Truace
Subscribe

LLM-assisted querying of ClinicalTrials.gov for psychedelic trials

Source article: AI-assisted characterization of psychedelic clinical trials from ClinicalTrials.gov

Abstract Background The registry record of psychedelic and related rapid-acting compounds is large, heterogeneous and laborious to synthesize by hand. We tested whether a large language model (LLM) with live access to the ClinicalTrials.gov application programming interface (API) could produce a timely, structured and reproducible characterization of this landscape. Methods Cross-sectional analysis of interventional trials on ClinicalTrials.gov as of 13 January 2026. The core-set comprised classic serotonin 5-HT…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1435· cite & verify
AI-assisted characterization of psychedelic clinical trials from ClinicalTrials.gov
Sanitary Trial Balance lantern slide (OHA 287), National Museum of Health and Medicine (4073709766) by National Museum of Health and Medicine. CC BY 2.0 · https://creativecommons.org/licenses/by/2.0
Trace impact readingContested
P 67The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 67The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Researchers connected the Claude language model to the ClinicalTrials.gov API and used it to survey interventional psychedelic trials registered as of 13 January 2026. After first trying open-ended natural-language queries, they switched to a documented protocol of field-specific query expressions, which produced every reported count.

The work suggests registries can be mapped quickly by an LLM, but only when each figure is traceable to an archived query. The authors state the method complements rather than replaces systematic review, and its accuracy against manual extraction has not yet been measured, so the counts are not validated against a human benchmark.

Main points

  1. Excluding withdrawn records, 412 core-set trials were found, 256 of them in patients, alongside 300 ketamine/esketamine trials.
  2. Psilocybin made up 244 trials (59%), MDMA 99, LSD 39, DMT/5-MeO-DMT 36, ayahuasca 6 and ibogaine 4.
  3. Of 63 phase 2/3 psilocybin depression trials, 35 used a depression rating scale as primary endpoint, most often the MADRS (25), and none enrolled adolescents.
  4. Only five patient trials (2%) had a site in a low- or middle-income country.

The gain

An LLM connected to the registry API, using documented field-specific query expressions, produced a structured and reproducible map of 412 core-set psychedelic trials plus 300 ketamine/esketamine trials.

The problem

Open-ended conversational querying by the LLM underestimated the registry record several-fold for some compounds, so unstructured use gave unreliable counts.

The rundown

The core set covered classic serotonin 5-HT2A agonists, MDMA, ibogaine and ayahuasca, with ketamine and esketamine in psychiatric indications reported separately as a comparator class. Populations and indications were defined from structured registry fields.

Depression led for psilocybin (74 of 181 patient trials) and the ketamine class (237 of 300), PTSD for MDMA (35 of 47) and anxiety for LSD (6 of 13).

Twenty-four core-set trials were registered as phase 3, 13 of them psilocybin. Psilocybin phase 3 programmes plan more than 1900 participants, the pivotal MDMA trials enrolled 100 and 121, and LSD phase 3 trials plan 149 to 245. No other psychedelic had a phase 3 trial.

Among the 63 phase 2/3 psilocybin depression trials, 34 (54%) assessed outcomes at weeks 3 to 6 and 26 (41%) excluded adults over 65.

What this doesn’t fix

The authors state the approach complements rather than replaces systematic review, and its accuracy against manual extraction remains to be quantified.

Sources

  1. Peer-reviewedTrials2026-10-14

The debate