Privacy

Opinion Meter is a browser extension that shows what people think of the things in your search results. This page says exactly what it sends and keeps.

What the extension sends

On Google search result pages only: the titles and addresses of the results on the page and the words you searched for. Nothing from any other page, ever. No page content, no cookies, no account details, no browsing history.

This includes sponsored results, shopping tiles and AI Overview reference links. The main reading concerns the search query; individual result readings concern the linked page, with a clearly labelled website fallback. Opening a result resends that one result's title and address and the original query to recover the reading if necessary.

Each install makes a random token so the server can limit how often one copy asks. The token identifies an install, not a person, and is never linked to an account.

What the server keeps

The server keeps only what it works out: the split of opinion, the count, the summary sentence and links to the public posts it read, for up to 24 hours per subject. It does not keep the posts themselves. The quotes shown when you open a subject are fetched live, so a post deleted on its platform disappears from Opinion Meter at once.

Request records used to prevent abuse are kept for at most 7 days and contain the install token and the time of the request, not the search.

Where the opinions come from

Public posts and comments read through the official interfaces of Reddit, YouTube, Hacker News and Bluesky, within the limits each platform sets. The text of those posts is sent to an AI model to be classified and summarised; usernames are not. Nothing is used to train any model.

What is a subject

Anything people talk about: products, films, apps, places, companies, articles, topics and public figures. For a person, the reading is about their work and public conduct, drawn only from public posts, and like every reading it is kept only as short-lived derived data.

Contact

Questions about this page: open an issue on the project's GitHub repository.