Detect AI-generated images
This feature is not available in the self-hosted deployment.
Enabled by default
The API automatically checks for AI-generated images and reports the result at verification.checks.visualCheck.generativeAiCheck.
{
"generativeAiCheck": {
"result": "Pass",
"passesAtOrBelowSensitivity": "Level10",
"firstSide": { "result": "Pass", "passesAtOrBelowSensitivity": "Level10" },
"secondSide": { "result": "Pass", "passesAtOrBelowSensitivity": "Level10" }
}
}
This is a tiered multi-side check.
Besides its own result (Pass, Fail, or NotPerformed), it reports passesAtOrBelowSensitivity and a separate sub-result for firstSide and secondSide.
Two settings change whether and how strictly the check runs:
- Setting
verification.useCase.verificationContexttoInPersondisables it outright, along with the screen presence and photocopy checks. See use cases. - The
HighConversionverification policy lowers it toLevel4.BalancedandHighAssuranceleave it atLevel5.
Tune the sensitivity
As with other checks, you can set your desired sensitivity, or you can disable this check if you don't need it.
{
"verification": {
"settings": { "generativeAiSensitivity": "Level10" }
}
}
Level1 is the least strict and Level10 is the most strict, and Disabled skips the check entirely.
If left unset, the default is Level5.
Impact on the overall result
If the generative AI check fails, then visualCheck fails, which in turn affects the overall verdict.