Photo checks for review platforms and user-generated content
Review sites can screen customer photos for AI-generated images before they appear next to a product or a restaurant. PhotoCheck scores each photo, reads its metadata, and finds faces that may need consent or blurring, at about a cent a photo.
The problem
Fake reviews used to be text. Now they come with photos: generated plates of food, flawless unboxings, hotel rooms that don't exist. A fake photo makes a fake review more convincing, and it's one of the few parts of a review you can check automatically.
The checks that matter
| AI score | Holds back photos that read as generated. |
|---|---|
| Faces | Customer photos with other people in them can need blurring. |
| Declared AI | Generated images that kept their labels are removed outright. |
A decision rule
Built on POST /api/v1/analyze. Thresholds are starting points; tune them to what a mistake costs you.
const r = await check(reviewPhoto);
if (r.summary.ai_generated === "declared") return "remove";
if (r.ai.score >= 0.8) return "hold_for_review";
if (r.faces.count > 0) return "publish_blurred";
return "publish";Questions
How much does a photo cost to check?
One credit, $0.0075 to $0.05 depending on the pack. /v1/analyze runs all three checks for that one credit.
Can we run it asynchronously?
A full check takes about a second. Call it inline at upload for single photos, or from a background job when checking many at once.