What is FalBot?
FalBot identifies callback traffic from fal.ai's asynchronous model queue. A client supplies a webhook_url when it submits a queued inference request. After the work finishes, fal.ai posts a JSON result to that URL with the request_id, a status of OK or ERROR, and either the model payload or error information.
The receiver acknowledges delivery with a 2xx response. fal.ai does not follow redirects, so the submitted URL must already be the final endpoint. Network failures, timeouts, and 4xx or 5xx responses can be retried while the stored result remains available, up to the documented maximum of 31 retries. Repeated deliveries share the same request_id and need idempotent handling.
Authentic callbacks carry request ID, user ID, timestamp, and signature headers. fal.ai signs a message derived from those values and the SHA-256 hash of the raw body with Ed25519. Receivers verify it against the public keys at fal.ai's JWKS endpoint and reject timestamps more than five minutes from the current time. The FalBot user-agent token is useful for log classification, but it does not prove who sent the request.
A FalBot payload can contain the output of an AI generation job, but the callback only posts that result to the supplied URL. The operation does not feed the receiving site's pages into AI search or training systems. Robots.txt is unrelated because the destination URL was supplied with the queue request.
