SafisAI scans modern pages locally for AI workflows, AI presence, image signals, and data privacy clues, then explains findings as OBSERVED, DISCLOSED, or POSSIBLE. No page upload. No SafisAI backend. We don't even own a server.
SafisAI is built around a simple question: is there any AI involved on this page? The extension favors evidence over hype, showing what was observed directly, what a site disclosed, and what is only possible.
When a page is using AI, people should be able to understand how it shows up, whether it is disclosed clearly, and whether their data may be involved in that system.
If a page is using AI, SafisAI aims to show whether it appears as a workflow, assistant, generation layer, embedded tool, or other page-level signal.
AI should not feel invisible by default. Some use is obvious, but disclosure and technical signals are not always visible without a closer look.
If AI features may touch prompts, text, images, or activity, users deserve clearer privacy context and careful evidence behind each claim.
Many AI features do not announce themselves in plain text. SafisAI is designed to inspect page structure, embedded UI, and privacy signals without turning every hint into a certainty.
SafisAI looks beyond simple page text and inspects exposed page structure, including open Shadow DOM content where the browser makes it available.
When AI workflows appear alongside analytics or session tooling, SafisAI can surface that privacy context so the page is not judged in isolation.
Scans run locally in the browser. SafisAI does not upload page contents to a server to decide whether AI is present.
SafisAI performs local snapshot scans of page structure, workflow behavior, image signals, and privacy clues, then limits rescans to avoid stale or noisy results. Hidden tabs skip work, repeated identical findings are deduplicated, and results stay on-device. The cache automatically empties with a browser restart.
Instead of one vague AI label, SafisAI breaks the page down into focused buckets so you can see what kind of evidence exists and how strong that evidence is.
Flags when a page behaves like an AI tool or AI agent workflow. Known platforms help, but workflow behavior matters too.
Surfaces scripts, widgets, references, and page-level signals that suggest AI is present without overstating certainty.
Checks AI-related image signals and provenance clues where available, while keeping wording conservative when proof is limited.
Separates privacy findings into OBSERVED, DISCLOSED, and POSSIBLE so direct technical evidence does not get mixed with softer inference.
SafisAI is designed to communicate what was actually found, not the strongest possible assumption.
Install SafisAI directly from the Chrome Web Store.