Run a free scan of the public evidence an agent can actually retrieve. Then get a fixed-scope evidence map and three prioritized repairs for catalogs, policies, checkout, payments, and attribution.
Built for merchants and agencies that need to know what an AI agent can discover, what it may attempt, and what evidence actually proves commercial impact.
Inspect public agent instructions, robots policy, structured profiles, product identity, price scope, freshness, route availability, and the audience, use-case, differentiation, and value evidence an agent needs to make a recommendation.
Separate technical capability from agent identity, customer-account ownership, delegated scopes, connector-specific access, buyer approval, publish/send/spend approval, rail-specific budget enforcement and revocation, provider automation and human-presence rules, policy effective dates, qualified review and disclosure duties, physical-action stop controls, AI-plan or credit consent, vendor fees, data sharing, payment authorization, replay-safe receipts, and intent-to-effect consistency.
Distinguish paid placement, affiliate influence, sponsorship disclosure, recommendation independence, ad clicks, conversion data, incremental lift, cart readiness, completed checkout, delivery, settlement, reversals, costs, and collected net revenue.
The scanner checks a bounded set of public HTTPS surfaces: the homepage, structured offer evidence, policy links, robots rules, llms.txt, and a UCP discovery profile. It does not log in, transact, bypass access controls, or store the report.
/.well-known/ucp and /.well-known/agent-card.json; MCP tools are served from /mcp.No signup. No result storage.
This is a public-evidence snapshot, not a certification. Missing or unreadable evidence is not proof that a capability is absent.
Answer fourteen questions using current, independently readable evidence. “Unsure” counts as a gap because an agent cannot safely infer what your systems do not prove.
The audit records what was checked, what the evidence supports, what remains unknown, and the smallest repair that can be independently retested.
This form is the dedicated business channel for the offer. It does not use personal email, phone, profiles, contacts, referrals, or advertising audiences.
Choose the surface where access, authority, or attribution is most likely to fail first.
Map what outside agents may discover, compare, place in a cart, purchase, or support—and make blocks and approvals explicit.
Test persistent jobs, event triggers, sub-agent identities, connector writes, memory, cost controls, checkpoints, and rollback.
Verify recommendation independence, disclosure, conversation-data boundaries, and handoff attribution.
Check property truth, booking identity, rate and fee transparency, guest-service authority, and completion receipts.