Skip to content
sighting.ai

Overview

One endpoint runs the same engine as the app and returns the full evidence report as JSON — verdict, confidence range, and whatever evidence that lane can produce: flagged spans and ranked signals for text, a ranked provenance record for image and video. Access is part of the Pro plan, which is not on sale yet.

API access is not on sale yet.

The endpoint is live and these docs describe it accurately. Access belongs to the Pro plan, which we have not opened for purchase — so a key cannot be issued today, and these pages are here to be read rather than acted on. Free accounts have no API access, and requests from one are refused with a 403. Nothing here is a waiting list; there is simply nothing to join yet.

Endpoint

One endpoint covers every modality:

POST https://sighting.ai/api/v1/detect

Authentication

Every request needs a bearer key in the Authorization header. Create and revoke keys at Settings → API keys.

Authorization: Bearer sk_live_...

Authentication covers scopes and what a rejected key looks like.

Your first request

curl
curl https://sighting.ai/api/v1/detect \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "modality": "text",
    "text": "In today'\''s rapidly evolving digital landscape..."
  }'

A successful request returns the full detection report: the verdict, the point estimate aiLikelihood, the confidence range, the ranked signals behind it, plain-language reasoning, and the evidence spans that were flagged. Detect documents every request and response field.

Good to know

  • Text runs on a transformer classifier we host ourselves — nothing is sent to a third-party detector, and the same input always produces the same report.
  • modality: "code" is accepted and answered, but its verdict is always uncertain: we measure the file and deliberately do not judge its author. Code returns no verdict says why.
  • modality: "image" and modality: "video" run an evidence ladder rather than one blended score, and the strongest signal that fires decides alone. A human verdict is reachable only here, and only from a signed capture manifest whose signer chains to a trust list we accept — never from a score.
  • Video is the narrowest lane. A signed file can be settled outright; an unsigned one gets a real provenance report and an uncertain verdict, because nothing that can honestly decide it exists yet. Image differs in one respect: a pixel classifier reads what no manifest settled, so an unsigned AI image can still be identified — but never cleared.
  • Keys are scoped; detection requests need the detect:write scope.
  • With redactExcerpts: true, stored evidence keeps offsets and scores but drops the raw text.

How the API earns trust

One open API
The docs match the endpoint, field for field.
Keys hashed, shown once
API credentials are never recoverable.
Zero-retention available
Verdicts without storing your content.