Mozilla Firefox

Since Firefox 133, the browser sidebar can host an AI chatbot. You choose the provider: Anthropic Claude, ChatGPT, Google Gemini or Le Chat Mistral, with Microsoft Copilot added in Firefox 143. Mozilla does not restrict the list by region, on the grounds that availability changes constantly and choice should not be pre-empted.

Mechanically, it is the provider's own web experience running in a panel next to your page, plus some plumbing that passes context across. That means your account, your history and your subscription live with the provider, exactly as if you had opened a tab — and it means Mozilla is not paying inference bills or training anything.

Select text on a page and Firefox offers pre-written prompts, whose exact wording Mozilla publishes: Summarize, Explain this, Quiz me and Proofread. Each is prefixed with a line naming the page and the selection.

Quiz me is the one worth knowing about. It instructs the chatbot to test you on the selection with a mix of multiple choice, true/false and short answer, waiting for your response before continuing. That converts any documentation page, article or textbook chapter into an active-recall exercise, in two clicks, for free. It is the single most useful thing in Firefox's AI feature set and almost nobody knows it exists.

There is also a Summarize Page option at the bottom of the sidebar and in the right-click menu, which sends the page content rather than a selection.

The support documentation is refreshingly concrete, and reviewers should hold other vendors to this standard. Opening the sidebar exposes the same data to the provider as visiting its site directly. A suggested prompt sends the selected text, the prompt and the page title. Summarize Page sends the page content, the prompt and the page title. Firefox has no access to your conversations. Mozilla does collect telemetry about feature usage — which provider was chosen, how often prompts are used — and says so on the same page.

It also spends several paragraphs warning users that chatbots predict text rather than deliver facts, that you agree to the provider's terms when you use one, and that some providers are more privacy-respecting than others, with a link to the Mozilla Foundation's guidance. A browser vendor talking down its own feature is not a marketing decision; it is an editorial one, and it is the reason this integration reads as trustworthy in a way most do not.

Untick AI chatbot in the sidebar customisation panel and the feature disappears. Hide the text-selection shortcut separately if you want the sidebar but not the prompt popover. And Firefox ships AI controls on desktop, a dedicated surface for reviewing and blocking generative AI features across the browser.

The existence of that surface tells you something about Firefox's user base — this is a community that objected loudly to AI features appearing at all — and something about Mozilla's position: it can afford to make its AI optional precisely because it is not betting the company on it.

Firefox's full-page translation runs on your own device. Not on a Mozilla server, not on Google's. The page text does not leave your machine to be translated, which is a materially different privacy posture from every cloud translator, and it is invisible in every discussion of "AI browsers" because it shipped before the category existed and does not have a chat interface.

The cost is coverage: an on-device translation model supports far fewer languages than Google Translate's 249. For European languages and common pairs it is there, and for anything sensitive it is the right default.

There is a reading of Firefox's AI features as underinvestment, and there is a reading of them as the only sane response to Mozilla's actual position. Both are partly right, and the second is more useful.

Mozilla cannot outspend Google or Microsoft on models, and any first-party assistant it shipped would be worse than the ones its users can already reach in a tab. What it can do is what browsers have always done: broker a relationship between the user and a service on terms the user controls. Choose your provider. See what gets sent. Switch whenever you like. Turn it off entirely.

That is a coherent product philosophy rather than an absence of one, and it is consistent with the reason Firefox still exists. The open question is whether it holds as browsing itself becomes more agentic — a frame around someone else's chat window is a thin position if the competition is a browser that completes tasks. Mozilla's answer so far is that it would rather be trusted than first, which is at least a choice.

Not competitive on AI capability, and not trying to be. Brave's Leo does more, works without an account, and can be pointed at your own local model. Chrome's Gemini integration goes deeper into the browser. Neither lets you swap the assistant for a competitor's on a whim, and neither ships a documented control panel for switching the AI off.

Choose Firefox for the reasons people have always chosen Firefox — independence from the two platform giants, control over your own browser, an organisation that publishes what it collects. The AI features are a competent, optional bonus on top of that, and the honest summary is that they add a convenient frame around tools you could already use in a tab.