How Drug Landscape fact-checks every drug profile
Every drug profile is automatically cross-referenced against four authoritative public sources. When a claim on our page conflicts with a primary source, we log the mismatch, demote the page's quality score, and regenerate the affected content. This page documents exactly how it works.
The four sources we cross-check against
- ClinicalTrials.gov (US National Library of Medicine) — authoritative for which conditions a drug is actually being studied in, sponsor names, trial design, primary endpoints. We search by intervention name and aggregate matching trials.
- ChEMBL (EMBL-EBI) — authoritative for drug class, molecular target, modality (small molecule, protein, antibody, etc.) and mechanism-of-action text. We query by molecule synonym and resolve target details via the linked target endpoint.
- DailyMed (US NLM) — full FDA-approved prescribing information for every US-marketed drug. We use it to verify approved indications + black-box warnings against the official label.
- Wikipedia — coarse-grained cross-check on common terms (drug class, broad indication). Lower confidence than the other three but high coverage.
How the comparison works
For each visited drug we run four parallel API calls (~6-10 seconds wall-clock) and then compare each authoritative fact against the corresponding field in our database:
| Field | Authoritative source | What triggers a flag |
|---|---|---|
| Drug class | ChEMBL action_type | Zero word overlap between our class and ChEMBL's — e.g. "Wnt/β-catenin inhibitor" vs "muscarinic antagonist". Critical |
| Molecular target | ChEMBL target endpoint | Our target string doesn't substring-match ChEMBL's preferred name. Major |
| Modality | ChEMBL molecule_type | Our modality classification disagrees (e.g. "small molecule" vs "protein"). Minor |
| Indications | ClinicalTrials.gov conditions | None of our indications overlap with the conditions the drug is being studied in. Major |
| Sponsor | ClinicalTrials.gov leadSponsor | Sponsor name disagrees with the registered trial sponsor. Minor (common for licenced drugs). |
What happens when a mismatch is detected
- The mismatch is logged to our public
fact_check_logtable with field, claimed value, source value, severity, and timestamp. - On critical or major severity, the drug's
quality_scoreis demoted to ≤10 andfact_check_statusis set to failed. - The drug's profile shows an Under review badge instead of the verified one.
- An AI summary is regenerated, grounded strictly in the verified source facts (no hallucination).
- Drugs with no critical or major mismatches earn the ✓ Verified [date] badge.
Recent mismatches caught
Live feed of the last 12 critical or major mismatches the system has caught. Each one was logged automatically.
| Drug | Field | Severity | Source | Detected |
|---|---|---|---|---|
| methylprednisolone-succinate | indications | major | ClinicalTrials.gov | 20 May 2026 |
| b-clorhexidine-2 | indications | major | ClinicalTrials.gov | 20 May 2026 |
| levofloxacin-500 | indications | major | ClinicalTrials.gov | 20 May 2026 |
| oxybuprocaine | indications | major | ClinicalTrials.gov | 20 May 2026 |
| course-b2-vin | indications | major | ClinicalTrials.gov | 20 May 2026 |
| tigerase-and-best-available-care | indications | major | ClinicalTrials.gov | 20 May 2026 |
| soolantra-ivermectin-cream-1 | drug_class | critical | ChEMBL | 20 May 2026 |
| isoflurane | indications | major | ClinicalTrials.gov | 20 May 2026 |
| isoflurane | molecular_target | major | ChEMBL | 20 May 2026 |
| isoflurane | drug_class | critical | ChEMBL | 20 May 2026 |
| bremen-eye-drops | indications | major | ClinicalTrials.gov | 20 May 2026 |
| ticagrelor-orodispersible-tablets | molecular_target | major | ChEMBL | 20 May 2026 |
Why this matters
Most pharma intelligence sources are either (a) paywalled and unauditable, or (b) AI-generated with no source attribution. Drug Landscape is open about exactly where each claim comes from and what we did to verify it. If you find a mismatch we haven't caught yet, the corrections email reaches our editor team within minutes.
Report an error: editor@druglandscape.com · Methodology questions: /methodology · Editorial policy: /editorial-policy