Google is adding its Gemini AI assistant to Chrome, the company announced at Google I/O on Tuesday.

Initially, Gemini will be able to “clarify complex information on any webpage you’re reading or summarize information,” according to a blog post from Google Labs and Gemini VP Josh Woodward. Google envisions that Gemini in Chrome will later “work across multiple tabs and navigate websites on your behalf.”

I saw a demo during a briefing ahead of Tuesday’s announcement. In Chrome, you’ll see a little sparkle icon in the top right corner. Click that and a Gemini chatbot window will open — it’s a floating UI that you can move and resize. From there, you can ask questions about the website.

In the demo, Charmaine D’Silva, a director of product management on the Chrome team, opened a page for a sleeping bag at REI and clicked on a suggested Gemini prompt to list the bag’s key features. Gemini read the entire page and listed a quick summary of the bag. D’Silva then asked if the sleeping bag was a good option for camping in Maine, and Gemini in Chrome responded by pulling information from the REI page and the web.

After that, D’Silva went to a shopping page on another retailer’s website for a different sleeping bag and asked Gemini to compare the two sleeping bags. Gemini did that and included a comparison table.

You’ll initially be able to keep a conversation going with Gemini as you navigate from tab to tab. But “later in the year,” Gemini in Chrome will let you select multiple tabs at once and ask a question about all of them.

D’Silva also showed a demo of a feature that will be available in the future: using Gemini to navigate websites. In the demo, D’Silva pulled up Gemini Live in Chrome to help navigate a recipe site. D’Silva asked Gemini to scroll to the ingredients, and the AI zipped to that part of the page. It also responded when D’Silva asked for help converting the required amount of sugar from cups to grams.

In Google’s selected demos, Gemini in Chrome seems like it could occasionally be useful, especially with comparison tables or in-the-moment ingredient conversions. I’d rather just read the website or do my own research instead of reading Gemini’s AI summaries, especially since AI can hallucinate incorrect information.

Gemini in Chrome is launching on Wednesday. It will initially release on Windows and macOS in early access to users 18 or older who use English as their language. It will be available to people who subscribe to Google’s AI Pro and Ultra subscriptions or users of Chrome’s beta, canary, and dev channels, Parisa Tabriz, Google’s VP and GM of Chrome, said in the briefing.

As for bringing Gemini to mobile Chrome, “it’s an area that we’ll think about,” Tabriz says, but right now, the company is “very focused on desktop.”

Correction, May 20th: Gemini in Chrome can keep a conversation going as you move from tab to tab; it doesn’t only work across two tabs, as we initially reported.