Suggest adding: DeepFakeCheck - layered deepfake analysis for video, audio, images, and text

Hi Bellingcat toolkit maintainers,

I would like to suggest DeepFakeCheck for consideration in the toolkit, potentially under image/video verification or media analysis:

https://deepfakecheck.io/?utm_source=bellingcat&utm_medium=referral&utm_campaign=directory_launch_2026q3

DeepFakeCheck is an online tool for reviewing video, audio, images, and text for signs of AI generation or manipulation. It combines multiple evidence layers: local C2PA provenance verification when credentials are available, metadata clues, and AI-based media analysis. It returns a risk rating with uncertainty notes instead of presenting the result as proof. Video checks sample up to 10 frames across the full timeline, and audio checks can support initial triage of suspected voice-cloning scams.

Why it may fit the toolkit: journalists, fact-checkers, and open source researchers sometimes need a quick first-pass review of suspicious media before deciding what to verify independently. The tool covers several media types in one interface, offers free checks, requires no registration, and is available in 18 languages.

Honest limitations:

- Results are supporting evidence, not proof, and should be combined with independent verification.
- AI-based detection can produce false positives and false negatives.
- C2PA credentials and useful metadata may be absent; metadata can also be edited.
- Sampling up to 10 video frames does not analyze every frame.
- DeepFakeCheck does not retain uploaded files after a request completes, but media processed by Google Gemini is also subject to Google privacy terms.
- Audio results should not be used alone for financial, identity, legal, or safety decisions.

Disclosure: I am connected with DeepFakeCheck. I am suggesting it because its evidence-first framing and explicit uncertainty may be useful in open source investigation workflows. No pressure if it is outside the toolkit scope.

Review the proposed DeepFakeCheck service and the toolkit's scope before deciding whether it belongs under image/video verification or media analysis. If accepted, add the entry with its evidence-first framing, limitations, privacy notes, and disclosure; done means the listing is accurate and aligned with the toolkit's curation standards.

Written by the indexing model from the issue text.