Updated September 27, 2026. This article first said Chrome Auto-Browse would arrive on Android on by default, on hundreds of millions of phones. It is opt-in and limited to paying Google AI Pro and Ultra subscribers in the US on phones running Android 12 or later, and the sentences below are corrected to say so.
Two of the world's biggest platforms put an AI agent inside the phone this month, and only one of them is coming to your website to do anything. That is the distinction that decides where your effort goes, and it is the one the launch coverage is skipping. Apple introduced its new Siri AI at WWDC on June 8. Google's Chrome Auto-Browse reached Android at the operating-system level in summer 2026. Both are real, and they bet on opposite surfaces. One reads the web to answer you. The other visits websites to act for you.
Two agents, same month, opposite surfaces
Within a few weeks, both dominant phone platforms made an AI agent part of the operating system. The coverage is scoring it as an assistant race: whose voice is more natural, whose answers are smarter. For anyone who runs a website, that is the wrong scoreboard. The question that matters is narrower. Which of these agents actually arrives at my website and tries to get something done. Only one of them does, and knowing which one tells you exactly what to prepare for.
HOW AI IS CHANGING THE INTERNET: THE GOOD, THE BAD, AND THE POINTLESS
Nobody asked us if we wanted the internet changed. So every week I send one email about one thing AI did to it, with the receipts, and what to do about it. Sometimes the honest answer is "not yet."
From someone who pays attention to this for a living.
What Apple's Siri AI actually does
Apple's own newsroom describes the new Siri AI in three moves. It can "get things done across apps with even more systemwide app actions." It can "answer questions related to the content on a user's screen." And it can "go out to the web to get up-to-date information using broad world knowledge and generate a helpful answer."
What Siri does is act inside apps, read what is on your screen, and go to the web for information to compose an answer. What it does not do, in Apple's own description, is navigate your website and complete a task on it. It is a reader of the web and an actor inside apps. It runs on-device and on Apple's Private Cloud Compute, on a custom model from Google's Gemini family, and it began rolling out in beta on September 14.
So Siri can absolutely surface your content. The path it takes to you is the same discipline as being a clean source any AI answer engine can read: well-structured, server-rendered content it can pull into an answer. That is a real relationship, and worth keeping. It is not the same as an agent showing up to do a job.
What Google's Auto-Browse actually does
Google's Chrome Auto-Browse is the other kind of agent entirely. It fills forms, books appointments, reserves parking, and runs comparison shopping by driving the browser the way a person would. In summer 2026 it landed on Android at the operating-system level, arriving inside Gemini in Chrome on Android rather than as a separate app or extension.
Two things make that different from every agent before it. It is built into the phone, and it has system-level authority. It does not wait for someone to install an extension or open a separate app, though it does wait for a paying Google AI Pro or Ultra subscriber in the US to opt in. It is the browser, and it visits websites to act. If you want the fuller map of who else is in this category, I keep a running landscape of agentic browsers, but Auto-Browse is the one that changes the math, because it comes built into Chrome on Android.
Only one of them shows up at your website to act
Apple's agent reads the web for answers and acts inside apps. Google's agent visits your website and completes tasks on it. For a website owner, that is the whole difference, and it is bigger than any feature comparison.
The real shift this month is bigger than two assistants getting smarter: the agent-as-visitor, the machine that lands on your website to finish a job, now ships with the phone instead of as something you install. It is still an opt-in: Auto-Browse on Android is for paying Google AI Pro and Ultra subscribers in the US who turn it on, not a default. The visitor class that used to be a power-user edge case moves closer to ambient, and the version of your website it meets is the one you have not tested against a machine yet.
One more detail is worth holding, because it says something about who holds the leverage. Both agents run on Gemini. Apple licenses a Gemini-derived model, and Google's Auto-Browse is Gemini in Chrome. Google is the model under both front doors, which does not soften the access question so much as sharpen it.
What to do now
You have two moves now that Auto-Browse is on Android.
First, run your highest-value task flows against an agent yourself. Open your own checkout, your booking form, your lead capture, and try to complete each one the way a non-human would, without leaning on visual cues. Where it breaks is where Auto-Browse will break for your customers.
Second, stay a clean source for the reading agent. Server-rendered content and clear structure are what let Siri go out to the web and come back with you in the answer instead of a competitor. The two agents ask for two different things from your website, and you can serve both, but only if you know which is which.
What you should not do is file this under assistant features to watch. One of these agents completes tasks on your website for any subscriber who turns it on. That is not a feature. That is a new visitor with a job to finish.
The race everyone is scoring is which assistant is smarter. The question for your website is narrower and more urgent. Which agent shows up to act, and does anything happen when it does. Right now that agent is Google's, it has been on Android since the summer for the subscribers who opt in, and the version of your website it meets is the one you have not tested.