AI Transparency

Google's Gemini model watches your video and produces structured, per-factor observations — what it saw at each point in the video, mapped to the GVIF framework's 39 factors. It never outputs an overall score, and it never decides how much any factor is worth.

All of that math — factor weights, dimension rollups, the final 0-100 score, and the classification band — is computed by GroVirally's own deterministic scoring engine, using fixed, versioned weights (see the Framework page). The same evidence always produces the same score, regardless of which AI model produced that evidence.

The video you submit (file or fetched from a supported URL) is sent to Google's Gemini API for analysis. No account credentials, payment information, or data about other users is included.

• We don't use your videos to train our own models.
• We don't ask the AI to guess whether a video will go viral — only to describe what it observed.
• We don't let the AI's output override the deterministic weights when computing your score.

Video content processed through the Gemini API is subject to Google's own API terms and data handling policies, which govern how Google itself processes data submitted through its API. We encourage you to review Google's published API terms directly for the specifics of their retention and training policies, since those are set and controlled by Google, not by us.

Every completed report permanently records which AI provider and model produced its evidence, which version of the GVIF framework scored it, and which version of our scoring engine ran the calculation — so a report from today stays fully interpretable even after any of those are upgraded later.

Every report includes a confidence value reflecting how complete the evidence was (e.g. shorter videos give the AI less to work with). Low confidence is surfaced, not hidden.

For more on data handling generally, see our Privacy Policy, or get in touch directly.