Psilocybin research in October 2026: a reproducible inventory
The public metadata core downloaded on 7 October 2026 contained 5,076 visible records: 4,315 labeled published, 450 preprints, 308 clinical-trial registrations and three protocols. This brief documents that compilation, with reusable analytical data and an executable reproduction script.
These are bibliographic records, not independent experiments or participants. Counts describe this index, not the entire field, scientific consensus or treatment effectiveness.
What kinds of records are in the compilation?
| Stored status | Records | Share |
|---|---|---|
| Published | 4,315 | 85.0% |
| Preprint | 450 | 8.9% |
| Clinical trial | 308 | 6.1% |
| Protocol | 3 | 0.1% |
“Published” is a metadata status, not an independent audit of peer review. Preprints and registrations are distinct categories. A protocol and a later report can describe the same study; deduplication of bibliographic manifestations does not make every row an independent experiment.
For example, the study-type label “Clinical Trial” appears on 977 records, while the publication-status category “clinical trial” contains 308 registrations. The fields answer different questions: design classification versus type of record. The two counts must not be treated as competing estimates of the number of trials.
Recorded years, with publication status kept separate
| Year | All records | Published | Preprints | Registrations | Other |
|---|---|---|---|---|---|
| 2026 (partial) | 985 | 731 | 88 | 165 | 1 |
| 2025 | 888 | 736 | 96 | 56 | 0 |
| 2024 | 654 | 541 | 80 | 33 | 0 |
| 2023 | 550 | 453 | 72 | 25 | 0 |
| 2022 | 423 | 368 | 44 | 11 | 0 |
| 2021 | 304 | 252 | 45 | 7 | 0 |
| 2020 | 170 | 149 | 15 | 4 | 2 |
The 2025 recorded-year cohort contains 888 records, including 736 labeled published; the 2020 cohort contains 170, including 149 labeled published. These observed differences are features of this compilation. Search coverage, later backfills, source indexing and curation affect the counts, so this brief does not estimate a field-wide growth rate.
Source labels and identifier coverage
4,334 records (85.4%) have a DOI field and 2,979 (58.7%) have a PubMed identifier. The groups overlap. Presence is not an independent resolution or identity audit; missing DOI fields are expected for many registrations.
| Source | Records | Share |
|---|---|---|
| PubMed | 2,582 | 50.9% |
| OpenAlex | 1,174 | 23.1% |
| Crossref | 418 | 8.2% |
| ClinicalTrials.gov | 308 | 6.1% |
| Europe PMC | 237 | 4.7% |
| PsyArXiv | 186 | 3.7% |
| bioRxiv | 127 | 2.5% |
| medRxiv | 44 | 0.9% |
A work may occur in several upstream databases and receive enrichment from another source. The retained source label does not measure exclusive coverage, acquisition yield or the reliability of a study. Read the source and dataset guide.
Download the frozen input and reproduce every table
- Record-level analytical CSV: all 5,076 rows, with record ID, year, status, source, study type, DOI and PubMed ID. It contains no abstracts or registry descriptions.
- All-year status counts, status totals, source totals and study-type totals.
- JSON summary and input checksums.
- Python reproduction script: standard library only; no package installation needed.
- Figure reproduction script: uses Matplotlib to create the SVG and PNG from
summary.json.
Save the CSV and script in the same directory, then run:
python3 research_brief.py --records records.csv --output reproducedThe resulting CSV summaries reproduce the frozen tables. To analyze a newer public SQLite download instead, run python3 research_brief.py --database downloaded.sqlite --output current. New live input can yield different counts.
- Published analytical CSV SHA-256
bb711860a81bfcfa741970ffc133c194de8658a856af478dbeb7e3f81c3804b7- Original public SQLite download SHA-256
0e6fd79ee574846aa2018990e5acaedaf2cb3c336be8250c6155dc7ad709426b
The CSV is a reduced analytical input, not a replacement for the full bibliographic database or an immutable Zenodo deposit. Compiler-held selection, arrangement and original analysis follow the scoped rights notice; third-party material is not relicensed.
Methods, limitations and citation
The input is the rights-sanitized SQLite core served by database.php, retrieved on 7 October 2026. That public core already excludes hidden and false-positive records. Every remaining row contributes once to its stored year and status. Empty classification fields are grouped as “Unclassified”; missing years are counted separately. Study labels are automatic discovery aids and do not establish independently verified design or certainty of evidence.
The reproduction script reads an explicit allowlist of analytical fields and calculates counts without opening abstracts, keywords, registry descriptions or raw importer payloads. The input hash identifies the exact downloaded bytes. The separately hashed analytical CSV allows readers to reproduce the reported aggregates even after the live index changes.
Related papers, protocols and registrations can remain separate legitimate records. Recent publication years are affected by future issue dates and indexing delays. No patient-level observations, treatment-effect estimates, causal conclusions or comprehensive-coverage claims follow from these counts.
Suggested citation: Germann, C. B. Psilocybin research in October 2026: a reproducible inventory. Psilocybin Research Tracker, 7 October 2026. https://psilocybin-research.com/psilocybin-research-brief-2026-10/. Cite the analytical CSV checksum when reproducing these tables. This web research brief is not a peer-reviewed journal article.