Brand safety monitoring
Watching what AI surfaces say about a brand for content that is false, defamatory, or exposes the company to regulatory or reputational risk, a regulated claim attributed incorrectly, a competitor’s product recommended in a brand’s place, an unsafe or non-compliant instruction tied to a product name. It works as a discipline only when each observation is a structured, repeatable capture you can diff over time, not a one-off screenshot nobody can re-derive.
In AI Search API
The same captures used for visibility and share-of-voice work double as brand-safety monitoring, because every one already returns the answer text and version-stamped provenance in one Envelope, nothing extra to opt into. A capture flagged for review carries the exact answer an AI surface gave, on a specific date, in a specific market, as structured fields you can query and diff. idempotencyKey keeps a scheduled risk-monitoring run replay-safe, and a surface returning nothing completes honestly as surface_absent rather than as a gap someone has to notice was never checked.
"answer": { "text": "…", "markdown": "…" },
"provenance": { "official": false, "surfacePresent": true }Related terms
See it in the product
Every definition here is enforced in the Envelope contract. Run a capture and read the fields yourself.