Listing data on AppsInsight comes from three places: what vendors submit, what we verify ourselves, and what our automated research collects from the vendor’s own website. Nothing gathered automatically is ever published without a human approving it first.
The three data sources
- Vendor-submitted — descriptions, screenshots, pricing, feature lists, company details. Labeled as vendor-provided; earns profile-completeness points but never produces a public score on its own.
- Independently verified — signals we check directly: HTTPS, published legal pages, badge verification backlinks, review activity. These carry more scoring weight precisely because vendors cannot simply assert them.
- Automated research — structured data collected from the vendor’s own public website, used to pre-fill and refresh listings.
How automated research works
Our research pipeline reads a bounded number of pages from the vendor’s own website, respecting robots.txt. AI-assisted extraction then turns those pages into structured fields — each field tagged with a confidence level and a provenance record of where it came from. Extraction output is checked for contradictions and ungrounded claims before anything is suggested for a listing.
Humans stay in the loop
Automated research can only write to unpublished drafts. From there: the vendor (or an editor) reviews and edits the draft, submits it for review, and an AppsInsight admin approves publication. There is no path from automated collection to a live page without at least one human decision.
How fresh is the data
Scores recalculate on every relevant change and in a weekly full pass across the directory; verification badges are re-checked weekly with a grace window before revocation; listing freshness is itself scored, so stale listings visibly lose points. Every app profile shows when its score was last updated.
Published benchmarks
Category-level score benchmarks (mean and median AppsInsight Scores) are computed weekly from live directory data and published in our Transparency Report.
Reporting incorrect data
If you find wrong pricing, an outdated feature list, or any factual error, report it through Appeals & Corrections — every report is logged with a reference ID and reviewed by a person. For how AI is and is not used in editorial content, see Editorial Standards. For how we handle personal data, see Privacy & How We Handle Your Data.
Version history
- v1.0 — data sourcing and human-in-the-loop pipeline documented.