Getting cited in AI Overviews tells you almost nothing about whether AI Mode can see you.

Two Google surfaces answer the same query, agree on the answer about 86% of the time, and cite the same URLs only 13.7% of the time. That last number comes from Ahrefs, which compared 540,000 query pairs across AI Overviews and AI Mode. The gap is the whole story. A citation in AI Overviews is weak evidence, close to none, that AI Mode will surface you for the same search.

So a founder checking whether his brand shows up in Google's AI answers is really checking two systems wearing one logo. One still rewards classic ranking. The other mostly ignores it. Counting a win in the first as proof of the second is how a visibility report ends up confidently wrong.

This matters more than it did a month ago, because the surface that ignores your rankings is now the one most people land on by default.

What changed at I/O 2026

Google spent years treating AI Mode as a tab you opted into. At I/O on May 19, the company rebuilt the search box for the first time in 25 years and set Gemini 3.5 Flash as the default model in AI Mode for everyone globally. The redesigned box routes queries into that conversational experience from the first keystroke. AI Mode had already passed one billion monthly users, with queries more than doubling every quarter since launch.

The ten blue links did not die. Liz Reid said from the stage that classic results still exist, and Google's post repeats that you keep getting a range of results. What shifted is the default path. Analysts covering the keynote described it bluntly as AI Mode going default, and defaults decide where attention goes.

For anyone measuring AI visibility, the timing is the problem. The surface you were probably reporting on, AI Overviews, is no longer the one capturing the bulk of intent. The surface that is now capturing it plays by different rules.

Two surfaces running two different selection systems

Start with how each one picks sources. AI Overviews appears beside the organic results, and it still leans on them. Around 81% of AI Overviews include at least one page from the top-ten organic results for that query. Rank well and you stay eligible.

AI Mode behaves differently. It runs a heavier query fan-out, issuing many sub-searches at once Google's own examples show up to 16 simultaneous queries), then assembles a long answer from passages across that spread. Analysts tracking the new default peg the overlap between traditional top-ten rankings and AI Mode sources at roughly 17 to 36%. The citation pools the two surfaces draw from barely touch.

The format follows from that split. AI Overviews sits above the blue links as a short summary, and a strong ranking plus a concise, liftable answer keeps you eligible for it. AI Mode replaces the links with a long conversational answer, and what earns a place there is entity presence and passage-level coverage of the sub-questions, not raw position.

Source preference diverges too. The Ahrefs study found Wikipedia in 28.9% of AI Mode citations versus 18.1% in AI Overviews, with AI Mode naming roughly 2.5 times as many brands per answer and leaning on encyclopedic and reference pages, while AI Overviews shows a stronger pull toward video and community posts. AI Mode is not a longer version of the overview. It researches the query again, from a wider and more reference-heavy pool.

Why a citation in one says little about the other

There is one bridge between the surfaces, and it runs in a single direction. If your brand is mentioned in an AI Overview, Ahrefs found a 61% chance it also appears in AI Mode's longer answer, because AI Mode tends to keep the entities the overview surfaced and pile more on top. Useful, but notice the limit. That is an entity mention carrying over, not your URL getting cited, and it still leaves four in ten cases where the overview presence buys you nothing in AI Mode.

Run it the other way and the bridge collapses. AI Mode pulls in sources the overview never touched, so plenty of its citations have no overview footprint at all. A clean AI Overviews report can sit next to near-total absence from the surface most users now see first.

This is the reporting trap. You track AI Overviews because it was easier to see, you watch your top-ten pages keep showing up there, and you call AI visibility healthy. Meanwhile the default experience is citing competitors and reference sites you never measured. The dashboard is not lying. It is answering a question you stopped needing to ask.

What a top-ten ranking still buys

None of this means ranking stopped mattering. It means a ranking buys a narrower thing than it used to.

A top-ten position still earns AI Overviews eligibility, and that surface has not gone away. It still drags click-through down by around 34.5% when it appears, so presence there protects whatever clicks remain. Ranking also keeps you in the live retrieval pool that every answer engine reads from, which is upstream of all of this.

What a ranking no longer buys is a seat in AI Mode. The fan-out reaches past the top ten, weights reference and entity signals the overview cares less about, and assembles answers where your organic position is one weak input among many. You can rank first for a query and stay invisible inside the answer most people read for it. The top-ten overlap on AI Overview citations itself has slid from 76% in mid-2025 to 38% by early 2026, and AI Mode sits further down that same slope.

Tracking both surfaces without enterprise tooling

You do not need a platform contract to see the divergence. You need to stop treating one number as both.

Pick ten queries that actually drive your business. For each one, run it twice in Google, once landing in AI Mode and once on a result that shows an AI Overview, and log which domains get cited in each. Do it monthly. Two columns, same queries, and the overlap (or the lack of it) shows up fast. Free tools help on the edges. Ahrefs Brand Radar lets you run a blank cited-domains search and filter AI Overviews and AI Mode separately, and SEO.com's AI Overview checker covers the overview side.

Know the blind spot in Google's own data while you do this. Search Console folds AI Overviews and AI Mode touchpoints into ordinary Search metrics, and you cannot filter them apart. Impressions and clicks from both land in the same bucket as classic search, so GSC will never tell you which surface is carrying you. Manual logging is crude, but right now it is the honest version of the picture.

The structural fix a rankings report can't see

If one change earns a place in AI Mode without ever moving a ranking, it is writing self-contained passages that each answer one specific sub-question.

The mechanism is the fan-out. AI Mode breaks a query into many smaller searches and retrieves passages rather than whole pages to answer each fragment. A page built as one long argument can rank beautifully for the head term and still offer no clean passage that resolves a sub-question the fan-out asked. So it never gets pulled. Restructure the same content into sections that each state a question and answer it completely, with the entities and specifics named in the passage itself, and you become retrievable across the spread of sub-queries the head keyword hides.

A rankings report cannot show you this. It tracks page-level positions for head terms. The passage that wins a fan-out fragment lives below that resolution, doing work no position column will ever display. You can pass every check in your SEO tool and still be missing from the answer your customer reads.

The reporting habit to break is older than AI Mode itself. As AI Mode settles in as the default, the teams that win it will be the ones who measured it on its own terms early, while everyone else was still reading an AI Overviews number and assuming the rest.