AI Mode passed a billion monthly users this year, with query volume more than doubling every quarter since launch. That's the headline everyone repeats. The eligibility rule behind that number takes one paragraph to explain, and it's public. Everything that happens after a page clears that bar is a mix of real documentation and industry guesswork, and the two get repeated interchangeably more often than they should. This piece separates them, section by section.
Key Takeaways
Q: What does Google actually require for a page to appear in AI Mode?
A: The same requirement as classic Search: indexed and eligible to show with a snippet. Google states there are no additional AI-specific requirements.
Q: How fast has AI Mode grown?
A: Google reported more than 1 billion monthly users at I/O 2026, with query volume more than doubling every quarter since launch.
Q: Can a brand see its own AI Mode impressions?
A: Yes, through Search Console's generative AI performance report, though it currently shows impressions only, with no clicks, position, or query data, and no split from AI Overviews.
Q: Does the well-known 8 percent click-rate statistic describe AI Mode?
A: No. That figure comes from Pew Research and measures AI Overviews specifically. No independent study of comparable scale has been published for AI Mode.
Q: Can a brand opt out of AI Mode without leaving classic Search?
A: Yes, through a Search Console control, but it applies to AI Overviews, AI Mode, and Discover's AI features together. There's no way to exclude AI Mode alone.
What AI Mode Actually Is
AI Mode is a tab inside Google Search. Ask it something and it answers with a generated response instead of handing you a list of links to sort through yourself.
Google runs it on a custom build of Gemini, wired into the same index, ranking systems, and safety infrastructure as classic Search. That matters more than the growth number does. AI Mode isn't a separate product a page has to earn its way into. It's a different way of reading the same index, which is also why almost everything below traces back to ordinary Search fundamentals rather than a new rulebook.
AI Mode vs AI Overviews vs Classic Search
AI Mode is the third surface a query can land on, and the three behave differently enough to budget for separately. The distinction that matters sits under the interface: which system decides what the reader sees, and where a link lands when it does.
Surface | What Triggers It | What the Reader Sees | Where the Links Sit | Search Share & Click Rate, Mar 2025 | Opt-Out Control |
|---|---|---|---|---|---|
Classic Search | Every query | Ranked list of links | Each link holds its own position | 82% of searches; 15% click rate | None short of noindex |
AI Overviews | Google judges a summary additive; often skipped | Short generated summary above the results | Inside or beside the summary; one shared position | 18% of searches; 8% click rate, 1% on a link inside the summary | Search generative AI control (also removes AI Mode, Discover) |
AI Mode | User opens the tab, or follows up from an Overview | Conversational response with follow-ups, files, images, voice | Inline in the response; standard page-position scoring | Not measured in this sample | Same control; can't be applied to AI Mode alone |
Search share and click-rate figures are Pew Research's March 2025 browsing data. Position and opt-out behavior are Google's own documentation.
The Eligibility Bar Is Lower Than Everyone Thinks
Google's documentation is direct: a page must be indexed and eligible to show with a snippet in Google Search, and there are no additional technical requirements beyond that.
No AI sitemap. No dedicated schema. No submission form. If a page already qualifies for a normal snippet, it qualifies for AI Mode. Meeting the bar gets a page into consideration. It doesn't get it cited, and Google is explicit that meeting every requirement still doesn't guarantee crawling, indexing, or serving.
How Google Decides What Goes Into a Given Answer
Google's own documentation describes the mechanism behind a single answer plainly: a longer or more complex question can be split into several smaller searches, run at the same time across different sources, then combined into one response. Google adds that this lets it surface a wider, more diverse set of supporting links than a classic search would.
That changes what ranking well even means for this surface. A page can get pulled into an answer through one of those smaller searches, for a question it was never written to target directly. The page that wins the citation isn't always the page built around the obvious keyword. It's whichever page happens to answer the narrower question sitting underneath it.
Why Two Identical Searches Don't Return the Same Answer
Run the same prompt twice and you can get two different answers back. Google documents at least three reasons why, and none of them is your page.
Personalization. Google expanded Personal Intelligence into AI Mode this year, letting it draw on Search history and, with a subscriber's consent, Gmail and Google Photos, to shape a response.
Model choice. Some users can select a different underlying model than the default, and Google says the two behave differently on the same question.
How the question is worded. Google's own documentation gives the example directly: "Why is an iguana a good pet?" versus "Why isn't an iguana a good pet?" can return different responses.
That's the practical case against screenshot evidence. One prompt, run once, signed in, on one device, tells you almost nothing you can act on, and it tells a client even less.
The Stat Everyone Quotes Describes a Different Feature
People click a normal search result in only 8 percent of visits when an AI summary shows up, against 15 percent when no summary appears, according to Pew Research's tracking of the actual browsing activity of 900 US adults. Sessions also ended entirely more often, 26 percent of the time on pages carrying a summary, against 16 percent on pages with only classic results.
Worth being precise about what this covers. It's AI Overviews, the short summary above the results. Not the AI Mode tab. Nobody has published independent research at that scale for AI Mode specifically, even though the two features share infrastructure. Most coverage online treats them as interchangeable. They aren't, and until someone runs the AI Mode version of Pew's study, that's a real gap in what anyone can say with confidence.
The Reporting Gap Search Console Hasn't Closed
Google shipped a generative AI performance report inside Search Console in June 2026, showing impressions, pages, countries, devices, and date-level trends for AI Overviews and AI Mode together. Real, first-party, free. Rolling out gradually, so not every property has it yet.
What it won't give you: clicks, position, or which query triggered the impression. And it doesn't separate AI Mode from AI Overviews, so a number moving in that report tells you the family moved. Not which member. If your reporting stack treats that report as a complete measurement of AI Mode, it's answering a narrower question than it looks like it is.
The Opt-Out Control, and Its One Real Limitation
Google also added a Search generative AI control under Settings in Search Console, letting a property choose to include or exclude its content from AI Overviews, AI Mode, and Discover's AI features. A third option lets a property inherit the setting from its parent.
Google states plainly that the control isn't used as a ranking or inclusion signal anywhere else in Search, so flipping it doesn't touch how a page performs in classic results. The real limitation is scope: it's all three AI surfaces or none of them. There's no way to keep a page inside AI Overviews while excluding it from AI Mode.
What Actually Moves the Needle, and What Doesn't
Google's own list for AI features is the same list it's had for ordinary SEO for years:
Crawling allowed, with no blocks at the robots.txt or hosting layer
Content reachable through real internal links, not buried three clicks deep
Important information in actual text, not locked inside an image
Structured data that matches what's visible on the page, nothing invented
Merchant Center and Business Profile data kept current
None of that is new. What's more useful is what's missing from it.
Google states plainly that there's no special schema, no AI-specific markup, and no separate machine-readable file needed to appear in these features. There's no required question-and-answer format either. Google publishes no format preference for AI features. There's no submission route. Indexing is still the only queue, and it's the same one it's always been. Anyone selling a technical AI Mode package beyond the list above is selling something Google hasn't documented.
What to Actually Do About It This Quarter
Stop treating Search Console's impressions number as a full measurement. It answers whether you're showing up, not whether it's working.
Keep the same prompt list for at least a full quarter before drawing conclusions. Personalization and model choice already add enough noise.
If you're deciding on the new opt-out control, remember it's all three AI surfaces or none. There's no way to keep AI Overviews and drop AI Mode alone.
Treat any single screenshot as an anecdote, not a data point. Sample across sessions and dates before reporting a trend to anyone.
If you want a starting point for that audit rather than building one from scratch, Geckocheck's Brand Intelligence tracks your presence across AI Mode and the other models that matter for your category, and updates as Google's own reporting keeps changing shape.
