AI Mode and AI Overviews Are Not the Same Surface

Most advice about ranking in Google's AI results is written as though there is one AI result to rank in. There are two, they behave differently, and the overlap between what they cite is far smaller than anyone expected.

The two surfaces, briefly

An AI Overview is the generated summary that appears above the traditional blue links. The links are still there underneath. A user can read the summary, ignore it, or scroll past it to the results.

AI Mode is a separate, end-to-end conversational search experience closer to ChatGPT. It cites sources, but there is no list of ten blue links beneath it to scroll to. The answer is the page.

Why the overlap is so low

AI Overviews are generated against a query that a person typed. AI Mode decomposes a question into several sub-queries, runs them, and synthesises across the results — a technique usually described as query fan-out. It is therefore pulling from a much wider and stranger set of pages, many of which would never rank on page one for the original phrasing.

That is genuinely good news for a small site. Fan-out rewards pages that answer a narrow sub-question completely, and those are exactly the pages a specialist can write and a large competitor cannot be bothered to.

What the traffic actually does

Two findings that sound contradictory but are not. When an AI Overview appears, click-through to the top organic result falls sharply — independent analyses have put the drop somewhere around 60%. But sites cited inside the Overview see materially more clicks than they would have got from an ordinary first position.

So the summary is not destroying traffic evenly. It is concentrating it. Being cited is worth more than it used to be, and not being cited is worth much less. There is no longer a comfortable middle.

How exposed is your industry?

Coverage is wildly uneven, and this is the first thing to check before panicking or relaxing. B2B technology queries draw AI Overviews something like 70% of the time. E-commerce product queries draw them around 4% of the time, because Google would rather show you shopping results it can monetise.

Local service queries — plumbers, accountants, clinics — sit in between, and are increasingly answered by the map pack rather than by an AI summary. If you are a trade in the GTA, your exposure to this whole problem is lower than the headlines suggest, and your Google Business Profile matters more than your blog.

What to actually do

1. 1Find out whether your queries even trigger AI results. Search ten real customer phrasings. If nothing generates a summary, spend your time elsewhere — this is not your bottleneck.
2. 2Write the direct answer first. Both surfaces lift short, self-contained passages. A paragraph of 40 to 60 words that answers the question with no preamble and no dependency on the sentence before it is the unit that gets extracted.
3. 3Answer the sub-questions, not just the headline question. Fan-out means the page that comprehensively covers one narrow thing gets pulled into answers about broader things.
4. 4Publish real numbers with dates attached. Models cite specifics because specifics are checkable. "Pricing is bespoke" gives an engine nothing to quote.
5. 5Track citations, not just rankings. Position tracking says nothing about whether you appear in either AI surface. If you have never checked, you do not know.

What not to do

Do not publish forty thin pages to increase your odds of being cited. Scaled low-value content is precisely what Google's spam policies now target, and a new domain has no authority to absorb the hit.

Do not buy a tool that promises AI Mode rankings. Nobody sells access to that. What is on offer is measurement, and you can approximate it for free by asking the questions yourself and reading who gets cited.

And do not rebuild your site for this before checking your exposure. A local business with 4% AI Overview coverage on its money queries has better things to do with a fortnight.