Chrome's Gemini and Edge's Copilot Are Reading Your Browsing: How to Turn Them Off (2026 Guide)

The short version: Chrome's built-in Gemini and Edge's built-in Copilot can read the content of every page you have open, and on both browsers the assistant ships enabled by default. Turning it off takes about two minutes per browser. This guide shows exactly where the toggles are, what data each assistant sends to the cloud before you disable it, and which tools to use instead if you still want an AI assistant while browsing.

Last updated: 2026-07-10

If you updated Chrome or Edge any time in the last year, you likely have an AI assistant sitting in your browser toolbar that can see and summarize the page you're looking at — your bank statement, your medical portal, an internal company dashboard, a client's contract in Google Docs — without you asking it to do anything first. On both browsers, the feature is opt-out, not opt-in.

This is a meaningfully different privacy posture than a chatbot you visit and paste text into deliberately. A page-reading browser assistant sits between you and everything you do online, by default, running in the background.

What Chrome's Gemini Actually Sees

Chrome's built-in Gemini integration (rolled out broadly through 2025 and 2026 under "AI features" in settings) has two separate capabilities worth distinguishing:

Tab and page content access. When you invoke the assistant — or in some configurations, automatically when it detects a summarizable page — Chrome sends the visible text content of the current tab to Google's servers for processing. This includes text rendered by JavaScript, not just the static HTML, so single-page apps and logged-in dashboards are included.

Browsing history context. A separate setting allows Gemini to reference your recent browsing history to answer questions like "what was that article I read yesterday about X." This requires Chrome to retain and index a searchable history that the model can query.

Neither of these is disclosed prominently at the point of use. The toggle for both lives under chrome://settings/ai (or Settings → AI innovations, depending on your Chrome version). Google's stated policy is that page content sent for a specific query is not used to train models and is retained only briefly for abuse prevention — but "briefly retained on Google's servers" is still a meaningfully different exposure than "never left your device."

What Edge's Copilot Actually Sees

Microsoft's implementation works similarly with a few differences that matter for privacy:

Page content is sent to Microsoft's cloud when Copilot is invoked, including content from pages behind a login if you have Copilot's "page context" permission enabled — which, on a fresh Edge install, defaults to on.

Copilot can be tied to your Microsoft account. If you're signed into Edge with a work or school Microsoft 365 account, Copilot's activity may be visible to your organization's admin depending on your employer's Purview and compliance configuration. If you use the same Edge profile for personal and work browsing, this is worth checking before you assume anything you do in that profile is private from your employer.

The sidebar auto-suggests on certain pages. Depending on your settings, the Copilot sidebar can proactively offer to summarize or explain content without an explicit request, which means a content-analysis request may fire before you've consciously decided to use the assistant at all.

Disable this at edge://settings/copilot — separately from the general Copilot toggle, you'll also want to check edge://settings/privacy for a related "optional connected experiences" setting that governs whether page content can be sent to Microsoft even outside an active Copilot session.

How to Turn Both Off

Chrome:

Go to chrome://settings/ai
Turn off "Help me write," "Tab organizer," and any "Page content" or "AI Overviews" toggles listed under AI innovations
Go to chrome://settings/privacy and confirm "Make searches and browsing better" is off if you don't want query and history data used for personalization
Restart the browser — some AI feature flags don't fully disengage until restart

Edge:

Go to edge://settings/copilot
Turn off "Show Copilot icon in the toolbar" and "Let Copilot access this page"
Go to edge://settings/privacy and turn off "Optional connected experiences"
If your Edge profile is a work/school account, also check with your IT admin what org-level Copilot policies apply — a local toggle does not override an organization-wide enforcement policy set through Intune or Purview

Neither company makes these settings easy to find, and both re-enable some AI features after major version updates. Check both settings pages again after your next browser update.

Why "I Never Click the Assistant Icon" Doesn't Fully Protect You

The instinct is to assume that if you never deliberately open the sidebar or click the assistant icon, no data has left your browser. This isn't reliably true. Background summarization, tab-organizer previews, and "smart suggestions" features can trigger content analysis without an explicit click, depending on your specific settings and browser version. The only way to be certain is to check the toggles directly rather than relying on your own usage habits as a privacy boundary.

This matters most on pages you'd never think to screenshot or paste into a chatbot manually: your bank's transaction history, an HR portal showing salary bands, a client's Google Doc shared with edit access, a patient chart if you work in healthcare. A deliberate chatbot query requires you to decide to share something. A background browser assistant removes that decision point.

What to Use Instead

Disabling the built-in assistants doesn't mean giving up AI help while you browse — it means choosing tools where the data flow is explicit and where you decide, query by query, what gets sent anywhere.

For research and current-events questions, Perplexity is a better substitute than a browser-native assistant for one specific reason: it's a tool you deliberately open and query, not a background process reading every tab. You control exactly what text goes into a Perplexity search, the same way you'd control what you type into any search engine — the browser itself isn't silently forwarding page content to a second AI vendor you never chose to use.

Affiliate Disclosure: This article may contain affiliate links. If you make a purchase through these links, we may earn a small commission at no extra cost to you. We only recommend products we genuinely believe in. This helps support our work and allows us to continue providing free content.

Storing What the Assistant Would Have Summarized

If part of why you liked the browser AI assistant was quick summaries of long documents you wanted to keep — meeting notes, research PDFs, contracts — replace that workflow with something under your control instead of the browser's cloud pipeline.

Tresorit gives you zero-knowledge encrypted storage for the documents themselves. Save the source file to an encrypted Tresorit folder, then run your local LLM or a deliberate Perplexity query against it when you actually need a summary — rather than having a browser extension parse it automatically the moment the tab opens. This keeps the sensitive document encrypted at rest between the moments you actually need AI assistance, instead of exposed to a background process running continuously.

A Practical Setup

For a privacy-conscious browsing setup that still lets you use AI deliberately:

| Layer | Tool | Purpose |

|---|---|---|

| Browser AI (disabled) | Chrome/Edge settings | Turn off background page-content access |

| Deliberate research | Perplexity | Query-by-query, no background tab reading |

| Sensitive document analysis | Local LLM via Ollama | Nothing leaves your device |

| Encrypted document storage | Tresorit | Zero-knowledge vault for source files |

| Account identity | Proton Mail | Decouples browser account from AI vendor relationship |

This setup costs nothing beyond what you're likely already paying if you use Perplexity Pro or a paid Proton or Tresorit plan — the browser-native assistants are "free" in the sense that Google and Microsoft aren't charging you a subscription fee, but the cost shows up as page content flowing to their servers by default.

The Honest Bottom Line

Browser-native AI assistants aren't malicious — they're a reasonable product to build, and for someone who doesn't handle sensitive information in their browser, the convenience trade-off may be fine. But "reasonable product" and "should be on by default without clear disclosure" are different claims, and right now both Chrome and Edge ship the second without adequately supporting the first.

If you handle client data, financial information, health records, or anything else you wouldn't want summarized and sent to Google or Microsoft's servers as a side effect of having a tab open, the two-minute settings check in this guide is worth doing today — and worth repeating after your next browser update, since these features have a track record of re-enabling themselves.

Get the PrivateAI Browser Lockdown Checklist — every AI and telemetry setting worth disabling in Chrome and Edge, in one printable page. Enter your email below.

Subscribe to PrivateAI Updates