AI Methodology
We use AI to gather structured data at scale, but never as the final word. Every step is designed to keep the catalog accurate, sourced and human-checked.
1 · Enrichment
When a tool is submitted, our pipeline reads its website and public sources to extract structured fields — pricing model, plans, features, APIs, integrations and more. AI models normalize this into a consistent schema.
2 · Evidence
Each extracted fact is attached to the source it came from, including the URL and a short excerpt. This provenance is stored so anyone can trace where a claim originated.
3 · Confidence scoring
Facts receive a confidence score based on source quality, agreement across sources and recency. Low-confidence or conflicting data is held back or flagged rather than published as certain.
4 · Human review
Editors and trusted contributors approve AI proposals before they go live. Verified makers can confirm or correct their own tool’s data. Community corrections feed back into the record.
5 · Continuous re-checking
Tools change, so we periodically re-verify facts and mark stale ones for refresh. See our editorial policy for the standards behind every listing.