Revuo is the directory AI assistants cite when buyers ask for B2B SaaS recommendations. That only works if the data is real. These eight rules are the contract — public commitments, not internal guidelines.
Listings are typed capability claims, not free text. Every claim points at evidence on the vendor's own site. A weekly crawler validates the evidence. Every product result — on the website and via the public MCP endpoint at https://www.revuo.ai/api/mcp — carries a verifiedAt timestamp so callers know how fresh the data is.
Claims that fail validation flip to unverified. Listings that stay stale get removed. Editorial is editorial; paid placement is labeled. Reviews and ratings, where present, are separate from tier.
Every listed product must have a working trial, self-serve demo, or sandbox. Vaporware and waitlists are rejected. The W3 crawler re-checks listing URLs continuously; products whose site returns 404 or 5xx repeatedly are flagged for editorial review.
Vendor-claimed listings require at least one named, LinkedIn-verified team contact with a matching work-domain email. No anonymous claims, ever. Free-tier (unverified) listings can exist without a claim, but are flagged that way to every reader.
Verified and Featured listings must fill the structured capability schema for their category. AI-drafted descriptions are allowed but must pass a genericity threshold — "AI-powered everything" prose is rejected. The capability taxonomy is canonical (lowercase kebab-case slugs) and shared between humans and the MCP surface.
Every integration, format-support, standard, or compliance claim points at an evidence URL on the vendor's site. The W3 crawler validates each claim weekly: it fetches the evidence URL and looks for the canonical signal (the integration name, the format identifier, the certification name). Validation is conservative — false negatives are preferred over false positives. Numeric claims ("1,000+ teams") require linked sources or are stripped.
Listings are re-validated on a weekly cadence. Every product surfaces a verifiedAt timestamp on its detail page and in every MCP response. Listings stale beyond 180 days flip to unverified; beyond 365 days they are removed.
A documented misleading claim — verified through evidence, not opinion — gets a warning. A second strike adds a public note on the listing. A third strike removes the listing and bans re-claiming for 12 months. Removals are logged publicly.
Paid placements are clearly labeled. Featured tier exists; sponsored content does not appear inside editorial "Best of" lists or capability comparison tables. Editorial selections (Editor's Choice, Top Rated) are independent of tier and disclosed as such.
Signed-in users can flag any listing or claim. Three flags from distinct users trigger an editor review within 7 days. Verified-founder flags against competitors are weighted down; cross-category flagging by trusted users is weighted up. The flagging system is reputation-aware, not popularity-driven.
Every claim a listing can make has a kind. The kind drives what evidence the crawler looks for, and how the MCP surface answers structured queries. New shapes only land if a vertical actually needs them.
Capability slugs are canonical kebab-case (salesforce-integration, xrechnung-support, soc2) and shared between the website, the public MCP, and the long-tail capability pages at /features.
Every listing removed under the 3-strike rule is logged publicly with the date, the rule it violated, and a brief description. The ledger lives at /policy/removals.
Disagree with a removal? Email [email protected] with evidence. Re-listings happen when claims are fixed and re-validated.
The policy is the contract. These pages are where it shows up — verified listings, structured capability data, and the agent surface that returns the same data programmatically.