Methodology

How Drug Landscape builds, scores, and validates its data. Every number we publish has a documented source and a documented formula. No black boxes.

Scoring & analysis

Data sources

AI grounding

Every LLM output on the platform — patent summaries, drug briefs, company briefs, landscape reports, claim analyses — uses a strict-grounding prompt: the model receives JSON facts pulled from our database and is forbidden from inventing companies, drugs, indications, or numbers not present in those facts. Outputs are post-validated with a hedge-detection filter that rejects responses containing speculative language ("may", "might", "possibly", "likely", "appears"). We use Groq's Llama 3.1 8B model for cost and latency.

Refresh schedule

Corrections

If you spot something wrong, email hello@druglandscape.com. We log every correction and re-validate the upstream source.