Google Chrome with Gemini Sidebar and Auto Browse

Chrome's AI layer has spent the past year feeling like a bolted-on afterthought: a floating box you summoned, asked something, and dismissed. The Gemini 3 refresh changes the geometry of that relationship. Gemini now lives in a persistent right-hand side panel, which sounds like a trivial UI change but is really the whole point — an assistant you can keep open while you work is a fundamentally different product from one you have to go fetch. Google is clearly aiming at the specialized AI browsers that have been nipping at Chrome's heels, and it is doing so from the comfortable position of already being on everyone's desktop.

Who is this for? Broadly, two groups. The first is anyone who routinely ends up with fifteen tabs open while comparison shopping or researching — the cross-tab comparison and review-summarization angle is the most immediately useful thing here, and it is available without paying extra. The second, much narrower group is U.S. Google AI Pro and Ultra subscribers who want to experiment with Auto Browse, the agentic mode that will actually go off and do multi-step work: pricing out flights and hotels, filling in forms, booking appointments, wrangling subscriptions. That second group is where the genuinely novel stuff lives, and it is also where the caveats pile up.

The Nano Banana integration deserves a mention as the sleeper feature. Being able to generate or transform an image inside the panel, without the download-edit-reupload shuffle, is the kind of small friction removal that people notice daily. It is also, notably, available to everyone rather than being paywalled.

Judged on ambition, this is the most serious thing Google has done with AI in Chrome. Judged on what most people can actually use today, it is considerably more modest. The free tier — sidebar, summarization, page and YouTube digests, image generation, cross-tab comparison — is a solid, sensible set of tools that makes Chrome competitive without asking anything of the user. That alone is worth the upgrade.

Auto Browse is the part that will decide whether this update is remembered. The multimodal shopping demo — identify an item from a reference photo, find similar products, cart them, apply codes, respect a budget — is a genuinely impressive description of an agent's workflow. But it is also exactly the sort of task where small errors compound into real-world consequences, and Google is asking for a lot of trust. The safeguards described are the right ones in principle: permissioned access to the built-in password manager, and a pause-and-confirm step before purchases or posts. I would want to live with those guardrails for a while before handing over credentials, and the fact that these behaviours are described rather than demonstrated at scale is reason for measured enthusiasm rather than excitement.

The access story is the clearest weakness. Auto Browse is paywalled behind AI Pro or Ultra and fenced into the U.S., European availability is limited, and Google has not committed to a Chrome version number for the full feature set — it is all "rolling out." For a browser whose entire advantage is ubiquity, shipping the flagship capability to a sliver of the user base is a strange look, and it means a lot of readers will install the update and find nothing much has changed. The promised Personal Intelligence memory feature is also still a roadmap item, so I have not factored it in.

Finally, an honest note on confidence: this assessment is based on announced capabilities and platform support rather than extended hands-on testing, and agentic browsing is precisely the category where the gap between demo and daily use tends to be widest. I am rating the direction and the free-tier substance highly, and withholding judgement on how reliably the agent actually performs.

Rating: 3.5/5 — a smart, well-targeted update with an excellent free baseline, held back by a heavily restricted headline feature and unproven agentic reliability.