AI Overviews Now Expand Into AI Mode by Default (August 2026): What Solo Founders Should Do When the Blue Links Drop Below the Fold

On August 27, 2026, Chris Long posted a recording of Google loading the AI Mode “Ask anything” box inside an ordinary AI Overview, with no “Show more” button. Barry Schwartz replicated it the next day at Search Engine Roundtable and got Google to confirm that AI Overviews now dynamically expand on some queries, pushing the blue links far down the page.

What changed in Google’s AI Overviews on August 27–28, 2026?

For a subset of queries, the AI Overview no longer loads as a truncated snippet with a “Show more” control. It loads the full answer plus the AI Mode follow-up prompt, already open. A Google spokesperson told Search Engine Roundtable that AI Overviews “may dynamically expand” where its systems judge that most useful; Robby Stein of Google added that the expansion stops if a searcher has already begun scrolling.

Two details matter. This is a test in progress, not a confirmed permanent default — Google called it dynamic and query-dependent, and no coverage percentage has been published by anyone. And the expansion fires on page load, before the searcher has done anything, which is exactly the moment your listing either exists on screen or does not.

How does a dynamic expansion actually push your listing down?

It does not touch your ranking at all. It changes how many vertical pixels sit above you. A collapsed AI Overview on a phone occupies maybe a third of the first screen; an expanded one plus an open prompt box can occupy two or three full screens before the first organic link appears.

That distinction is the mechanism, and it lets you predict cases nobody has written about yet. Ask two questions of any query: does a synthesized answer serve it better than a list of links, and does the searcher have an obvious follow-up? Where both are yes — comparisons, definitions, multi-step how-tos, “is X worth it” — expect expansion. Where either is no — a brand name, “open now near me,” a SKU — expect the classic layout. Your position 3 is worth what it was last month on the second group and much less on the first, and no rank tracker shows the difference.

Why does positioning decide what to do about this?

Because the constraint is now space, not effort. Positioning — the school Al Ries and Jack Trout built around owning one distinct slot in a prospect’s mind rather than competing on completeness — is the right lens precisely when the surface will not show a comprehensive answer. A generative summary quotes narrow, checkable, attributable claims. It does not quote a 3,000-word guide that covers everything adequately and nothing distinctly.

Read through a coverage lens, the right response is “publish more, chase more citations” — which a two-person team cannot execute and would not benefit from. Read through positioning, it is the opposite: pick the narrow question you can answer better than anyone in your category, answer it with a number or a rule a model can lift verbatim, and accept invisibility on broad queries. That is less work, which is the only reason it is available to you.

Does this make paid search more valuable or less?

More valuable in the short term, and more expensive because of it. Text ads above the AI Overview are the only placement on an expanded results page that is guaranteed to sit above the fold, which makes them the last predictable piece of real estate on an informational query.

Here is the arithmetic that should worry a small advertiser. Take a two-person B2B SaaS selling scheduling software, running $2,000 a month at a $6.80 average CPC, and getting 40 trials a month free from blog posts that rank for how-to queries. Those are precisely the queries that expand. If half those clicks stop arriving, replacing 20 trials at a 4% landing-page conversion rate takes 500 more paid clicks — about $3,400 a month, which is 170% of the entire existing ad budget. The general form is worth writing down: replacement cost = (lost conversions ÷ landing page conversion rate) × CPC. Run it on your own numbers before you decide this is a story about publishers.

Third-party trackers have estimated ads on roughly a quarter of AI Overview results in 2026, up from low single digits in early 2025. That is an outside estimate, not a Google figure — read it as a direction, and the direction is that the ad slot follows the eyeballs down the page.

Is Google’s explanation the right way to read this?

Partly, and the omission is what costs you money. Google’s stated reason is that users find Search more helpful this way and engage more deeply in follow-up exploration. That is probably true and beside the point for a small business, because deeper follow-up exploration describes a second prompt typed into Google, not a visit to your site. The metric being optimized is engagement inside the answer, and every unit of it used to be a click.

The scroll-cancel behavior is the tell. Google built a rule so the expansion will not disrupt someone already reading — which means Google knows it is disruptive, and shipped it anyway for the moment before the searcher starts reading. That is a considered decision about where attention should go. Disagree with the framing if you like, but plan around the decision.

How should a three-person team respond in the next seven days?

A week of work with a spreadsheet and Search Console, not a quarter of platform migration. In this order.

1. Export your top 50 queries from Search Console for the last 90 days, sorted by clicks. That is your exposure list.
2. Tag each query “summarizable” or “not.” Comparisons, definitions, how-to and “should I” phrasings go first; brand terms, “near me,” and anything with a SKU go second. One judgment call per row.
3. Total the clicks in each bucket. Above 40% in the summarizable bucket is a concentration problem. Under 15%, file this and move on.
4. Run the replacement-cost formula on that bucket at a 50% loss assumption. The output is the monthly paid budget you would need to stand still — and that number, not the traffic chart, is what to react to.
5. Rewrite your top three summarizable pages to lead with an answer. One direct sentence under each heading, containing a number, a date, or a named rule. Generative surfaces lift extractable claims, not throat-clearing.
6. Add one claim to each page that only you can make. A four-person bookkeeping practice publishing real monthly fee ranges by client size is citable; one offering “affordable bookkeeping tailored to your business” is not.
7. Bid defensively on your five highest-intent non-summarizable queries. Those results pages have not changed, and they are where a small budget still converts predictably.
8. Set a reminder for October 1 to re-pull the same 50 queries. One measurement now, one in five weeks. Checking daily just reads noise.

Before you attribute any September traffic drop to this change, rule out the alternatives — the same discipline we laid out for separating a real algorithmic hit from a reporting glitch after the August spam update applies here without modification.

What dates should be on your calendar between now and Q4?

• September 1–30, 2026: the Google Ads AI Max auto-migration window. If your Search campaigns change behavior next month, this is a likelier cause than the AI Overview expansion, and the two are easy to confuse — we covered how to tell the upgrade from everything else in your account on August 25.
• October 1, 2026: Google Local Services Ads begins charging for some missed calls, per Search Engine Land. If you run LSAs, that answer-rate audit and this query audit belong in the same sitting.
• October 1, 2026: your five-week re-pull of the query list. Compare bucket by bucket, not in aggregate.
• November–December 2026: seasonal auction pressure. If the step-4 number was uncomfortable in August, recompute it at Q4 CPCs before committing a holiday budget.

Frequently asked questions

Is the “Show more” button gone for good?

No. Google described the expansion as dynamic and limited to queries where its systems judge the fuller answer more useful, so collapsed AI Overviews with a “Show more” control still appear. Treat it as a rollout in progress. What has changed is that you can no longer assume the collapsed version is the default state on informational queries.

Will this show up in my Search Console impressions?

Expect impressions to hold up better than clicks, which is the signature of a visibility problem rather than a ranking problem. If your average position is stable, impressions are flat or up, and clicks fall on a specific set of query types, layout is the likeliest explanation. A genuine ranking loss usually moves average position and impressions together.

Should I pause my blog and move the money to ads?

Not on this news alone. Content still earns citations inside the AI answer, and citations still carry brand recognition into the auction later. What should change is the shape of what you publish: fewer broad guides, more narrow pages that answer one question with a specific, attributable claim. Reallocate effort within content before reallocating budget out of it.

Does this affect local businesses the same way?

Less severely, and for a structural reason. “Near me” and other local-intent queries return map packs and location results rather than synthesized explanations, so they are poor candidates for expansion. The exposure for a local business sits in its informational content — the cost and how-to pages written to attract people who have not chosen a provider yet.

The one thing worth deciding this week

Not whether to fight the layout — you cannot. The decision is which single question you intend to be the obvious answer to, narrow enough that a model has a reason to name you and a searcher has a reason to type your brand into the follow-up box. Everything on the seven-day list serves that decision. Tracking which queries have quietly changed shape is the kind of measurement campaignpress.ai makes routine, but the positioning call is yours.

A team of three cannot out-publish a summarizer. It can be the source the summarizer cannot paraphrase without crediting.