ChatGPT in Chrome: Side Chat, Open Tabs, and Browser History Explained

ChatGPT is spreading out from its own website into the places you already work — starting with your browser. The official ChatGPT Chrome extension adds a Side Chat that can read the page you're on, reference your open tabs, and answer questions about a YouTube video. Meanwhile the desktop app is getting browser-native touches: URL suggestions as you type, access to your browsing history, and controls for how all of that is managed. Here's what actually changed, how to use it, and where the privacy tradeoffs sit.

What "ChatGPT around the web" means

Instead of copying text into a chat window, ChatGPT now sits beside the page and pulls context from what you're already looking at. OpenAI is expanding ChatGPT beyond its website with a Chrome extension that can understand the contents of the webpage you're viewing — you can ask questions about a page, summarize articles, explain concepts, or kick off longer tasks without leaving the browser (Digital Trends).

That puts ChatGPT in the same lane as Gemini in Chrome, which already lives in Chrome's side panel and reads the active tab. The bet from both companies is the same: AI assistants are becoming browser-native productivity tools rather than standalone chatbots.

In the Chrome extension: the Side Chat

The extension opens ChatGPT in a panel next to the page you're viewing, so you can ask about the page or continue into tasks that use its context alongside local files and connected apps (ChatGPT docs).

Ask about a YouTube video

Open a video, open the Side Chat, and ask for a summary, the key points, or a specific timestamp's argument — the assistant works from the page context instead of you scrubbing the whole thing. It's the fastest way to decide whether a 40-minute talk is worth your time.

Reference your open tabs

The Side Chat can use context from your open tabs when a task needs it. That means "compare the pricing on these three tabs" or "pull the shipping dates from the two orders I have open" without copy-pasting between windows. ChatGPT reads across the tabs you've got open and answers in one place.

Highlight text and ask

Select text on any page and ask about it directly — define a term, translate a paragraph, rewrite an email draft, or explain a dense clause. This is the highlight-to-ask pattern that browser assistants have converged on, now built into the official extension.

It carries over to the desktop app

Chats you start in Chrome are available in the ChatGPT app, and you can open recent ChatGPT chats in Chrome, so you can continue the same thread in either place. The panel stays with the tab where you opened it, so context doesn't jump around as you browse.

In the desktop app: URL suggestions, history, and controls

On the desktop side, the browsing experience gets more assistant-like:

• URL suggestions as you type — a faster path to the page you mean, surfaced as you start typing.
• Revisit your browser history — ChatGPT can use relevant history entries as context for a task, so "reopen that article I read yesterday about X" becomes a prompt instead of a scroll.
• Choose how it's managed — you decide what the assistant can touch, from website access down to whether history is in play at all.

History access is the sensitive one. Per OpenAI's documentation, browser history can include internal URLs, search terms, and activity across signed-in devices, so ChatGPT asks each time it wants to use history and scopes that access to the specific request — there's deliberately no always-allow option for history (ChatGPT docs).

The permission model, in plain terms

Because the extension can act in your browser, not just read it, the controls matter. A few things worth knowing:

• Per-site approvals. By default ChatGPT asks before it interacts with each new website, based on the site's host. You can allow once, allow for that site, allow for all sites, or decline. "Allow for all sites" is flagged as elevated risk for a reason.
• Allowlist and blocklist. In the desktop app under Settings, you can manage which domains ChatGPT can use without asking and which it should never touch.
• Untrusted page content. OpenAI's own guidance is to treat page content as untrusted and review a site before letting the assistant continue — a nod to prompt-injection risk when an AI reads arbitrary web pages.
• What's stored. OpenAI says it doesn't keep a separate complete record of your Chrome actions; browser activity is stored only when it becomes part of the chat context, such as text or screenshots the assistant reads (ChatGPT docs).

The short version: convenience goes up, and so does the surface area. Reading your tabs and history is powerful precisely because that data is sensitive.

How to set it up

1. In the ChatGPT desktop app, open the Plugins Directory and install Chrome (other Chromium browsers aren't supported yet).
2. Approve Chrome's permission prompts and confirm the Side Chat loads.
3. Open a page, select ChatGPT from the Chrome toolbar (on macOS, Cmd + Shift + .), and ask away.
4. For file uploads inside a Chrome task, enable Allow access to file URLs on the extension's details page.

If the Side Chat loads but Chrome tasks won't run, the usual fixes are updating the desktop app, restarting Chrome, confirming the Chrome plugin is on in the app, and making sure you're in the Chrome profile where the extension is installed.

Why this matters beyond ChatGPT

The browser is turning into the interface for AI agents. Whoever the assistant is — ChatGPT, Gemini, or something you run yourself — the pattern is the same: read the page, use the tabs, remember the history, and eventually take action on your behalf. The interesting questions are no longer "can it summarize this article" but "where does my data live, and who approves each action?"

For anything involving logged-in sites, internal tools, or client data, that's where local and permissioned beats cloud-by-default. A browser assistant that reads your history is convenient; an agent workflow you run on your own machine, with your own approvals, is auditable.

Do this with your own AI workforce

ChatGPT's browser features are great for quick, in-tab help. When the job is a repeatable, multi-step workflow — pulling data across tabs, filling forms on logged-in sites, then dropping results into a doc — you may want agents you actually control. Eigent is an open-source "Cowork" desktop app: a local multi-agent workforce that automates real browser and file workflows, with per-action permissions instead of blanket access.

If you're weighing the cloud-agent approach against a workspace you own, see our breakdown of ChatGPT Work vs Eigent and what ChatGPT Work actually does. Or just download Eigent and point a team of agents at your next repetitive task.