Have you ever wanted to browse the Internet but find typing URLs in the address bar a pain? Google is here to help. Today, the company announced significant enhancements to its existing Gemini in Chrome functionality. Highlights include a new look for AI companions, more integrated image editing tools, and perhaps most impressively (but also creepily) the launch of automatic browsing. This allows Gemini to take the wheel when you’re online.
New side panel view
Credit: Google
Previously, Chrome’s Gemini appeared in a small box at the top of the browser, which made it a bit inconvenient to use, especially when moving between tabs. Google’s update moves it into a slightly larger, scrollable side panel view that doesn’t obscure other content. Instead, it appears on the right side of the webpage you’re viewing, so you can easily compare what you’re seeing with the answers Gemini provides, and continue the conversation as you move back and forth between multiple tabs. All the same functionality as before is retained, including the ability to browse multiple open tabs in a prompt. It’s a small change, but should help with usability.
Right-click the image and edit it with Nano Banana
Credit: Google
Google’s Nano Banana image generation AI has been having a few issues, but Chrome’s new Gemini update makes it easier to use. Instead of downloading images and re-uploading them to Gemini, you can now edit them with just a right-click using Nano Banana. Alternatively, you can use natural language to start editing by showing the image you want to edit on-screen and instructing Gemini to edit in the side panel. Google says this should work with almost any image that can be displayed on a browser.
Google showed this off to journalists during a demo using Google Photo Library, but there’s nothing that says you have to stick to your own images. This immediately set off alarm bells for me, given that Elon Musk’s X is currently in a quagmire for using Grok to allow people to directly edit other people’s images on social media platforms without their permission. The tool was curtailed a bit after some users started using it to generate explicit content from other people’s photos, but Google doesn’t seem concerned. When asked about the security of this feature, a Google spokesperson said:
“We have clear policies that prohibit the use of AI tools to generate sexually explicit content, and our tools are continually refined to reflect these policies. We invested in safety from the beginning and added technical guardrails to limit problematic output, such as violent, offensive, or sexually explicit content.”
The company hasn’t said anything about how users can use Nano Banana in Chrome to circumvent copyright, but technically the new update doesn’t actually add any new features to Google’s AI image generator, it just makes it more accessible. Indeed, the same goes for Grok’s recent updates, and even with the best of intentions, easy access can mean opening the floodgates.
Auto-browse in Chrome using AI
Credit: Google
Last but not least, “Agentic” has become a hot buzzword in AI these days, and Google doesn’t want Chrome to be left behind. So Gemini can now not only answer your questions, but also control your browser.
Currently, this feature is limited to Google AI Pro and Ultra subscribers, but starting today, those subscribers will be able to ask Chrome to “auto-browse” so they can complete research, navigate to different websites, and fill out forms.
Watch the AI move around the web and click to go to different tabs while it works in the background. You can auto-browse multiple tabs at the same time, so you can perform several tasks at the same time. AI lists the steps to take during navigation in a side panel for easy check-in.
Google demonstrated this to journalists by showing the AI finding a specific product, navigating to its store page, singing into the buyer’s account (using Google Password Manager), and adding it to the cart. The company also suggested that you could use auto-references to schedule reservations, fill out online forms using information in uploaded PDFs, collect tax documents, and compare apartments listed on sites like Redfin. I can’t speak to how well it performs these tasks as I haven’t used it yet, but it seemed to work fine in a controlled demo.
What do you think so far?
Do you really trust AI to browse for you?
My concerns with auto-browsing are mainly with sketchy websites and permissions, but Google says they have plans for those as well. Autobrows will require your permission before accessing Google Password Manager, and if it encounters a link that the AI deems incorrect, it will likely use Chrome’s existing unsafe browsing protections to navigate you away. A Google spokesperson told me this feature is “as secure as it can be,” but I’d like to keep monitoring it, at least for the first few requests.
This feature also has one limitation at this time. Although you can have multiple tabs open at once, auto-browse tabs cannot communicate with each other. This means that each instance of autoreference is isolated, but that may change in the future.
Personally, I don’t think I use this much, especially for sensitive tasks like “gathering tax documents,” but the ability to automatically fill out basic forms seems useful. Google says automated browsing will stop and users will be asked to take over sensitive steps in tasks that may require automated browsing, such as actually making a purchase or submitting a form. The final step is not (or should not be) executed. This gives you a chance to see how it works. In that respect, it’s similar to the Gemini app’s existing shopping functionality.
Existing and upcoming features
Gemini in Chrome can use most of Gemini’s existing features, including connecting to apps like Gmail and accessing your chat history with bots. But there’s one big thing planned for “the coming months.”
Recently, the Gemini app rolled out a beta version of “Personal Intelligence” for paid users. This allows the AI to see all of your past conversations and connected apps without you having to tell it where to look. This is essentially an extension of the existing connected apps and history functionality, with an inference model applied on top of it. For example, when you ask Google to help you find new tires for your car, it looks through your Gmail and Google Photos to automatically recognize the model of car you have and when you last bought tires.
This feature is still in development, but being in development means that Google is working quickly to make all the different ways you can access Gemini equivalent. All other features described in this article are already available or currently being deployed.
