025, Hong Kong) is 
Real-time Smoking Cessation Instant Messaging Support Using a Large Language Model (LLM)-Based Chatbot Via "Quit to Win" 2025 (QTW2025)
 and has an Overall Status of COMPLETED as of its 2025-05 verification. (https://clinicaltrials.gov/study/NCT06914492)
- [2026-06] The successor trial NCT07649044, 
Promoting Smoking Cessation in the Community Via "Quit to Win" Contest 2026: Engagement-Focused GenAI-based Chatbot for Real-Time Smoking Cessation Support
, is listed as NOT_YET_RECRUITING with an estimated start date of 2026-06-27, meaning it could plausibly shift to Recruiting before November 1, 2026. (https://clinicaltrials.gov/study/NCT07649044)
- [2026-04-14] NCT07280650, 
Feasibility and Acceptability of an LLM-based Chatbot for Family Caregivers: Evaluation Study
 (University of Michigan), is NOT_YET_RECRUITING with an estimated start date of 2026-08, so it may have moved to Recruiting by late 2026. (https://clinicaltrials.gov/study/NCT07280650)
- [2025-10] NCT07214831, a Massachusetts General Hospital study of 
A Feasibility and Acceptability Study of a Large Language Model-based Chatbot for Brief Alcohol Intervention Among Emerging Adults
, is NOT_YET_RECRUITING with an estimated start date of 2027-06-01, so it will almost certainly remain not-yet-recruiting through November 1, 2026. (https://clinicaltrials.gov/study/NCT07214831)
- [2026] NCT07741266, 
a pilot randomized controlled trial to assess the feasibility and preliminary effectiveness of an LLM-driven chatbot versus electronic brochure in addressing reproductive concerns among AYA cancer survivors
, is a newly registered study (high NCT number, 2026 cohort); its current recruiting status could not be confirmed in this session. (https://clinicaltrials.gov/study/NCT07741266)
- [2026] NCT07410689, SAFE.AI, describes a 
hybrid SAFE.ai chatbot that builds on an existing rule-based system by integrating a large language model (LLM) layer to support more flexible, open-ended, and conversational interactions
; its exact intervention tag and current recruiting status were not confirmed. (https://clinicaltrials.gov/study/NCT07410689)
- [2026-10-08/09] Independent of the registry count, media coverage this week (NYT, The Independent, Newswise/BIDMC) confirms heavy real-world momentum behind LLM-chatbot clinical research, with the BIDMC/Google AMIE trial — a prospective, physician-supervised, patient-facing chatbot study — 
conducted between April and November 2025, the study enrolled 114 patients, with 98 completing both the AI interaction and their primary care appointment
, illustrating the broader surge of similar interventional registrations feeding into the ClinicalTrials.gov "LLM-based Chatbot" tag, even though that specific trial itself was not found tagged with that exact intervention term. (https://www.newswise.com/articles/bidmc-researchers-conduct-first-real-world-study-of-safety-and-quality-of-patient-facing-ai-in-primary-care)
- [2026-09-01] A related mixed-methods evaluation of an LLM-based behavioral activation chatbot for youth depression was published, underscoring that academic/clinical interest in LLM chatbot interventions is accelerating, which is consistent with new ClinicalTrials.gov registrations appearing throughout 2026. (https://pmc.ncbi.nlm.nih.gov/articles/PMC13533316/)
- [2025, 206 days old] A JMIR methodological review protocol notes that 
since ChatGPT's launch in November 2022, interest in patient-facing conversational agents whose replies are generated by large language models (LLM), in other words, LLM-based chatbots, has surged, with growing attention to their potential clinical applications
, and that 
LLM-based chatbots are reported being adopted at scale by consumers and patients
, supporting a trend of more such trials being registered over time, though this does not itself report a current recruiting count. (https://doi.org/10.2196/90507)

BOTTOM LINE: The exact ClinicalTrials.gov registry page used for resolution could not be fetched directly in this session, but individual records tagged with the "LLM-based Chatbot" intervention show a small, mixed pool (several NOT_YET_RECRUITING or COMPLETED, a few of unconfirmed status), while the broader research field is clearly expanding rapidly (new 2026 trial registrations, high-profile Lancet/BIDMC-Google publication this week); this suggests the Recruiting count by November 1, 2026 is likely modest but could tick upward as 2026-started studies (e.g., NCT07649044, NCT07280650) transition out of "not yet recruiting."

NOT FOUND: The live, current total number of studies shown as "Recruiting" on the exact clinicaltrials.gov search/filter URL specified in the resolution criteria; confirmed current recruiting status for NCT07741266 and NCT07410689; any official ClinicalTrials.gov aggregate count or "results found" figure for the intr=LLM-based Chatbot + status:rec filter.

- [2026-10] Direct access to the clinicaltrials.gov live tracker page (the exact URL used for resolution) and its API endpoint were not fetchable in this session; findings below are inferred from individual study records returned by search, not from a single authoritative count — https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec
- [2026-10] Multiple recently-registered trials using "LLM-based Chatbot" or closely related intervention names are currently listed as NOT_YET_RECRUITING rather than Recruiting, including a University of Michigan caregiver chatbot study 
whose status as of 2026-04 was "NOT_YET_RECRUITING"
 — https://clinicaltrials.gov/study/NCT07280650
- [2026-10] An MD Anderson informed-consent chatbot trial posted in August 2026 is also not yet recruiting, 
sponsored by M.D. Anderson Cancer Center
 with an estimated 2027-01-01 start date, showing new registrations continue to lag actual recruitment — https://clinicaltrials.gov/study/NCT07756515
- [2026-10] A Massachusetts General Hospital alcohol-intervention chatbot trial (NCT07214831) and a Hong Kong "Quit to Win" 2026 smoking-cessation GenAI chatbot trial (NCT07649044) were both listed as NOT_YET_RECRUITING as of mid-2026, reinforcing that many registered LLM-chatbot studies have not yet reached active recruitment status — https://clinicaltrials.gov/study/NCT07214831, https://clinicaltrials.gov/study/NCT07649044
- [2026-10] A Stanford "LLM-Generated Coaching Prompts" physical-activity study was updated as recently as July 2026 and remains 
"NOT_YET_RECRUITING"
 with an estimated January 2027 start — https://clinicaltrials.gov/study/NCT06880315
- [2026-02-ish] An academic review protocol explicitly framed LLM-based chatbot interventional trials as a newly emerging and rapidly expanding category, noting 
"interest in patient-facing conversational agents whose replies are generated by large language models...has surged, with growing attention to their potential clinical applications"
 since ChatGPT's 2022 launch — https://doi.org/10.2196/90507
- [2026-04-23] A broader bibliometric analysis of AI in clinical trials registry data found 
5,828 AI-related clinicaltrials.gov records as of a search conducted 23 April 2026, representing slightly over 1% of all ClinicalTrials.gov records
, and specifically noted that 
"chatbot" has risen since 2018, and "ChatGPT"/"GPT" has substantially risen since 2022
 — https://arxiv.org/pdf/2605.29096
- [2026-10-08] Independent of the registry count, the field received major news momentum this week via the AMIE/BIDMC/Google Lancet publication, which could spur sponsors to register new similarly-labeled LLM chatbot trials, but this is a lagging indicator (new registrations take weeks-months) rather than something that would change the November 1 count — https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
- [2026-10] A handful of trials do appear to reach Recruiting status under similar labels, e.g. an intelligent-chatbot tobacco-support pilot RCT and a breast-cancer chatbot knowledge study used "Recruiting"/"Not yet recruiting" terminology consistent with the registry's standard lifecycle, but exact current counts tagged specifically "LLM-based Chatbot" (the precise intervention-name string used by the resolution link) could not be enumerated from search snippets alone — https://clinicaltrials.gov/study/NCT06952725, https://clinicaltrials.gov/study/NCT06943911
- [general] No historical analog exists for "how many times this exact clinicaltrials.gov filtered count hit a particular number," since this is a narrow, newly-coined intervention-name filter (LLM-based Chatbot) rather than a recurring contest with past outcomes; thus no favourite/odds/poll base rate exists for this specific metric.

BOTTOM LINE: There is no direct "contest" base rate for this metric since it's a live registry count rather than a historical recurring event; the available evidence shows the "LLM-based Chatbot" intervention category is small but growing, with many newly registered trials sitting in Not-Yet-Recruiting status for months before transitioning to Recruiting, suggesting the November 1, 2026 recruiting count will likely remain in a modest range rather than show a dramatic jump, modestly favoring continuity/slow-growth outcomes over large discrete jumps.

NOT FOUND: The exact current live count of studies at the resolution URL (https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec) could not be retrieved, as both the search-page URL and the clinicaltrials.gov API endpoint were blocked from direct fetching in this session; no historical snapshot/trend chart of this specific filtered count over time was found; no existing third-party forecasts, polls, or odds on this specific question were found.

- [2026-10-06] A brand-new MGH study, CHAT-HX (NCT07860762), was first posted on 2026-10-06 with intervention listed as an "AI-Based Structured History-Taking Chatbot" described as 
"A large language model-based agentic chatbot that conducts a structured, pre-visit history interview on a study tablet and generates a history summary, differential diagnosis list, and workup suggestions"
 and is currently Not yet recruiting, illustrating that new LLM-chatbot trials are still entering the registry pipeline literally days before the question's close (https://clinicaltrials.gov/study/NCT07860762).
- [2026-09-25] Another freshly registered trial, PersianAI-CBT (NCT07840469, University of Isfahan), confirms continued global registration momentum, with a large language model supporting 
"the conversational functionality of the chatbot"
 while it remains Not yet recruiting as of its last update (https://clinicaltrials.gov/study/NCT07840469).
- [2026-03-31] NCT07504367 ("Large Language Models Assist in Tumor MDT," Sun Yat-Sen Memorial Hospital) is one of the few comparable LLM-intervention trials already showing Overall Status RECRUITING, with an actual start date of 2026-01-01 and estimated completion 2026-12-31, demonstrating that at least some trials in this space do successfully convert to and remain in Recruiting status through the resolution window (https://clinicaltrials.gov/study/NCT07504367).
- [2026-04-30] NCT06949462 (PEAR, Singapore General Hospital, LLM for anaesthesia consent) was also marked RECRUITING as of its last update, but its actual/estimated completion date of 2026-04-27 has already passed, meaning it may have since transitioned to Completed by November 2026 — a reminder that the recruiting pool is a moving target with inflows and outflows (https://clinicaltrials.gov/study/NCT06949462).
- [2026-07-23] The Stanford "LLM-Generated Coaching Prompts" study pushed its estimated start date back to 2027-01-20 in a July 2026 update, confirming that sponsor-side delays (not registry mechanics) are a key factor keeping some pipeline trials out of Recruiting status past November 1 (https://clinicaltrials.gov/study/NCT06880315).
- [2026-07] University of Michigan Rogel Cancer Center's AI supportive-care chatbot trial (NCT07772596) lists an estimated Start Date of 2026-10-01, among the closest upcoming start dates found, making it a plausible candidate to flip to Recruiting by November 1 if the sponsor updates status on schedule (fetched record, search result 15).
- [2026-08-13] Brigham and Women's Hospital's prostate-RT patient-education chatbot (NCT07719062) has an estimated Start Date of 2026-08, already in the past relative to today, meaning the sponsor-controlled step of updating status from Not yet recruiting to Recruiting is overdue and could occur any time before November 1 (https://clinicaltrials.gov/study/NCT07719062 via search result 13).
- [2026-08] ChatACP (NCT07448649, University of Hong Kong) explicitly states per its JMIR protocol that 
"This project was recommended for support in September 2025... Recruitment and data collection will commence in September 2026"
, indicating the sponsor's own timeline places its transition to Recruiting squarely within the question's window (https://pubmed.ncbi.nlm.nih.gov/42585665/).
- [2026-06] A CABG complication-prediction LLM chatbot trial (NCT07432893) states 
"Phase II usability evaluation is projected from May to July 2026. Phase III recruitment is anticipated from August 2026 to September 2027"
, another sponsor-stated timeline implying a possible shift into Recruiting before the resolution date (https://doi.org/10.2196/103717).
- [2026-06-11] Tsinghua University's "Large Language Model in Outpatient Care" trial (NCT07641478) lists an estimated Start Date of 2026-06-12 and estimated Completion Date of 2026-09-04 — both already in the past relative to today — meaning this trial has likely already moved through and possibly out of Recruiting status before the resolution date, a mechanism by which the count can also shrink (https://clinicaltrials.gov/study/NCT07641478).

BOTTOM LINE: The key drivers of the November 1 count are (1) sponsor-controlled status updates — several studies with 2026 Q3/Q4 estimated start dates (CHAT-HX, Rogel Cancer Center, BWH prostate education, ChatACP, Tumor MDT) sit right at the threshold of flipping from Not-yet-recruiting to Recruiting, and (2) continued weekly new registrations (two more added just in September-October 2026), both pushing the count up, partially offset by trials whose completion dates have already passed (PEAR, Tsinghua) that may exit Recruiting before the deadline.

NOT FOUND: The live resolution-URL page and ClinicalTrials.gov API count could not be directly fetched (domain access blocked); no authoritative current total of studies matching the exact "LLM-based Chatbot" intervention term with status Recruiting was obtainable, so the analysis relies on individual study records rather than the aggregate search count.

- [2026-08-07] Kalshi and Polymarket have expanded into biotech/clinical-trial-related prediction markets, but 
Kalshi launched a suite of biotech prediction markets on July 16, letting users wager on Phase 3 clinical trial outcomes and full FDA approvals from major pharmaceutical companies
, not on registry metadata/intervention-tag counts. (https://cryptobriefing.com/kalshi-polymarket-drug-approval-betting/)
- [2026-05 onward] A niche clinical-trial-specific market, Endpoint Arena, has also launched, but 
Endpoint Arena is still in pilot mode and currently only allows... less than $50,000 is currently being wagered across all biotech-related contracts on Kalshi combined, and Endpoint Arena has yet to move beyond paper trading
 — i.e., even the specialized market infrastructure is thin and focused on drug-approval/endpoint outcomes, not trial-registry search counts. (https://www.fiercebiotech.com/biotech/betting-biotech-prediction-markets-set-sights-clinical-trials)
- No Metaculus, Manifold, Polymarket, or Kalshi question could be found that forecasts the number of ClinicalTrials.gov studies tagged with intervention "LLM-based Chatbot" and status "Recruiting." Searches for this exact metric and for related "bot tournament" forecasting questions returned no public market or community-forecast page.
- The only "tournament" context found is the metadata tag embedded in the question itself (`bot_tournament_question`), indicating this question was likely authored for an internal/private AI-forecasting-bot benchmarking exercise rather than posted to a public crowd-forecasting platform; no externally visible community prediction distribution exists for it.
- Academic/clinical commentary (not a market, but the closest thing to "expert opinion" available) is cautious rather than prescriptive about growth in registered chatbot trials. Coverage of the BIDMC/Google AMIE study (Oct 8, 2026, published in The Lancet) emphasizes that the evidence "does not yet support autonomous use for diagnosis, triage, or management," per the researchers, and UK RCGP president Victoria Tzortziou Brown called for larger, independent studies before wider clinical adoption — suggesting expert sentiment leans toward cautious, incremental expansion rather than a rapid surge in new recruiting trials. (https://www.independent.co.uk/news/uk/home-news/ai-chatbot-patients-diagnosis-gp-b3063668.html)
- Related academic surveys note a broader proliferation of LLM-chatbot studies registering on ClinicalTrials.gov across many clinical domains (oncology, diabetes transition care, depression screening, mental health, alcohol intervention, caregiver support), several already registered or recently posted (e.g., NCT07741266, NCT07280650, NCT06801925), indicating an active and growing pipeline feeding into the "Recruiting" status category, though with no single source aggregating or forecasting the exact future count. (https://clinicaltrials.gov/study/NCT07280650)

BOTTOM LINE: No prediction market, poll, or formal expert forecast exists for this specific ClinicalTrials.gov intervention-tag count; the closest available signals are indirect — active biotech prediction markets (Kalshi, Polymarket, Endpoint Arena) focus on trial outcomes/approvals rather than registry metadata, and clinical experts commenting on recent high-profile chatbot trials (e.g., AMIE/Lancet study) express caution about rapid clinical deployment, suggesting no strong external consensus pushes toward either a large increase or decrease in the recruiting-study count by November 1, 2026.

NOT FOUND: Any Polymarket, Kalshi, Manifold, or Metaculus market/question directly tracking this ClinicalTrials.gov intervention-filtered count; any published analyst projection or poll of the expected number of "Recruiting" LLM-based Chatbot studies; any public track record of forecasters on this specific or closely analogous registry-count questions.

- [2026-04] The University of Michigan caregiver chatbot trial (NCT07280650) has an official intervention title "LLM-based Chatbot" and was listed with an estimated start of 2026-08, which has now passed, making it the single strongest near-term candidate to flip to Recruiting by Nov 1, 2026. 
The study is titled "Feasibility and Acceptability of an LLM-based Chatbot for Family Caregivers: Evaluation Study" and was registered with status Not Yet Recruiting as of April 2026
, with 
a start date estimated for 2026-08 and a status verified date of 2026-04
.
- [2026-08-03] A second strong candidate, NCT07741266 (LLM-driven chatbot for reproductive concerns in AYA cancer survivors), was first posted August 3, 2026 with 
status "Not yet recruiting"
 and 
an estimated study start of September 1, 2026
 — a date that has already passed as of the Nov 1 resolution point, making a shift to Recruiting plausible.
- [2026-06] NCT07649044 ("Quit to Win" 2026 GenAI chatbot, University of Hong Kong) remains the Hong Kong program's current entry, verified 
as "Promoting Smoking Cessation in the Community Via 'Quit to Win' Contest 2026: Engagement-Focused GenAI-based Chatbot for Real-Time Smoking Cessation Support"
 with 
status "NOT_YET_RECRUITING" and an estimated start date of 2026-06-27
 — its own target start date is already four months in the past relative to resolution, suggesting high likelihood it has transitioned to Recruiting (or already completed early phases) by Nov 1.
- [2026] The SAFE.AI hybrid chatbot study (NCT07410689) is a plausible marginal case: it uses a "Hybrid Chatbot" intervention with an LLM layer rather than the exact "LLM-based Chatbot" label, meaning it may or may not be captured by the precise intervention-filter string used in the resolution URL. 
The intervention is named "Hybrid Chatbot," described as building on the existing rule-based system by integrating a large language model (LLM) layer to support more flexible, open-ended, and conversational interactions
.
- [2026-07-28] The reproductive-health chatbot trial's registry history shows a typical ~1-month lag from first submission to first posting 
(first submitted July 28, 2026; first posted August 3, 2026)
, confirming the general pattern that new "LLM-based Chatbot" registrations continue to appear roughly monthly, feeding the pool of studies that could reach Recruiting status by November.
- [2026-10] Field-wide academic framing continues to describe the "LLM-based Chatbot" intervention class as a rapidly expanding category rather than a niche: 
"interest in patient-facing conversational agents whose replies are generated by large language models (LLM), in other words, LLM-based chatbots, has surged, with growing attention to their potential"
 clinical applications, supporting continued net registration growth into the category that underlies the resolution count.
- [2026] No trial using the exact "LLM-based Chatbot" intervention tag was found with a Recruiting status as of this search session; all directly-tagged records identified (NCT06914492, NCT07280650, NCT07741266, NCT07649044, NCT07214831) are either Completed or Not Yet Recruiting, indicating the live Recruiting count under the strict filter is currently small (plausibly zero or very low) rather than in a high range.
- [2026-10] Attempts to query the ClinicalTrials.gov v2 API directly (query.intr=LLM-based Chatbot, filter.overallStatus=RECRUITING) and to load the resolution's exact search URL were both blocked in this session (domain/API access restricted), so no authoritative live count could be independently confirmed beyond individual study record checks.

BOTTOM LINE: Among studies explicitly tagged with the intervention "LLM-based Chatbot," the only plausible near-term additions to a "Recruiting" count by Nov 1, 2026 are a small handful of 2026-registered trials (Michigan caregiver study, AYA cancer-survivor reproductive chatbot, HKU "Quit to Win" 2026) whose estimated start dates have already passed and could have flipped from Not-Yet-Recruiting to Recruiting; this favors a low-but-nonzero number (likely in the single digits) rather than a large count, consistent with the "modest range" leader identified in prior research.

NOT FOUND: The authoritative, current total shown on the exact resolution URL/API query (blocked from direct access); confirmed post-October 2026 status updates for NCT07280650, NCT07741266, or NCT07649044 showing an actual transition to Recruiting; any comprehensive list enumerating every study carrying the precise "LLM-based Chatbot" intervention tag.

**Case for an upset: the true November 1, 2026 "Recruiting" count is higher than the modest/low number implied by the leading estimate.**

- [2026-06] NCT07649044 ("Quit to Win" Contest 2026, University of Hong Kong) had its 
Status Verified Date: 2026-06, Overall Status: NOT_YET_RECRUITING, Start Date: 2026-06-27, Type: ESTIMATED
 — meaning its own estimated start date has already passed as of this writing (Oct 9, 2026), so the record is stale and the study is very likely to have flipped to Recruiting well before November 1, strengthening the case for at least one more "Recruiting" hit than a snapshot taken months ago would suggest. https://clinicaltrials.gov/study/NCT07649044
- [2026-04] NCT07280650 (University of Michigan caregiver chatbot) listed an 
Overall Status: NOT_YET_RECRUITING ... Start Date: 2026-08, Type: ESTIMATED
 — that estimated start month (August 2026) has also already passed, meaning this record too is plausibly outdated and the study may already be actively recruiting by November 1, which would independently add to the count rather than keep it flat. https://clinicaltrials.gov/study/NCT07280650
- [2026, undated] NCT07741266, the AYA cancer-survivor reproductive-concerns LLM chatbot trial, is a freshly registered high-numbered NCT (2026 cohort) whose detailed protocol describes an active randomization and recruitment pipeline 
Simple random sampling with a fixed sample size will be employed to recruit subjects... Randomization will be automatically performed using a computer-generated randomization sequence embedded in the chatbot backend
 — this phrasing suggests the trial infrastructure is essentially recruitment-ready, not a years-out protocol, making a near-term flip to "Recruiting" plausible before Nov 1. https://clinicaltrials.gov/study/NCT07741266
- [2026-10-08] The Lancet/AMIE publication and its accompanying NYT, Independent, and Newswise coverage this week generated unusually high-profile validation that LLM chatbots can meaningfully assist primary-care physicians, with doctors finding the bot helpful in "approximately 75 percent of patient cases" — this kind of landmark positive result historically triggers rapid follow-on trial registrations and accelerated recruitment launches at peer institutions within weeks, which could pull additional studies into "Recruiting" status before the Nov 1 cutoff even though the lag is typically measured in months. https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
- [2026-10-07] A newly published Nature paper on a BERT/LLM-based medical chatbot with cryptographic protection shows the technical research pipeline for "medical chatbot" trial designs is still very active and diversifying (security-focused, diagnostic, behavioral-health variants), which expands the universe of candidate studies that could newly post or transition into the registry's "LLM-based Chatbot" intervention tag by Nov 1 — a mechanism the leading (likely static/low) estimate may underweight. https://www.nature.com/articles/s41598-026-72040-w
- [general, multiple records] Both NCT07649044 and NCT07280650 share a structural pattern seen repeatedly across this intervention tag: studies are registered with "Estimated" start dates that, once passed, are frequently updated to Recruiting with a lag of only a few weeks on ClinicalTrials.gov administrative updates — because at least two identified trials have already sailed past their listed estimated start dates without a confirmed status update being captured in this research session, the true current count is plausibly undercounted by official-looking "Not yet recruiting" labels that are simply stale, not accurate.

BOTTOM LINE: The strongest upset case is for a higher recruiting count than the leader's modest/low estimate: at least two specifically identified "LLM-based Chatbot"-tagged trials (NCT07649044, NCT07280650) have already passed their own estimated start dates as of Oct 9, 2026 and are likely to have flipped to Recruiting by Nov 1, while a newly registered 2026 trial (NCT07741266) appears recruitment-ready, and the Oct 8 Lancet/AMIE publicity wave could accelerate additional launches.

NOT FOUND: Confirmed, dated (post-June/April 2026) status updates for NCT07649044 and NCT07280650 showing an actual flip to Recruiting; any live fetch of the exact resolution-URL count (domain access was blocked in this session); official total enumeration of all currently Recruiting studies under this precise intervention-name tag.

14 AskNews articles (newest first; 6 from the last 48 h, 10 from the archive). Summaries are AskNews's own; dates are publication dates.

[2026-10-09] BIDMC and Google Conduct First Real-World Safety Study of Patient-Facing AI in Primary Care
newswise — https://www.newswise.com/articles/bidmc-researchers-conduct-first-real-world-study-of-safety-and-quality-of-patient-facing-ai-in-primary-care
  - Researchers at Beth Israel Deaconess Medical Center (BIDMC), in collaboration with Google, conducted the first prospective real-world study of the Articulate Medical Intelligence Explorer (AMIE), a patient-facing AI chatbot, in primary care settings.
  - Conducted between April and November 2025, the study enrolled 114 patients, with 98 completing both the AI interaction and their primary care appointment.
  - Every conversation was monitored in real time by a board-certified internal medicine physician who could intervene if a patient appeared at risk of harm.
  - Across all 98 completed encounters, none required a safety-stop intervention; physicians identified one hallucination and provided additional clinical clarification in five cases.
  - Patients rated the quality of AMIE's conversations favorably, and their attitudes toward AI improved after interacting with the chatbot.
  - However, concerns remained regarding trust in the confidentiality of information and the chatbot's honesty.
  - Primary care providers reviewed AI-generated summaries for 44 cases; in 75% of these cases, physicians said the summary helped them prepare, and in 57% it may have influenced their clinical approach.
  - Only one provider reported the interaction as somewhat harmful, citing patient anxiety over a possible lymphoma diagnosis.
  - The study, published in The Lancet, was designed to evaluate feasibility and safety rather than health outcomes.
  - The research was funded by Alphabet, and Adam Rodman, MD, served as a visiting researcher at Google during part of the study.

[2026-10-08] AI Chatbots Show Promise in Pre-GP Diagnosis but Require Oversight, Study Finds
The Independent — https://www.independent.co.uk/news/uk/home-news/ai-chatbot-patients-diagnosis-gp-b3063668.html
  - A small study conducted in the US involving 98 patients at a GP clinic in Boston suggests that AI chatbots could assist in diagnosing patients days before their scheduled GP visits.
  - Published in The Lancet on October 08 2026, the research found that the AI's suggested diagnoses matched confirmed diagnoses in more than half of cases, with an accuracy rate of 75 per cent when considering the chatbot's top three suggestions.
  - Doctors monitored conversations in real time and received transcripts prior to appointments.
  - In 75% of cases reviewed via a questionnaire, medics stated the AI chatbot helped them prepare for patient visits, and it may have changed their approach for more than half of cases.
  - Patient attitudes towards AI in primary care improved after interacting with the chatbot.
  - However, the study noted safety concerns: one conversation was rated 'somewhat harmful' by a doctor after the chatbot listed a possible cancer diagnosis which caused patient anxiety, and in another instance, the chatbot hallucinated, adding clinical information in five interactions.
  - Researchers stress that while the study shows initial feasibility, the evidence 'does not yet support autonomous use for diagnosis, triage, or management.' Professor Victoria Tzortziou Brown, president of the Royal College of GPs, emphasized that the study was small, conducted in a single US practice, and relied on close doctor supervision, meaning it cannot determine safety or effectiveness without that level of oversight.
  - She noted that GPs bring clinical judgment and history knowledge that AI should support, not replace, and called for larger, independent studies to examine safety, inequalities, workload, and patient outcomes before introduction into NHS care.

[2026-10-08] Clinical Trial Shows AI Chatbot AMIE Assists Doctors in 75% of Cases
The New York Times — https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
  - A clinical trial conducted over eight months at Beth Israel Deaconess Medical Center in Boston tested a specialized Google AI chatbot named AMIE (Articulate Medical Intelligence Explorer) to determine if it could assist primary care physicians without providing misleading information.
  - Published on Thursday in the medical journal The Lancet, the study found that doctors found the bot helpful in approximately 75 percent of patient cases, with the technology influencing how they handled more than half of those cases.
  - Peter Brodeur, a resident physician at the hospital, described the bot's performance as comparable to 'a high-performing medical student or resident.' The chatbot, powered by Google's Gemini technology, explored symptoms and generated summaries for doctors to review before patient visits.
  - The trial monitored hundreds of conversations and found the bot 'hallucinated' (made up information) only once, when it misrepresented a date.
  - Adam Rodman, an internist and medical historian at Beth Israel, noted that patients only trusted the bot because it operated alongside trained physicians, stating, 'None of our patients trusted it until they realized that it was handing-off their doctor.' During the trial, physicians monitored the chatbot and intervened to provide context on five occasions, including one instance where the bot listed cancer as a potential diagnosis.
  - Robert Califf, a cardiologist at Duke and former FDA commissioner, commented that while the technology represents the 'future,' it remains unclear if it will reduce the load on doctors or unearth new problems.

[2026-10-07] BERT-Based Medical Chatbot with Application-Level Cryptographic Protection Achieves 99% Accuracy
Nature — https://www.nature.com/articles/s41598-026-72040-w
  - Published on October 07, 2026, in the journal Nature, researchers proposed a BERT-based medical chatbot architecture that integrates contextual natural language processing with an application-level cryptographic protection layer for client-server communication.
  - The study utilized the HealthFC and TREC-Health datasets, implementing the NLP component via the Hugging Face Transformers library with TensorFlow and spaCy.
  - For security, the system employs RSA for AES secret key distribution, AES for encrypting the medical payload, and SHA-256-based HMAC for message integrity and authentication.
  - The model achieved 99% accuracy, 99% precision, 97% recall, 98% F1-score, and 97% AUC-ROC.
  - In a five-fold stratified cross-validation, accuracy values were recorded as 1.00, 0.99, 0.99, 0.98, and 0.99.
  - These results outperformed baseline models including LSTM (0.88 accuracy), SVM (0.91 accuracy), and Bi-LSTM (0.94 accuracy).
  - The authors clarify that this cryptographic mechanism serves as an additional layer complementing standard TLS, rather than replacing it, and does not protect plaintext data during server-side inference.
  - The study notes that clinical safety, external validation, paraphrase-aware leakage analysis, end-to-end latency, and multilingual or voice-based capabilities require further experimental evaluation.

[2026-10-07] CONIFAS2 Study Protocol: Feasibility of Nature-Based Intervention for Children with ADHD
PLOS — https://journals.plos.org/plosone/article
  - The CONIFAS2 study, a two-arm feasibility randomised controlled trial (RCT) protocol published by PLOS on October 07, 2026, aims to assess the feasibility and acceptability of a co-produced, nature-based intervention for children aged 6-11 with ADHD.
  - The study is being conducted across children's neurodiversity ADHD services, initially within the Humber Teaching NHS Foundation Trust (HTFT) and Navigo Community Interest Company (CiC) in North East Lincolnshire, with phased inclusion of Cornwall Partnership NHS Foundation Trust and Leeds Community Healthcare NHS Trust.
  - The intervention, delivered over a 6-week period, consists of a nature-based activity box (designed for at least ten minutes of daily nature exposure), psychoeducational resources, and fortnightly telephone support (three calls of approximately 30 minutes) from a neurodiversity ADHD service clinician, alongside an optional online community forum moderated by a Kids charity organisation.
  - The study will compare the intervention plus treatment as usual (TAU) to TAU alone, with a proposed sample size of 84 participants (42 per arm) to ensure 35 evaluable participants per group, accounting for an anticipated 15% attrition rate.
  - As a feasibility trial, it is not powered to formally detect clinical effectiveness but aims to provide data on recruitment, retention, and intervention fidelity to inform a future large-scale effectiveness trial.
  - The study collects health economic data to assess the potential social and economic impact of the intervention.
  - Primary outcome measures include the Child Health Utility 9D (CHU9D) index, while secondary measures include the SNAP-IV 26-item scale, RCADS-P, and the NCI.
  - The protocol adheres to SPIRIT 2025 reporting guidelines and will be reported according to the CONSORT 2010 statement extension for pilot and feasibility trials.

[2026-10-06] German Study Tests AI Chatbots for Rheumatism Patient Queries
الوطن — https://www.alwatan.com.sa/article/1186817
  - A German study published in the 'Journal of Medical Systems' tested 10 AI-powered chatbots designed to answer questions from rheumatism patients, utilizing 6,291 real patient inquiries collected between September 2025 and January 2026.
  - The chatbots, which were developed based on German clinical guidelines and deployed across 13 rheumatism centers and 6 patient organizations, addressed topics where questions about the disease itself accounted for 50.2%, followed by medications and follow-up at 37.3%, and diagnosis at 28.2%.
  - While 92.9% of 2,671 user-rated responses received positive feedback and 80.2% of participants viewed the robots as a useful addition to patient education, the study found that only 45% of the answers fully adhered to medical guidelines.
  - Researchers emphasize the need to verify accuracy and safety before expanding the technology's use in healthcare, noting that insufficient detail was the most common reason for rejecting answers, and 4.2% of interactions yielded no response.

[2026-09-22] AI Chatbot 'Suzy' Implements Multi-Tier Safety Guardrails and Human Monitoring for Suicide Risk in Opioid Treatment Clinics
Nature — https://www.nature.com/articles/s41746-026-03288-9
  - Published on September 22, 2026, in Nature's npj Digital Medicine, a study by researchers at a large urban general hospital in Massachusetts describes the implementation of 'Suzy,' an AI chatbot designed to monitor suicide risk and respond to patient distress in high-risk addiction medicine settings.
  - The chatbot, built using Open Chat Studio (Dimagi, Inc) and OpenAI LLMs (initially GPT 4o, then 5.0), serves adults receiving medication treatment for opioid use disorder (OUD).
  - The system employs a 'Safety Router' to classify user inputs into three risk tiers: Tier 1 (clear suicide/self-harm risk), Tier 2 (ambiguous distress), and Tier 3 (no risk).
  - When Tier 1 is triggered, the chatbot outputs an urgent, supportive safety disclaimer directing users to emergency services like 911 or 988, while Tier 2 messages are sent to a 'Disambiguate Node' to gather more context.
  - A critical component of the safety framework is a human-monitoring layer where a trained bachelor-level staff member reviews chat transcripts twice daily on weekdays, using the Columbia-Suicide Severity Rating scale (C-SSRS) to assess risk; if imminent risk is found, a licensed psychologist joins the call for triage.
  - The authors, including co-authors JS, LG, DS, and SC, emphasize that while AI guardrails are necessary, they are insufficient without human oversight, recommending that all behavioral health chatbots incorporate rigorous safety architectures and real-time human triage.
  - The study, which was IRB-approved with written informed consent from all patients, notes that patient data was not retained via a zero data retention agreement with OpenAI to ensure HIPAA compliance.
  - The researchers conclude that a multi-pronged approach involving safety guardrails, human triage, ongoing expert testing, and hybrid AI architectures is essential for scaling AI support in high-risk clinical environments.

[2026-08-20] Scoping Review Identifies Mental Health Harms of LLM-Based Chatbots (analysis)
Nature — https://www.nature.com/articles/s41746-026-03054-x
  - A scoping review published in npj Digital Medicine on August 20, 2026, synthesizes research on the mental health harms of Large Language Model (LLM)-based chatbots.
  - The authors conducted a PRISMA-based systematic literature search across five databases (ACM, IEEE, PubMed, Science.gov, Google Scholar) and a supplementary search, identifying N=3,137 articles.
  - Of these, N=119 articles were included that focused on LLM-based chatbots and their potential harms to mental health.
  - The literature was categorized into five areas: (1) Conceptual works identifying harms from chatbot limitations such as hallucinations, sycophancy, and bias, alongside data security risks and dangers for high-risk psychiatric cases like suicide and psychosis; (2) Vignette studies showing chatbots respond inappropriately to mental health queries compared to clinical standards; (3) Cognitive overreliance linked to decreased cognitive and academic performance; (4) Problematic use characterized by emotional or social dependency and withdrawal, which correlates with mental health symptoms; and (5) Articles on AI psychosis suggesting chatbots may reinforce and validate delusional beliefs.
  - The review concludes that both theoretical and empirical work identify various hypothetical and observed harms associated with LLM chatbot use in mental health contexts.

[2026-07-27] Protocol for Randomized Controlled Trial on Chatbot Intervention for Adolescent Digital Addiction
Springer — https://link.springer.com/article/10.1186/s12888-026-08375-5
  - The study aims to develop and evaluate a low-intensity, secondary intervention using a Stage of Change (SoC) and Motivation Interviewing (MI)-based Chatbot to reduce digital addictions (Internet Gaming Disorder [IGD] and Social Media Addiction [SMA]), sedentary lifestyles, and mental distress among adolescents.
  - The research is a two-arm randomized controlled trial (RCT) involving an expected 330 participants with clinical digital addiction symptoms, comparing a five-week Chatbot-assisted intervention group against a psycho-educational control group, both delivered via WhatsApp.
  - The primary outcome is the composite score of IGD and SMA, with secondary outcomes including quality of life, psychological distress, and eHealth literacy, while perceived autonomy and competence are examined as mediators.
  - Evaluations are scheduled at baseline (T0), post-program (T1), and a 3-month follow-up (T2).
  - The trial, pre-registered at ClinicalTrials.gov (NCT06821373) on February 09, 2025, seeks to determine if digital interventions offer a more cost-effective and scalable alternative to traditional methods, particularly in resource-limited settings.

[2026-07-21] Study Reveals High Attrition in LMIC Chatbot Health Interventions Due to Structural and Platform Barriers (analysis)
Nature — https://www.nature.com/articles/s44482-026-00027-5
  - A mixed-methods study published in npj Digital Public Health (Source: Nature, July 21, 2026) evaluated the 'ParentText' chatbot intervention across three Low- and Middle-Income Countries (LMICs): Malaysia, Jamaica, and the Philippines.
  - The study found that while enrolment figures were high, 'effective access' was low, with 93.8% of participants completing less than 25% of the content and a median engagement of only two days.
  - Despite the programme being designed for 38 days, most participants disengaged within the first week.
  - Retention rates by Day 7 varied significantly by site: 51.3% in Malaysia, 18.9% in Jamaica, and 13.3% in the Philippines.
  - Mean engagement duration was 11.18 days in Malaysia versus 4.69 in Jamaica and 3.45 in the Philippines.
  - Qualitative data indicated that caregivers valued the content but disengaged due to structural barriers, including message frequency, length, and platform automation policies that restricted re-contacting users who missed a response window.
  - To address these issues, a consensus panel identified 29 redesign optimisations, with 69% rated as highly feasible.
  - Key recommendations include reducing message burden, shifting to progress-based delivery, and treating re-engagement as a core feature.
  - Selected optimisations were subsequently tested in randomised trials in South Africa and Malaysia, with plans to extend to Jamaica and the Philippines contingent on funding.

[2026-06-25] HopeBot: LLM-Powered Chatbot for Depression Screening Evaluated in UK and China
PLOS — https://journals.plos.org/digitalhealth/article
  - The World Health Organisation (WHO) estimates that depression affects approximately 3.8% of the global population, with 12-month prevalence rates around 4-5% in the United Kingdom (UK) and 3-4% in China, yet only about half of affected individuals receive adequate treatment.
  - To address barriers such as the need for trained professionals and societal stigma, researchers developed HopeBot, an LLM-powered (GPT-4o) conversational chatbot designed to administer the Patient Health Questionnaire-9 (PHQ-9) screening tool.
  - Unlike previous systems that relied on rigid rule-based intents, HopeBot utilizes retrieval-augmented generation (RAG) to maintain the validated structure and scoring logic of the PHQ-9 while allowing for flexible, empathic dialogue and real-time clarification of ambiguous user inputs.
  - A prospective mixed-methods study conducted by the authors with 132 participants from diverse educational and cultural backgrounds in the UK and China evaluated the system's feasibility and score-level concordance against self-administered PHQ-9 assessments.
  - The study focused on format-level validation rather than diagnostic validity, as establishing diagnostic accuracy would require comparison with a clinician-administered reference.
  - The research aims to provide a scalable, user-friendly method for early depression detection that preserves psychometric integrity while enhancing procedural efficiency.

[… older articles omitted]

## Tool outputs

### resolution_source (SUCCESS)
RESOLUTION SOURCE (the page the question resolves on — treat as the status-quo anchor):
  Current state: The fetched clinicaltrials.gov search page returned no actual study listing data—only generic navigation/glossary text—so the current count of Recruiting studies with intervention 'LLM-based Chatbot' cannot be determined from this fetch.
  Last updated: unknown | already resolved: unclear | page matches criteria: True
  Note: The clinicaltrials.gov page content was not actually loaded/rendered in the fetch (likely JS-driven search results not captured); only boilerplate text appeared. The IEEE Spectrum article is background context and does not contain the study count needed for resolution. The question resolves based on the live count on Nov 1, 2026, which is a future date relative to this fetch, so the outcome cannot yet be known.
  Fetched: https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec (HTTP 200)
  Fetched: https://spectrum.ieee.org/ai-clinical-decision-support (HTTP 200)

### web_search (SUCCESS)
Web search summary (20 results):
## Summary of Findings

### Query 1: "clinicaltrials.gov LLM-based chatbot recruiting studies count"

No single aggregate "official count" of LLM-based chatbot trials with Recruiting status was found, but several relevant data points emerged:

- **Systematic review (PLOS Digital Health)**: 
A review of clinical trial registries uncovered 22 registered trials currently exploring the applications of LLMs in healthcare, registered between 2023 and 2024, in various stages, with 9 actively recruiting participants (41% of the total), 7 have not yet started, and the remaining 6 are either ongoing or nearing completion.
 (Source: journals.plos.org/digitalhealth, article dated ~Feb 2024 report references)

- **Individual trial examples found on ClinicalTrials.gov**:
  - NCT06943911 ("BC-CHAT") — 
a breast cancer AI chatbot study, "Not yet recruiting," with 122 breast cancer patients recruited from Kafr El-Sheikh University Hospital, randomly assigned to intervention and control groups.

  - NCT07250516 — a China-based study on 
"LLM-based AI Dialogue Bot"
 with 
an enrollment count of 6,360 (actual).

  - NCT07860762 ("CHAT-HX") — 
"Augmenting Clinical Workflow With a Structured Chatbot for Patient History Collection and Diagnostic Support," sponsored by Massachusetts General Hospital, first submitted 2026-09-30.

  - NCT06880315 (Stanford) — 
"LLM-Generated Coaching Prompts," status "Not Yet Recruiting," with an estimated start date of 2027-01-20.


**Date range**: Sources span 2023–2026 (most recent trial records from September 2026).

---

### Query 2: "clinicaltrials.gov API query.intr='LLM-based Chatbot' filter.overallStatus=RECRUITING countTotal"

No direct result returning an actual numeric count for this specific query was found. However, extensive technical documentation on how this API query works was found:

- The ClinicalTrials.gov v2 API supports 
query.intr for intervention/drug/treatment searches, filter.overallStatus as a comma-separated status filter like RECRUITING, and countTotal=true to get a totalCount in the response on the first request.

- Example syntax given: 
`curl "https://clinicaltrials.gov/api/v2/studies?query.intr=semaglutide&filter.overallStatus=RECRUITING&pageSize=20&countTotal=true"`
 (GitHub cheatsheet, noble-ronin/clinicaltrials-api)
- 
The v2 base URL is https://clinicaltrials.gov/api/v2/, pagination uses pageToken (cursor-based) rather than offset, and filters use query.cond, query.intr, filter.overallStatus, filter.phase, and query.spons.

- 
Accepted overallStatus values include RECRUITING, COMPLETED, ACTIVE_NOT_RECRUITING, NOT_YET_RECRUITING, ENROLLING_BY_INVITATION, SUSPENDED, TERMINATED, WITHDRAWN, UNKNOWN, with multiple values pipe-separated.


**No actual count number** for "LLM-based Chatbot" interventions specifically with RECRUITING status was returned in search results — this would require directly querying the live API endpoint, which returns real-time data not indexed in search results. Documentation sour

You are an elite superforecaster. Estimate the probability of each option for this Metaculus multiple-choice question.

## Question
How many clinicaltrials.gov studies with an LLM-based Chatbot as the intervention/treatment will be recruiting on November 1, 2026?

## Description / Resolution Criteria
## Description
IEEE Spectrum: [Can AI Chatbots Reason Like Doctors?](https://spectrum.ieee.org/ai-clinical-decision-support)

`{"format": "bot_tournament_question", "info": {"hash_id": "aee13c93ea76b1e9", "sheet_id": "8"}}`

## Resolution Criteria
This question resolves as the number of studies with an Intervention/treatment of *LLM-based Chatbot* that have a study status of Recruiting are listed on November 1, 2026 at this link: https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec

## Options
  - 0 or 1
  - 2
  - 3
  - 4
  - >4

## Research tool outputs (raw; the resolution-source read is the status-quo anchor)
## Tool outputs

### resolution_source (SUCCESS)
RESOLUTION SOURCE (the page the question resolves on — treat as the status-quo anchor):
  Current state: The fetched clinicaltrials.gov search page returned no actual study listing data—only generic navigation/glossary text—so the current count of Recruiting studies with intervention 'LLM-based Chatbot' cannot be determined from this fetch.
  Last updated: unknown | already resolved: unclear | page matches criteria: True
  Note: The clinicaltrials.gov page content was not actually loaded/rendered in the fetch (likely JS-driven search results not captured); only boilerplate text appeared. The IEEE Spectrum article is background context and does not contain the study count needed for resolution. The question resolves based on the live count on Nov 1, 2026, which is a future date relative to this fetch, so the outcome cannot yet be known.
  Fetched: https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec (HTTP 200)
  Fetched: https://spectrum.ieee.org/ai-clinical-decision-support (HTTP 200)

### web_search (SUCCESS)
Web search summary (20 results):
## Summary of Findings

### Query 1: "clinicaltrials.gov LLM-based chatbot recruiting studies count"

No single aggregate "official count" of LLM-based chatbot trials with Recruiting status was found, but several relevant data points emerged:

- **Systematic review (PLOS Digital Health)**: 
A review of clinical trial registries uncovered 22 registered trials currently exploring the applications of LLMs in healthcare, registered between 2023 and 2024, in various stages, with 9 actively recruiting participants (41% of the total), 7 have not yet started, and the remaining 6 are either ongoing or nearing completion.
 (Source: journals.plos.org/digitalhealth, article dated ~Feb 2024 report references)

- **Individual trial examples found on ClinicalTrials.gov**:
  - NCT06943911 ("BC-CHAT") — 
a breast cancer AI chatbot study, "Not yet recruiting," with 122 breast cancer patients recruited from Kafr El-Sheikh University Hospital, randomly assigned to intervention and control groups.

  - NCT07250516 — a China-based study on 
"LLM-based AI Dialogue Bot"
 with 
an enrollment count of 6,360 (actual).

  - NCT07860762 ("CHAT-HX") — 
"Augmenting Clinical Workflow With a Structured Chatbot for Patient History Collection and Diagnostic Support," sponsored by Massachusetts General Hospital, first submitted 2026-09-30.

  - NCT06880315 (Stanford) — 
"LLM-Generated Coaching Prompts," status "Not Yet Recruiting," with an estimated start date of 2027-01-20.


**Date range**: Sources span 2023–2026 (most recent trial records from September 2026).

---

### Query 2: "clinicaltrials.gov API query.intr='LLM-based Chatbot' filter.overallStatus=RECRUITING countTotal"

No direct result returning an actual numeric count for this specific query was found. However, extensive technical documentation on how this API query works was found:

- The ClinicalTrials.gov v2 API supports 
query.intr for intervention/drug/treatment searches, filter.overallStatus as a comma-separated status filter like RECRUITING, and countTotal=true to get a totalCount in the response on the first request.

- Example syntax given: 
`curl "https://clinicaltrials.gov/api/v2/studies?query.intr=semaglutide&filter.overallStatus=RECRUITING&pageSize=20&countTotal=true"`
 (GitHub cheatsheet, noble-ronin/clinicaltrials-api)
- 
The v2 base URL is https://clinicaltrials.gov/api/v2/, pagination uses pageToken (cursor-based) rather than offset, and filters use query.cond, query.intr, filter.overallStatus, filter.phase, and query.spons.

- 
Accepted overallStatus values include RECRUITING, COMPLETED, ACTIVE_NOT_RECRUITING, NOT_YET_RECRUITING, ENROLLING_BY_INVITATION, SUSPENDED, TERMINATED, WITHDRAWN, UNKNOWN, with multiple values pipe-separated.


**No actual count number** for "LLM-based Chatbot" interventions specifically with RECRUITING status was returned in search results — this would require directly querying the live API endpoint, which returns real-time data not indexed in search results. Documentation sour

## Question dates and conventions
Dates: opened 2026-10-09; forecasting closes 2026-10-09; resolves 2026-11-01. Only events AFTER the open date can satisfy the criteria; assume the event has NOT happened unless a source shows it has. A question phrased 'before <date>' or 'by <date>' asks whether it happens between now and that date. The deadline for the outcome is the RESOLUTION date, not the close date.

## Recent news (AskNews, raw article summaries, newest first — read these directly)
14 AskNews articles (newest first; 6 from the last 48 h, 10 from the archive). Summaries are AskNews's own; dates are publication dates.

[2026-10-09] BIDMC and Google Conduct First Real-World Safety Study of Patient-Facing AI in Primary Care
newswise — https://www.newswise.com/articles/bidmc-researchers-conduct-first-real-world-study-of-safety-and-quality-of-patient-facing-ai-in-primary-care
  - Researchers at Beth Israel Deaconess Medical Center (BIDMC), in collaboration with Google, conducted the first prospective real-world study of the Articulate Medical Intelligence Explorer (AMIE), a patient-facing AI chatbot, in primary care settings.
  - Conducted between April and November 2025, the study enrolled 114 patients, with 98 completing both the AI interaction and their primary care appointment.
  - Every conversation was monitored in real time by a board-certified internal medicine physician who could intervene if a patient appeared at risk of harm.
  - Across all 98 completed encounters, none required a safety-stop intervention; physicians identified one hallucination and provided additional clinical clarification in five cases.
  - Patients rated the quality of AMIE's conversations favorably, and their attitudes toward AI improved after interacting with the chatbot.
  - However, concerns remained regarding trust in the confidentiality of information and the chatbot's honesty.
  - Primary care providers reviewed AI-generated summaries for 44 cases; in 75% of these cases, physicians said the summary helped them prepare, and in 57% it may have influenced their clinical approach.
  - Only one provider reported the interaction as somewhat harmful, citing patient anxiety over a possible lymphoma diagnosis.
  - The study, published in The Lancet, was designed to evaluate feasibility and safety rather than health outcomes.
  - The research was funded by Alphabet, and Adam Rodman, MD, served as a visiting researcher at Google during part of the study.

[2026-10-08] AI Chatbots Show Promise in Pre-GP Diagnosis but Require Oversight, Study Finds
The Independent — https://www.independent.co.uk/news/uk/home-news/ai-chatbot-patients-diagnosis-gp-b3063668.html
  - A small study conducted in the US involving 98 patients at a GP clinic in Boston suggests that AI chatbots could assist in diagnosing patients days before their scheduled GP visits.
  - Published in The Lancet on October 08 2026, the research found that the AI's suggested diagnoses matched confirmed diagnoses in more than half of cases, with an accuracy rate of 75 per cent when considering the chatbot's top three suggestions.
  - Doctors monitored conversations in real time and received transcripts prior to appointments.
  - In 75% of cases reviewed via a questionnaire, medics stated the AI chatbot helped them prepare for patient visits, and it may have changed their approach for more than half of cases.
  - Patient attitudes towards AI in primary care improved after interacting with the chatbot.
  - However, the study noted safety concerns: one conversation was rated 'somewhat harmful' by a doctor after the chatbot listed a possible cancer diagnosis which caused patient anxiety, and in another instance, the chatbot hallucinated, adding clinical information in five interactions.
  - Researchers stress that while the study shows initial feasibility, the evidence 'does not yet support autonomous use for diagnosis, triage, or management.' Professor Victoria Tzortziou Brown, president of the Royal College of GPs, emphasized that the study was small, conducted in a single US practice, and relied on close doctor supervision, meaning it cannot determine safety or effectiveness without that level of oversight.
  - She noted that GPs bring clinical judgment and history knowledge that AI should support, not replace, and called for larger, independent studies to examine safety, inequalities, workload, and patient outcomes before introduction into NHS care.

[2026-10-08] Clinical Trial Shows AI Chatbot AMIE Assists Doctors in 75% of Cases
The New York Times — https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
  - A clinical trial conducted over eight months at Beth Israel Deaconess Medical Center in Boston tested a specialized Google AI chatbot named AMIE (Articulate Medical Intelligence Explorer) to determine if it could assist primary care physicians without providing misleading information.
  - Published on Thursday in the medical journal The Lancet, the study found that doctors found the bot helpful in approximately 75 percent of patient cases, with the technology influencing how they handled more than half of those cases.
  - Peter Brodeur, a resident physician at the hospital, described the bot's performance as comparable to 'a high-performing medical student or resident.' The chatbot, powered by Google's Gemini technology, explored symptoms and generated summaries for doctors to review before patient visits.
  - The trial monitored hundreds of conversations and found the bot 'hallucinated' (made up information) only once, when it misrepresented a date.
  - Adam Rodman, an internist and medical historian at Beth Israel, noted that patients only trusted the bot because it operated alongside trained physicians, stating, 'None of our patients trusted it until they realized that it was handing-off their doctor.' During the trial, physicians monitored the chatbot and intervened to provide context on five occasions, including one instance where the bot listed cancer as a potential diagnosis.
  - Robert Califf, a cardiologist at Duke and former FDA commissioner, commented that while the technology represents the 'future,' it remains unclear if it will reduce the load on doctors or unearth new problems.

[2026-10-07] BERT-Based Medical Chatbot with Application-Level Cryptographic Protection Achieves 99% Accuracy
Nature — https://www.nature.com/articles/s41598-026-72040-w
  - Published on October 07, 2026, in the journal Nature, researchers proposed a BERT-based medical chatbot architecture that integrates contextual natural language processing with an application-level cryptographic protection layer for client-server communication.
  - The study utilized the HealthFC and TREC-Health datasets, implementing the NLP component via the Hugging Face Transformers library with TensorFlow and spaCy.
  - For security, the system employs RSA for AES secret key distribution, AES for encrypting the medical payload, and SHA-256-based HMAC for message integrity and authentication.
  - The model achieved 99% accuracy, 99% precision, 97% recall, 98% F1-score, and 97% AUC-ROC.
  - In a five-fold stratified cross-validation, accuracy values were recorded as 1.00, 0.99, 0.99, 0.98, and 0.99.
  - These results outperformed baseline models including LSTM (0.88 accuracy), SVM (0.91 accuracy), and Bi-LSTM (0.94 accuracy).
  - The authors clarify that this cryptographic mechanism serves as an additional layer complementing standard TLS, rather than replacing it, and does not protect plaintext data during server-side inference.
  - The study notes that clinical safety, external validation, paraphrase-aware leakage analysis, end-to-end latency, and multilingual or voice-based capabilities require further experimental evaluation.

[2026-10-07] CONIFAS2 Study Protocol: Feasibility of Nature-Based Intervention for Children with ADHD
PLOS — https://journals.plos.org/plosone/article
  - The CONIFAS2 study, a two-arm feasibility randomised controlled trial (RCT) protocol published by PLOS on October 07, 2026, aims to assess the feasibility and acceptability of a co-produced, nature-based intervention for children aged 6-11 with ADHD.
  - The study is being conducted across children's neurodiversity ADHD services, initially within the Humber Teaching NHS Foundation Trust (HTFT) and Navigo Community Interest Company (CiC) in North East Lincolnshire, with phased inclusion of Cornwall Partnership NHS Foundation Trust and Leeds Community Healthcare NHS Trust.
  - The intervention, delivered over a 6-week period, consists of a nature-based activity box (designed for at least ten minutes of daily nature exposure), psychoeducational resources, and fortnightly telephone support (three calls of approximately 30 minutes) from a neurodiversity ADHD service clinician, alongside an optional online community forum moderated by a Kids charity organisation.
  - The study will compare the intervention plus treatment as usual (TAU) to TAU alone, with a proposed sample size of 84 participants (42 per arm) to ensure 35 evaluable participants per group, accounting for an anticipated 15% attrition rate.
  - As a feasibility trial, it is not powered to formally detect clinical effectiveness but aims to provide data on recruitment, retention, and intervention fidelity to inform a future large-scale effectiveness trial.
  - The study collects health economic data to assess the potential social and economic impact of the intervention.
  - Primary outcome measures include the Child Health Utility 9D (CHU9D) index, while secondary measures include the SNAP-IV 26-item scale, RCADS-P, and the NCI.
  - The protocol adheres to SPIRIT 2025 reporting guidelines and will be reported according to the CONSORT 2010 statement extension for pilot and feasibility trials.

[2026-10-06] German Study Tests AI Chatbots for Rheumatism Patient Queries
الوطن — https://www.alwatan.com.sa/article/1186817
  - A German study published in the 'Journal of Medical Systems' tested 10 AI-powered chatbots designed to answer questions from rheumatism patients, utilizing 6,291 real patient inquiries collected between September 2025 and January 2026.
  - The chatbots, which were developed based on German clinical guidelines and deployed across 13 rheumatism centers and 6 patient organizations, addressed topics where questions about the disease itself accounted for 50.2%, followed by medications and follow-up at 37.3%, and diagnosis at 28.2%.
  - While 92.9% of 2,671 user-rated responses received positive feedback and 80.2% of participants viewed the robots as a useful addition to patient education, the study found that only 45% of the answers fully adhered to medical guidelines.
  - Researchers emphasize the need to verify accuracy and safety before expanding the technology's use in healthcare, noting that insufficient detail was the most common reason for rejecting answers, and 4.2% of interactions yielded no response.

[2026-09-22] AI Chatbot 'Suzy' Implements Multi-Tier Safety Guardrails and Human Monitoring for Suicide Risk in Opioid Treatment Clinics
Nature — https://www.nature.com/articles/s41746-026-03288-9
  - Published on September 22, 2026, in Nature's npj Digital Medicine, a study by researchers at a large urban general hospital in Massachusetts describes the implementation of 'Suzy,' an AI chatbot designed to monitor suicide risk and respond to patient distress in high-risk addiction medicine settings.
  - The chatbot, built using Open Chat Studio (Dimagi, Inc) and OpenAI LLMs (initially GPT 4o, then 5.0), serves adults receiving medication treatment for opioid use disorder (OUD).
  - The system employs a 'Safety Router' to classify user inputs into three risk tiers: Tier 1 (clear suicide/self-harm risk), Tier 2 (ambiguous distress), and Tier 3 (no risk).
  - When Tier 1 is triggered, the chatbot outputs an urgent, supportive safety disclaimer directing users to emergency services like 911 or 988, while Tier 2 messages are sent to a 'Disambiguate Node' to gather more context.
  - A critical component of the safety framework is a human-monitoring layer where a trained bachelor-level staff member reviews chat transcripts twice daily on weekdays, using the Columbia-Suicide Severity Rating scale (C-SSRS) to assess risk; if imminent risk is found, a licensed psychologist joins the call for triage.
  - The authors, including co-authors JS, LG, DS, and SC, emphasize that while AI guardrails are necessary, they are insufficient without human oversight, recommending that all behavioral health chatbots incorporate rigorous safety architectures and real-time human triage.
  - The study, which was IRB-approved with written informed consent from all patients, notes that patient data was not retained via a zero data retention agreement with OpenAI to ensure HIPAA compliance.
  - The researchers conclude that a multi-pronged approach involving safety guardrails, human triage, ongoing expert testing, and hybrid AI architectures is essential for scaling AI support in high-risk clinical environments.

[2026-08-20] Scoping Review Identifies Mental Health Harms of LLM-Based Chatbots (analysis)
Nature — https://www.nature.com/articles/s41746-026-03054-x
  - A scoping review published in npj Digital Medicine on August 20, 2026, synthesizes research on the mental health harms of Large Language Model (LLM)-based chatbots.
  - The authors conducted a PRISMA-based systematic literature search across five databases (ACM, IEEE, PubMed, Science.gov, Google Scholar) and a supplementary search, identifying N=3,137 articles.
  - Of these, N=119 articles were included that focused on LLM-based chatbots and their potential harms to mental health.
  - The literature was categorized into five areas: (1) Conceptual works identifying harms from chatbot limitations such as hallucinations, sycophancy, and bias, alongside data security risks and dangers for high-risk psychiatric cases like suicide and psychosis; (2) Vignette studies showing chatbots respond inappropriately to mental health queries compared to clinical standards; (3) Cognitive overreliance linked to decreased cognitive and academic performance; (4) Problematic use characterized by emotional or social dependency and withdrawal, which correlates with mental health symptoms; and (5) Articles on AI psychosis suggesting chatbots may reinforce and validate delusional beliefs.
  - The review concludes that both theoretical and empirical work identify various hypothetical and observed harms associated with LLM chatbot use in mental health contexts.

[2026-07-27] Protocol for Randomized Controlled Trial on Chatbot Intervention for Adolescent Digital Addiction
Springer — https://link.springer.com/article/10.1186/s12888-026-08375-5
  - The study aims to develop and evaluate a low-intensity, secondary intervention using a Stage of Change (SoC) and Motivation Interviewing (MI)-based Chatbot to reduce digital addictions (Internet Gaming Disorder [IGD] and Social Media Addiction [SMA]), sedentary lifestyles, and mental distress among adolescents.
  - The research is a two-arm randomized controlled trial (RCT) involving an expected 330 participants with clinical digital addiction symptoms, comparing a five-week Chatbot-assisted intervention group against a psycho-educational control group, both delivered via WhatsApp.
  - The primary outcome is the composite score of IGD and SMA, with secondary outcomes including quality of life, psychological distress, and eHealth literacy, while perceived autonomy and competence are examined as mediators.
  - Evaluations are scheduled at baseline (T0), post-program (T1), and a 3-month follow-up (T2).
  - The trial, pre-registered at ClinicalTrials.gov (NCT06821373) on February 09, 2025, seeks to determine if digital interventions offer a more cost-effective and scalable alternative to traditional methods, particularly in resource-limited settings.

[2026-07-21] Study Reveals High Attrition in LMIC Chatbot Health Interventions Due to Structural and Platform Barriers (analysis)
Nature — https://www.nature.com/articles/s44482-026-00027-5
  - A mixed-methods study published in npj Digital Public Health (Source: Nature, July 21, 2026) evaluated the 'ParentText' chatbot intervention across three Low- and Middle-Income Countries (LMICs): Malaysia, Jamaica, and the Philippines.
  - The study found that while enrolment figures were high, 'effective access' was low, with 93.8% of participants completing less than 25% of the content and a median engagement of only two days.
  - Despite the programme being designed for 38 days, most participants disengaged within the first week.
  - Retention rates by Day 7 varied significantly by site: 51.3% in Malaysia, 18.9% in Jamaica, and 13.3% in the Philippines.
  - Mean engagement duration was 11.18 days in Malaysia versus 4.69 in Jamaica and 3.45 in the Philippines.
  - Qualitative data indicated that caregivers valued the content but disengaged due to structural barriers, including message frequency, length, and platform automation policies that restricted re-contacting users who missed a response window.
  - To address these issues, a consensus panel identified 29 redesign optimisations, with 69% rated as highly feasible.
  - Key recommendations include reducing message burden, shifting to progress-based delivery, and treating re-engagement as a core feature.
  - Selected optimisations were subsequently tested in randomised trials in South Africa and Malaysia, with plans to extend to Jamaica and the Philippines contingent on funding.

[2026-06-25] HopeBot: LLM-Powered Chatbot for Depression Screening Evaluated in UK and China
PLOS — https://journals.plos.org/digitalhealth/article
  - The World Health Organisation (WHO) estimates that depression affects approximately 3.8% of the global population, with 12-month prevalence rates around 4-5% in the United Kingdom (UK) and 3-4% in China, yet only about half of affected individuals receive adequate treatment.
  - To address barriers such as the need for trained professionals and societal stigma, researchers developed HopeBot, an LLM-powered (GPT-4o) conversational chatbot designed to administer the Patient Health Questionnaire-9 (PHQ-9) screening tool.
  - Unlike previous systems that relied on rigid rule-based intents, HopeBot utilizes retrieval-augmented generation (RAG) to maintain the validated structure and scoring logic of the PHQ-9 while allowing for flexible, empathic dialogue and real-time clarification of ambiguous user inputs.
  - A prospective mixed-methods study conducted by the authors with 132 participants from diverse educational and cultural backgrounds in the UK and China evaluated the system's feasibility and score-level concordance against self-administered PHQ-9 assessments.
  - The study focused on format-level validation rather than diagnostic validity, as establishing diagnostic accuracy would require comparison with a clinician-administered reference.
  - The research aims to provide a scalable, user-friendly method for early depression detection that preserves psychometric integrity while enhancing procedural efficiency.

[… older articles omitted]

## Research reports (one per dimension: web search + page reads, dated, with sources)
### CURRENT STATE
- [2026-10-09] Attempts to directly load the resolution-source URL (clinicaltrials.gov/search?...intr=LLM-based%20Chatbot&aggFilters=status:rec) and the ClinicalTrials.gov API returned no accessible content in this research session, so the live, current count of "Recruiting" studies tagged with the exact intervention/treatment name "LLM-based Chatbot" could not be directly verified today. (https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec)
- [2026-10-09] Individual study records that use this specific intervention tag show mixed, mostly non-recruiting statuses: NCT06914492 ("Quit to Win" 2025, Hong Kong) is 
Real-time Smoking Cessation Instant Messaging Support Using a Large Language Model (LLM)-Based Chatbot Via "Quit to Win" 2025 (QTW2025)
 and has an Overall Status of COMPLETED as of its 2025-05 verification. (https://clinicaltrials.gov/study/NCT06914492)
- [2026-06] The successor trial NCT07649044, 
Promoting Smoking Cessation in the Community Via "Quit to Win" Contest 2026: Engagement-Focused GenAI-based Chatbot for Real-Time Smoking Cessation Support
, is listed as NOT_YET_RECRUITING with an estimated start date of 2026-06-27, meaning it could plausibly shift to Recruiting before November 1, 2026. (https://clinicaltrials.gov/study/NCT07649044)
- [2026-04-14] NCT07280650, 
Feasibility and Acceptability of an LLM-based Chatbot for Family Caregivers: Evaluation Study
 (University of Michigan), is NOT_YET_RECRUITING with an estimated start date of 2026-08, so it may have moved to Recruiting by late 2026. (https://clinicaltrials.gov/study/NCT07280650)
- [2025-10] NCT07214831, a Massachusetts General Hospital study of 
A Feasibility and Acceptability Study of a Large Language Model-based Chatbot for Brief Alcohol Intervention Among Emerging Adults
, is NOT_YET_RECRUITING with an estimated start date of 2027-06-01, so it will almost certainly remain not-yet-recruiting through November 1, 2026. (https://clinicaltrials.gov/study/NCT07214831)
- [2026] NCT07741266, 
a pilot randomized controlled trial to assess the feasibility and preliminary effectiveness of an LLM-driven chatbot versus electronic brochure in addressing reproductive concerns among AYA cancer survivors
, is a newly registered study (high NCT number, 2026 cohort); its current recruiting status could not be confirmed in this session. (https://clinicaltrials.gov/study/NCT07741266)
- [2026] NCT07410689, SAFE.AI, describes a 
hybrid SAFE.ai chatbot that builds on an existing rule-based system by integrating a large language model (LLM) layer to support more flexible, open-ended, and conversational interactions
; its exact intervention tag and current recruiting status were not confirmed. (https://clinicaltrials.gov/study/NCT07410689)
- [2026-10-08/09] Independent of the registry count, media coverage this week (NYT, The Independent, Newswise/BIDMC) confirms heavy real-world momentum behind LLM-chatbot clinical research, with the BIDMC/Google AMIE trial — a prospective, physician-supervised, patient-facing chatbot study — 
conducted between April and November 2025, the study enrolled 114 patients, with 98 completing both the AI interaction and their primary care appointment
, illustrating the broader surge of similar interventional registrations feeding into the ClinicalTrials.gov "LLM-based Chatbot" tag, even though that specific trial itself was not found tagged with that exact intervention term. (https://www.newswise.com/articles/bidmc-researchers-conduct-first-real-world-study-of-safety-and-quality-of-patient-facing-ai-in-primary-care)
- [2026-09-01] A related mixed-methods evaluation of an LLM-based behavioral activation chatbot for youth depression was published, underscoring that academic/clinical interest in LLM chatbot interventions is accelerating, which is consistent with new ClinicalTrials.gov registrations appearing throughout 2026. (https://pmc.ncbi.nlm.nih.gov/articles/PMC13533316/)
- [2025, 206 days old] A JMIR methodological review protocol notes that 
since ChatGPT's launch in November 2022, interest in patient-facing conversational agents whose replies are generated by large language models (LLM), in other words, LLM-based chatbots, has surged, with growing attention to their potential clinical applications
, and that 
LLM-based chatbots are reported being adopted at scale by consumers and patients
, supporting a trend of more such trials being registered over time, though this does not itself report a current recruiting count. (https://doi.org/10.2196/90507)

BOTTOM LINE: The exact ClinicalTrials.gov registry page used for resolution could not be fetched directly in this session, but individual records tagged with the "LLM-based Chatbot" intervention show a small, mixed pool (several NOT_YET_RECRUITING or COMPLETED, a few of unconfirmed status), while the broader research field is clearly expanding rapidly (new 2026 trial registrations, high-profile Lancet/BIDMC-Google publication this week); this suggests the Recruiting count by November 1, 2026 is likely modest but could tick upward as 2026-started studies (e.g., NCT07649044, NCT07280650) transition out of "not yet recruiting."

NOT FOUND: The live, current total number of studies shown as "Recruiting" on the exact clinicaltrials.gov search/filter URL specified in the resolution criteria; confirmed current recruiting status for NCT07741266 and NCT07410689; any official ClinicalTrials.gov aggregate count or "results found" figure for the intr=LLM-based Chatbot + status:rec filter.

### BASE RATE
- [2026-10] Direct access to the clinicaltrials.gov live tracker page (the exact URL used for resolution) and its API endpoint were not fetchable in this session; findings below are inferred from individual study records returned by search, not from a single authoritative count — https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec
- [2026-10] Multiple recently-registered trials using "LLM-based Chatbot" or closely related intervention names are currently listed as NOT_YET_RECRUITING rather than Recruiting, including a University of Michigan caregiver chatbot study 
whose status as of 2026-04 was "NOT_YET_RECRUITING"
 — https://clinicaltrials.gov/study/NCT07280650
- [2026-10] An MD Anderson informed-consent chatbot trial posted in August 2026 is also not yet recruiting, 
sponsored by M.D. Anderson Cancer Center
 with an estimated 2027-01-01 start date, showing new registrations continue to lag actual recruitment — https://clinicaltrials.gov/study/NCT07756515
- [2026-10] A Massachusetts General Hospital alcohol-intervention chatbot trial (NCT07214831) and a Hong Kong "Quit to Win" 2026 smoking-cessation GenAI chatbot trial (NCT07649044) were both listed as NOT_YET_RECRUITING as of mid-2026, reinforcing that many registered LLM-chatbot studies have not yet reached active recruitment status — https://clinicaltrials.gov/study/NCT07214831, https://clinicaltrials.gov/study/NCT07649044
- [2026-10] A Stanford "LLM-Generated Coaching Prompts" physical-activity study was updated as recently as July 2026 and remains 
"NOT_YET_RECRUITING"
 with an estimated January 2027 start — https://clinicaltrials.gov/study/NCT06880315
- [2026-02-ish] An academic review protocol explicitly framed LLM-based chatbot interventional trials as a newly emerging and rapidly expanding category, noting 
"interest in patient-facing conversational agents whose replies are generated by large language models...has surged, with growing attention to their potential clinical applications"
 since ChatGPT's 2022 launch — https://doi.org/10.2196/90507
- [2026-04-23] A broader bibliometric analysis of AI in clinical trials registry data found 
5,828 AI-related clinicaltrials.gov records as of a search conducted 23 April 2026, representing slightly over 1% of all ClinicalTrials.gov records
, and specifically noted that 
"chatbot" has risen since 2018, and "ChatGPT"/"GPT" has substantially risen since 2022
 — https://arxiv.org/pdf/2605.29096
- [2026-10-08] Independent of the registry count, the field received major news momentum this week via the AMIE/BIDMC/Google Lancet publication, which could spur sponsors to register new similarly-labeled LLM chatbot trials, but this is a lagging indicator (new registrations take weeks-months) rather than something that would change the November 1 count — https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
- [2026-10] A handful of trials do appear to reach Recruiting status under similar labels, e.g. an intelligent-chatbot tobacco-support pilot RCT and a breast-cancer chatbot knowledge study used "Recruiting"/"Not yet recruiting" terminology consistent with the registry's standard lifecycle, but exact current counts tagged specifically "LLM-based Chatbot" (the precise intervention-name string used by the resolution link) could not be enumerated from search snippets alone — https://clinicaltrials.gov/study/NCT06952725, https://clinicaltrials.gov/study/NCT06943911
- [general] No historical analog exists for "how many times this exact clinicaltrials.gov filtered count hit a particular number," since this is a narrow, newly-coined intervention-name filter (LLM-based Chatbot) rather than a recurring contest with past outcomes; thus no favourite/odds/poll base rate exists for this specific metric.

BOTTOM LINE: There is no direct "contest" base rate for this metric since it's a live registry count rather than a historical recurring event; the available evidence shows the "LLM-based Chatbot" intervention category is small but growing, with many newly registered trials sitting in Not-Yet-Recruiting status for months before transitioning to Recruiting, suggesting the November 1, 2026 recruiting count will likely remain in a modest range rather than show a dramatic jump, modestly favoring continuity/slow-growth outcomes over large discrete jumps.

NOT FOUND: The exact current live count of studies at the resolution URL (https://clinicaltrials.gov/search?viewType=Card&intr=LLM-based%20Chatbot&aggFilters=status:rec) could not be retrieved, as both the search-page URL and the clinicaltrials.gov API endpoint were blocked from direct fetching in this session; no historical snapshot/trend chart of this specific filtered count over time was found; no existing third-party forecasts, polls, or odds on this specific question were found.

### KEY FACTORS
- [2026-10-06] A brand-new MGH study, CHAT-HX (NCT07860762), was first posted on 2026-10-06 with intervention listed as an "AI-Based Structured History-Taking Chatbot" described as 
"A large language model-based agentic chatbot that conducts a structured, pre-visit history interview on a study tablet and generates a history summary, differential diagnosis list, and workup suggestions"
 and is currently Not yet recruiting, illustrating that new LLM-chatbot trials are still entering the registry pipeline literally days before the question's close (https://clinicaltrials.gov/study/NCT07860762).
- [2026-09-25] Another freshly registered trial, PersianAI-CBT (NCT07840469, University of Isfahan), confirms continued global registration momentum, with a large language model supporting 
"the conversational functionality of the chatbot"
 while it remains Not yet recruiting as of its last update (https://clinicaltrials.gov/study/NCT07840469).
- [2026-03-31] NCT07504367 ("Large Language Models Assist in Tumor MDT," Sun Yat-Sen Memorial Hospital) is one of the few comparable LLM-intervention trials already showing Overall Status RECRUITING, with an actual start date of 2026-01-01 and estimated completion 2026-12-31, demonstrating that at least some trials in this space do successfully convert to and remain in Recruiting status through the resolution window (https://clinicaltrials.gov/study/NCT07504367).
- [2026-04-30] NCT06949462 (PEAR, Singapore General Hospital, LLM for anaesthesia consent) was also marked RECRUITING as of its last update, but its actual/estimated completion date of 2026-04-27 has already passed, meaning it may have since transitioned to Completed by November 2026 — a reminder that the recruiting pool is a moving target with inflows and outflows (https://clinicaltrials.gov/study/NCT06949462).
- [2026-07-23] The Stanford "LLM-Generated Coaching Prompts" study pushed its estimated start date back to 2027-01-20 in a July 2026 update, confirming that sponsor-side delays (not registry mechanics) are a key factor keeping some pipeline trials out of Recruiting status past November 1 (https://clinicaltrials.gov/study/NCT06880315).
- [2026-07] University of Michigan Rogel Cancer Center's AI supportive-care chatbot trial (NCT07772596) lists an estimated Start Date of 2026-10-01, among the closest upcoming start dates found, making it a plausible candidate to flip to Recruiting by November 1 if the sponsor updates status on schedule (fetched record, search result 15).
- [2026-08-13] Brigham and Women's Hospital's prostate-RT patient-education chatbot (NCT07719062) has an estimated Start Date of 2026-08, already in the past relative to today, meaning the sponsor-controlled step of updating status from Not yet recruiting to Recruiting is overdue and could occur any time before November 1 (https://clinicaltrials.gov/study/NCT07719062 via search result 13).
- [2026-08] ChatACP (NCT07448649, University of Hong Kong) explicitly states per its JMIR protocol that 
"This project was recommended for support in September 2025... Recruitment and data collection will commence in September 2026"
, indicating the sponsor's own timeline places its transition to Recruiting squarely within the question's window (https://pubmed.ncbi.nlm.nih.gov/42585665/).
- [2026-06] A CABG complication-prediction LLM chatbot trial (NCT07432893) states 
"Phase II usability evaluation is projected from May to July 2026. Phase III recruitment is anticipated from August 2026 to September 2027"
, another sponsor-stated timeline implying a possible shift into Recruiting before the resolution date (https://doi.org/10.2196/103717).
- [2026-06-11] Tsinghua University's "Large Language Model in Outpatient Care" trial (NCT07641478) lists an estimated Start Date of 2026-06-12 and estimated Completion Date of 2026-09-04 — both already in the past relative to today — meaning this trial has likely already moved through and possibly out of Recruiting status before the resolution date, a mechanism by which the count can also shrink (https://clinicaltrials.gov/study/NCT07641478).

BOTTOM LINE: The key drivers of the November 1 count are (1) sponsor-controlled status updates — several studies with 2026 Q3/Q4 estimated start dates (CHAT-HX, Rogel Cancer Center, BWH prostate education, ChatACP, Tumor MDT) sit right at the threshold of flipping from Not-yet-recruiting to Recruiting, and (2) continued weekly new registrations (two more added just in September-October 2026), both pushing the count up, partially offset by trials whose completion dates have already passed (PEAR, Tsinghua) that may exit Recruiting before the deadline.

NOT FOUND: The live resolution-URL page and ClinicalTrials.gov API count could not be directly fetched (domain access blocked); no authoritative current total of studies matching the exact "LLM-based Chatbot" intervention term with status Recruiting was obtainable, so the analysis relies on individual study records rather than the aggregate search count.

### EXPERT VIEW
- [2026-08-07] Kalshi and Polymarket have expanded into biotech/clinical-trial-related prediction markets, but 
Kalshi launched a suite of biotech prediction markets on July 16, letting users wager on Phase 3 clinical trial outcomes and full FDA approvals from major pharmaceutical companies
, not on registry metadata/intervention-tag counts. (https://cryptobriefing.com/kalshi-polymarket-drug-approval-betting/)
- [2026-05 onward] A niche clinical-trial-specific market, Endpoint Arena, has also launched, but 
Endpoint Arena is still in pilot mode and currently only allows... less than $50,000 is currently being wagered across all biotech-related contracts on Kalshi combined, and Endpoint Arena has yet to move beyond paper trading
 — i.e., even the specialized market infrastructure is thin and focused on drug-approval/endpoint outcomes, not trial-registry search counts. (https://www.fiercebiotech.com/biotech/betting-biotech-prediction-markets-set-sights-clinical-trials)
- No Metaculus, Manifold, Polymarket, or Kalshi question could be found that forecasts the number of ClinicalTrials.gov studies tagged with intervention "LLM-based Chatbot" and status "Recruiting." Searches for this exact metric and for related "bot tournament" forecasting questions returned no public market or community-forecast page.
- The only "tournament" context found is the metadata tag embedded in the question itself (`bot_tournament_question`), indicating this question was likely authored for an internal/private AI-forecasting-bot benchmarking exercise rather than posted to a public crowd-forecasting platform; no externally visible community prediction distribution exists for it.
- Academic/clinical commentary (not a market, but the closest thing to "expert opinion" available) is cautious rather than prescriptive about growth in registered chatbot trials. Coverage of the BIDMC/Google AMIE study (Oct 8, 2026, published in The Lancet) emphasizes that the evidence "does not yet support autonomous use for diagnosis, triage, or management," per the researchers, and UK RCGP president Victoria Tzortziou Brown called for larger, independent studies before wider clinical adoption — suggesting expert sentiment leans toward cautious, incremental expansion rather than a rapid surge in new recruiting trials. (https://www.independent.co.uk/news/uk/home-news/ai-chatbot-patients-diagnosis-gp-b3063668.html)
- Related academic surveys note a broader proliferation of LLM-chatbot studies registering on ClinicalTrials.gov across many clinical domains (oncology, diabetes transition care, depression screening, mental health, alcohol intervention, caregiver support), several already registered or recently posted (e.g., NCT07741266, NCT07280650, NCT06801925), indicating an active and growing pipeline feeding into the "Recruiting" status category, though with no single source aggregating or forecasting the exact future count. (https://clinicaltrials.gov/study/NCT07280650)

BOTTOM LINE: No prediction market, poll, or formal expert forecast exists for this specific ClinicalTrials.gov intervention-tag count; the closest available signals are indirect — active biotech prediction markets (Kalshi, Polymarket, Endpoint Arena) focus on trial outcomes/approvals rather than registry metadata, and clinical experts commenting on recent high-profile chatbot trials (e.g., AMIE/Lancet study) express caution about rapid clinical deployment, suggesting no strong external consensus pushes toward either a large increase or decrease in the recruiting-study count by November 1, 2026.

NOT FOUND: Any Polymarket, Kalshi, Manifold, or Metaculus market/question directly tracking this ClinicalTrials.gov intervention-filtered count; any published analyst projection or poll of the expected number of "Recruiting" LLM-based Chatbot studies; any public track record of forecasters on this specific or closely analogous registry-count questions.

### FAVORITE CASE
- [2026-04] The University of Michigan caregiver chatbot trial (NCT07280650) has an official intervention title "LLM-based Chatbot" and was listed with an estimated start of 2026-08, which has now passed, making it the single strongest near-term candidate to flip to Recruiting by Nov 1, 2026. 
The study is titled "Feasibility and Acceptability of an LLM-based Chatbot for Family Caregivers: Evaluation Study" and was registered with status Not Yet Recruiting as of April 2026
, with 
a start date estimated for 2026-08 and a status verified date of 2026-04
.
- [2026-08-03] A second strong candidate, NCT07741266 (LLM-driven chatbot for reproductive concerns in AYA cancer survivors), was first posted August 3, 2026 with 
status "Not yet recruiting"
 and 
an estimated study start of September 1, 2026
 — a date that has already passed as of the Nov 1 resolution point, making a shift to Recruiting plausible.
- [2026-06] NCT07649044 ("Quit to Win" 2026 GenAI chatbot, University of Hong Kong) remains the Hong Kong program's current entry, verified 
as "Promoting Smoking Cessation in the Community Via 'Quit to Win' Contest 2026: Engagement-Focused GenAI-based Chatbot for Real-Time Smoking Cessation Support"
 with 
status "NOT_YET_RECRUITING" and an estimated start date of 2026-06-27
 — its own target start date is already four months in the past relative to resolution, suggesting high likelihood it has transitioned to Recruiting (or already completed early phases) by Nov 1.
- [2026] The SAFE.AI hybrid chatbot study (NCT07410689) is a plausible marginal case: it uses a "Hybrid Chatbot" intervention with an LLM layer rather than the exact "LLM-based Chatbot" label, meaning it may or may not be captured by the precise intervention-filter string used in the resolution URL. 
The intervention is named "Hybrid Chatbot," described as building on the existing rule-based system by integrating a large language model (LLM) layer to support more flexible, open-ended, and conversational interactions
.
- [2026-07-28] The reproductive-health chatbot trial's registry history shows a typical ~1-month lag from first submission to first posting 
(first submitted July 28, 2026; first posted August 3, 2026)
, confirming the general pattern that new "LLM-based Chatbot" registrations continue to appear roughly monthly, feeding the pool of studies that could reach Recruiting status by November.
- [2026-10] Field-wide academic framing continues to describe the "LLM-based Chatbot" intervention class as a rapidly expanding category rather than a niche: 
"interest in patient-facing conversational agents whose replies are generated by large language models (LLM), in other words, LLM-based chatbots, has surged, with growing attention to their potential"
 clinical applications, supporting continued net registration growth into the category that underlies the resolution count.
- [2026] No trial using the exact "LLM-based Chatbot" intervention tag was found with a Recruiting status as of this search session; all directly-tagged records identified (NCT06914492, NCT07280650, NCT07741266, NCT07649044, NCT07214831) are either Completed or Not Yet Recruiting, indicating the live Recruiting count under the strict filter is currently small (plausibly zero or very low) rather than in a high range.
- [2026-10] Attempts to query the ClinicalTrials.gov v2 API directly (query.intr=LLM-based Chatbot, filter.overallStatus=RECRUITING) and to load the resolution's exact search URL were both blocked in this session (domain/API access restricted), so no authoritative live count could be independently confirmed beyond individual study record checks.

BOTTOM LINE: Among studies explicitly tagged with the intervention "LLM-based Chatbot," the only plausible near-term additions to a "Recruiting" count by Nov 1, 2026 are a small handful of 2026-registered trials (Michigan caregiver study, AYA cancer-survivor reproductive chatbot, HKU "Quit to Win" 2026) whose estimated start dates have already passed and could have flipped from Not-Yet-Recruiting to Recruiting; this favors a low-but-nonzero number (likely in the single digits) rather than a large count, consistent with the "modest range" leader identified in prior research.

NOT FOUND: The authoritative, current total shown on the exact resolution URL/API query (blocked from direct access); confirmed post-October 2026 status updates for NCT07280650, NCT07741266, or NCT07649044 showing an actual transition to Recruiting; any comprehensive list enumerating every study carrying the precise "LLM-based Chatbot" intervention tag.

### UPSET CASE
**Case for an upset: the true November 1, 2026 "Recruiting" count is higher than the modest/low number implied by the leading estimate.**

- [2026-06] NCT07649044 ("Quit to Win" Contest 2026, University of Hong Kong) had its 
Status Verified Date: 2026-06, Overall Status: NOT_YET_RECRUITING, Start Date: 2026-06-27, Type: ESTIMATED
 — meaning its own estimated start date has already passed as of this writing (Oct 9, 2026), so the record is stale and the study is very likely to have flipped to Recruiting well before November 1, strengthening the case for at least one more "Recruiting" hit than a snapshot taken months ago would suggest. https://clinicaltrials.gov/study/NCT07649044
- [2026-04] NCT07280650 (University of Michigan caregiver chatbot) listed an 
Overall Status: NOT_YET_RECRUITING ... Start Date: 2026-08, Type: ESTIMATED
 — that estimated start month (August 2026) has also already passed, meaning this record too is plausibly outdated and the study may already be actively recruiting by November 1, which would independently add to the count rather than keep it flat. https://clinicaltrials.gov/study/NCT07280650
- [2026, undated] NCT07741266, the AYA cancer-survivor reproductive-concerns LLM chatbot trial, is a freshly registered high-numbered NCT (2026 cohort) whose detailed protocol describes an active randomization and recruitment pipeline 
Simple random sampling with a fixed sample size will be employed to recruit subjects... Randomization will be automatically performed using a computer-generated randomization sequence embedded in the chatbot backend
 — this phrasing suggests the trial infrastructure is essentially recruitment-ready, not a years-out protocol, making a near-term flip to "Recruiting" plausible before Nov 1. https://clinicaltrials.gov/study/NCT07741266
- [2026-10-08] The Lancet/AMIE publication and its accompanying NYT, Independent, and Newswise coverage this week generated unusually high-profile validation that LLM chatbots can meaningfully assist primary-care physicians, with doctors finding the bot helpful in "approximately 75 percent of patient cases" — this kind of landmark positive result historically triggers rapid follow-on trial registrations and accelerated recruitment launches at peer institutions within weeks, which could pull additional studies into "Recruiting" status before the Nov 1 cutoff even though the lag is typically measured in months. https://www.nytimes.com/2026/10/08/technology/ai-chatbots-urgent-care.html
- [2026-10-07] A newly published Nature paper on a BERT/LLM-based medical chatbot with cryptographic protection shows the technical research pipeline for "medical chatbot" trial designs is still very active and diversifying (security-focused, diagnostic, behavioral-health variants), which expands the universe of candidate studies that could newly post or transition into the registry's "LLM-based Chatbot" intervention tag by Nov 1 — a mechanism the leading (likely static/low) estimate may underweight. https://www.nature.com/articles/s41598-026-72040-w
- [general, multiple records] Both NCT07649044 and NCT07280650 share a structural pattern seen repeatedly across this intervention tag: studies are registered with "Estimated" start dates that, once passed, are frequently updated to Recruiting with a lag of only a few weeks on ClinicalTrials.gov administrative updates — because at least two identified trials have already sailed past their listed estimated start dates without a confirmed status update being captured in this research session, the true current count is plausibly undercounted by official-looking "Not yet recruiting" labels that are simply stale, not accurate.

BOTTOM LINE: The strongest upset case is for a higher recruiting count than the leader's modest/low estimate: at least two specifically identified "LLM-based Chatbot"-tagged trials (NCT07649044, NCT07280650) have already passed their own estimated start dates as of Oct 9, 2026 and are likely to have flipped to Recruiting by Nov 1, while a newly registered 2026 trial (NCT07741266) appears recruitment-ready, and the Oct 8 Lancet/AMIE publicity wave could accelerate additional launches.

NOT FOUND: Confirmed, dated (post-June/April 2026) status updates for NCT07649044 and NCT07280650 showing an actual flip to Recruiting; any live fetch of the exact resolution-URL count (domain access was blocked in this session); official total enumeration of all currently Recruiting studies under this precise intervention-name tag.

## Required pre-forecast walkthrough

Before giving probabilities, address these explicitly in your rationale:
  (0) Resolution check: do the sources show the outcome already decided as of today? If so, which option and by which source.
  (a) The time left until the question resolves.
  (b) The status quo outcome — the option most likely if nothing changes from today.
  (c) An outcome that would be SURPRISING but plausible.

## Calibration guidance

- **Status quo anchoring.** Good forecasters put extra weight on the status quo option because the world changes slowly most of the time.
- **Leave moderate probability on most options.** Good forecasters don't drop options to ~0% — leave ≥3-5% on every option to account for unexpected outcomes, unless you have strong specific evidence ruling them out.
- Avoid uniform-distribution laziness — most options are not equally likely.
- **Prediction markets are a prior, not an anchor.** If the evidence lists a SAME QUESTION market for these options, treat its prices as a strong prior and name the specific evidence it lacks whenever you move away from it. A RELATED market is context only. If no market is listed, there is none — do not assume one.

## Output

Return ONLY valid JSON, no markdown fences:

{
  "rationale": "<address (a) (b) (c) above, then explain your distribution — 4-7 sentences total>",
  "probabilities": {
    "0 or 1": <prob>,
        "2": <prob>,
        "3": <prob>,
        "4": <prob>,
        ">4": <prob>
  }
}

Probabilities MUST sum to ~1.0 (renormalization is applied; close enough is fine).