How many tabs do you have open right now? Ten? Forty? Be honest. Now picture an assistant that can read all of them, then fill in the boring form sitting in tab number twelve. That’s the pitch behind the latest ChatGPT Chrome extension update, which OpenAI announced on 15 October.
OpenAI says it’s “bringing the power of computer use to anyone who uses Chrome.” In plain terms, computer use means the AI can click, type and move around a screen the way you would. So, is this the moment AI agents finally go mainstream? Let’s look at what you actually get.
So what changed in the ChatGPT Chrome extension?
The extension now comes with two modes, Chat and Work. They sound similar, but they do very different jobs.
Chat is the gentle one. In OpenAI’s words, “If you’re in Chat, it can look across your open tabs and gather info.” So you can ask a question and let ChatGPT pull context from the pages you’re already reading.
Work is the bold one. Here, “ChatGPT can control your Chrome tabs to get things, like sending messages, filling out forms, or downloading files,” as 9to5Google reported. It can also tidy that chaotic pile of tabs into groups, which might be the most useful trick of all.
There’s also a redesigned sidebar. 9to5Google compares it to the Gemini panel already built into Chrome. And because “your entire ChatGPT history is right there,” you don’t have to jump between windows to find an old chat.
If you’re new to the idea, our explainer on what an AI agent is covers the basics. In short, an agent is software that takes steps toward a goal, instead of only answering questions.
What can Work mode actually do?
OpenAI shared a few sample prompts. One of them reads: “Use the details in this event brief to fill out the registration form, and stop before submitting.” That last bit is smart, since you still get the final say.
The Chrome Web Store listing goes further, according to Unite.AI. It mentions research on live and signed-in sites, plus creating and editing documents, spreadsheets and slides. It also lists data entry, filing expenses, bulk downloads, website testing and tab cleanup.
So who’s this for? Honestly, anyone who spends half the day copying details from one page into another. Think expense reports, event sign-ups or moving data into a spreadsheet. Those jobs aren’t hard. They’re just tedious, and that’s exactly where an agent can save you time.
One small catch: setup runs through the Computer Use section in the ChatGPT desktop app’s settings. Also, OpenAI says the extension is “available for all ChatGPT plans.” But it isn’t fully clear yet whether Work’s tab control reaches every plan, because Work first rolled out to paid tiers in July.
Wait, is it safe to let an AI drive your browser?
Good question, and it’s the one you should ask first. Your browser is where you’re logged into email, work tools and maybe your bank. Unite.AI notes that ChatGPT can act on sites where you’re already signed in, such as LinkedIn, Salesforce and Gmail.
OpenAI has built in some brakes. By default, ChatGPT asks before it touches each new website. You can pick Allow once, Allow for this site, Allow for all sites, or Decline. OpenAI flags “Allow for all sites” and browser-history access as Elevated Risk. It also always asks before it reads your history or downloads or uploads files.
Then there’s prompt injection. That’s a trick where a web page hides instructions meant to hijack the AI. OpenAI’s own advice is to keep “treating page content as untrusted” and to avoid “sending secrets through browser tasks unless they are required and reviewed.”
For a sense of scale, PPC Land points to Anthropic’s own testing of its Chrome agent. Prompt injection worked 23.6% of the time without safeguards and 11.2% with them. OpenAI hasn’t published similar numbers for this update. We’d like to see them.
Why OpenAI moved into Chrome instead of its own browser
Here’s the backstory. OpenAI launched its own browser, Atlas, in October 2025. Then, on 9 July 2026, it said it was deprecating Atlas, which was set to stop working on 9 August.
So instead of asking you to switch browsers, OpenAI is going where you already are. The extension’s store page lists 7,000,000 users. Since 31 August it has also worked in Edge, Brave, Opera and Vivaldi, though Opera doesn’t get the side chat.
That’s a sensible pivot. Most people don’t want to change browsers for one feature. But it also puts ChatGPT right next to Google’s own Gemini tools, inside Google’s own browser. That’s a bold place to pick a fight.
This also fits a pattern we’ve noticed. Earlier this month, ChatGPT started answering with buttons and maps, as we covered in our post on GPT-6 Intelligent UI. The chat box is slowly turning into something that does things for you.
Okay, but should you turn it on?
Our take? Try Chat mode first. It’s low risk, because it reads your tabs instead of acting in them. You’ll quickly see whether the extra context actually helps.
If you try Work mode, start small. Pick a boring task on a site with nothing sensitive, then watch what it does. Use “Allow once” rather than “Allow for all sites,” at least at the start. And keep prompts like “stop before submitting” in your toolkit.
Also, read the install prompt carefully. According to Unite.AI, Chrome’s install prompts can include access to data on all websites, your browsing history, bookmarks and downloads.
For small teams and startups, this could be handy for admin work, since nobody got into business to fill in forms. Still, set some house rules first. Decide which sites are off limits, and don’t let it near anything involving passwords or payments.
Agents can also wander. We wrote about OpenAI agents poking around Wikipedia, and that story is a good reminder to keep a human in the loop.
Where the ChatGPT Chrome extension goes from here
The ChatGPT Chrome extension turns a chatbot into a helper that can actually click things. That’s a real shift, and it’ll probably save plenty of people a lot of dull work.
But the trade-off is simple. The more an AI can do in your browser, the more you have to trust it. Right now, the safety story leans on permission pop-ups and your own good judgment.
On privacy, OpenAI says it only stores browser activity when it becomes part of your ChatGPT context, and Computer Use follows your Memories setting. That’s reassuring, but you’ll still want to check those settings yourself.
So here’s what we’ll watch next. First, whether OpenAI shares real safety test results. Next, whether Work mode reaches free users. Finally, how fast Google answers back.
And here’s the question for you. Which boring browser task would you happily hand to an AI tomorrow, and which one would you never let it touch?