Privacy by architecture

Legitimeter is built around a short list of design rules. A design that breaks one is rejected.

This page describes the extension's analysis engine and sync. Website analytics are covered in the Privacy Policy.

1. You stay in charge

Legitimeter acts for you, on pages you already open, using preferences you control. It is not a crawler. Its findings describe evidence. What happens next (hide, label, de-emphasize, ignore) is your setting.

2. It observes; it does not acquire

Legitimeter analyzes what your browser has already received during your own browsing. It does not make extra requests to the sites you view unless you turn on image re-reading, which re-requests only an image the page already showed you, without cookies, and never to enlarge a dataset. This is a product rule, not a legal guarantee: we review releases against applicable law.

3. Local stays local

In the Legitimeter analysis engine, information that can stay on your device stays there. When the extension syncs, the payload is built from an allow-list of fields marked shareable; private messages, drafts, passwords and unrelated browsing are not included. The only deletion currently offered is erasure of testimonials you submitted.

4. Trust is earned

We assume any participant could misbehave. Contributions are submitted over signed, account-authenticated requests. Network evidence is capped by contributor trust and revocable.

5. Using AI is not being a bot

A person writing with AI help is not scored as an automated account. "AI-assisted" is a note, not one of the five signals (automation, synthetic media, coordination, spam, impersonation).

Labels are probabilistic estimates, not verdicts. Legitimeter reports evidence and confidence; it does not claim to catch every fake or to be always right. We publish no accuracy figure until one has been measured and can be substantiated.

See also: How it works and the Privacy Policy.