TL;DR
- Ranking still matters, but it carries less of the click than it used to. Pew Research tracked 68,879 Google searches by 900 US adults in March 2025 and found 8 percent click on a traditional result when an AI summary was present, versus 15 percent when none appeared.
- Ahrefs measured a 34.5 percent reduction in position-1 click-through rate when an AI Overview is present, across 300,000 keywords.
- The goal worth chasing is selection, not rank alone. Selection means your page is one of the handful the AI summary quotes. Rank is now a strong prior on selection, not a guarantee.
- Ranking still matters as a floor. A page that does not rank anywhere on page one is rarely retrieved at all, and brand-name plus transactional queries still draw the click on rank one.
- Before chasing rank one, run the two free checks — read the AI summary on your target query to see if you are already cited, and compare your rank changes in Search Console to whether the clicks rose with them.
You open Google Search Console for a page you worked hard on, and the position chart shows you crept up from five to three.
The clicks line, the one underneath it, did not move. You hover over a few days to be sure. The line is flat.
That gap is what this article is about.
The headline number is real, and the headline number does not tell the whole story for a 2026 small site. A Pew Research study of 68,879 Google searches found that when an AI summary appeared at the top of the results, only 8 percent of visits ended in a click on a traditional search result, compared with 15 percent of visits when no AI summary appeared. Ranking still moves the needle, and the needle measures less than it did.
What does ‘ranking’ actually mean in 2026?
Ranking still names a position in an ordered list of blue links beneath the search box.
The thing that changed is what sits above that list. Half a screen of AI summary, a People Also Ask block, a Featured Snippet, and a Local Pack now sit between the user and position one on a typical informational query. The same rank one as 2018 reaches the eye of a different reader.
The user who scrolls past all of it to click your blue link is a different person than the one who clicked your blue link three years ago. They scrolled past an answer that was good enough for most people, and they came to your page anyway. That click carries information your rank chart cannot show you.
So when you read your Search Console position chart, you are reading two things at once. Where you sit in the ordered list. And how much of the screen the user has already consumed before they reach you.
The first number is still useful. The second number is the one that explains why your clicks did not move.
Why does position one carry less weight than it used to?
Two pieces of evidence sit at the centre of the answer, and they were published a year apart by two independent teams.
The first is the Pew Research finding that users who saw an AI summary clicked a traditional search result in 8 percent of visits, compared with 15 percent of visits when no AI summary appeared. The study tracked 900 US adults across 68,879 unique Google searches using passive browser tracking — the most rigorous behavioural methodology in the literature so far. Nearly half the click pressure on the traditional results disappears when an AI summary takes the slot above them.
The second is the Ahrefs measurement of position-1 click-through rate specifically.
In a comparison of 300,000 keywords — 150,000 that triggered an AI Overview and 150,000 informational keywords that did not — Ahrefs found that the presence of an AI Overview reduces position-1 click-through rate by about 34.5 percent compared to a forecast baseline. The two studies measure different things and arrive at the same direction. Ranking one still beats ranking ten. Ranking one in 2026 carries roughly half the click weight it carried in the pre-summary era.
That is the honest version. Not "ranking is dead," and not "nothing changed."
What replaced ranking as the goal worth chasing?
Selection replaced it for the half of queries that now return an AI summary.
The engine reads the top ten or so traditional results plus a wider pool of pages from query fan-out, picks a handful of those pages as sources, and quotes one passage from each into the synthesized answer. Position one is a strong prior on being in that handful. It is not a guarantee, and a page at position four can be selected over a page at position one if the position-four page carries the cleaner extractable answer.
This is the move from an ordered list to a selected set. The same library and a different librarian.
A rank-one page that does not contain an extractable answer to the sub-question the user asked is passed over. A rank-three page that does — a short paragraph that names the answer, an honest source-attributed claim, a heading that mirrors the question — is selected and quoted. The quotation is the new shop window.
Selection is the discipline that decides whose passage ends up there, and it is what the search-everywhere-optimization layer extends across the AI chat surfaces beyond Google. A small site that builds for selection earns answer-space across half a dozen surfaces, not only the one ordered list beneath the Google search box.
Where does ranking still matter for a small site?
Ranking still matters as the floor underneath everything else.
A page that does not rank anywhere on page one is rarely retrieved into the selection pool. The traditional top ten is one of the two main inputs the AI summary engine reads, alongside the wider fan-out pool. A page that lives on page four is invisible to both. The first step toward being quoted is still being rankable enough to be seen.
Three query types still draw the click on rank one with the old weight.
The first is brand-name queries. A buyer who types your name came to find you, and the AI summary does not interrupt the visit because there is no shared answer to summarise.
The second is transactional queries — buy, hire, book, near-me. The user came to choose a destination, not to read an answer.
The third is queries the engine cannot synthesise confidently, which still default to the traditional ranked list. For all three, the on-page basics that earn position one — clean title tags, descriptive H1s, fast load, structured data — still earn the click that follows.
So the right frame is not "ranking still matters" or "ranking does not matter." It is "ranking is the floor, and selection is the goal."
How do you tell whether your rank changes still matter?
Two checks, both free, both take a few minutes per page.
The first check is whether your page is cited as a source inside the AI summary that triggers on your target query. Run the query in an incognito window from the country your readers live in. Read the summary.
Look at the source list at the bottom of the summary. Your page is either there, or not there. If it is there, the citation is doing work the rank chart cannot show you — a brand impression every time the summary fires.
The second check is whether your rank changes still move your Search Console click line.
Open Search Console for the query in question, compare a 28-day window before the rank change to a 28-day window after, and read both numbers. If your rank climbed from five to three and your clicks rose by a third, the position is still pulling its weight. If your rank climbed from five to three and your clicks stayed flat, the position is doing less work than it used to — and the work has moved somewhere else, usually into the summary slot above it. When the engine summarises your topic, a page cited inside the answer earns the brand mention the rank line cannot record.
You run both checks once a quarter on the five pages that matter most. Twenty minutes, no tools, no spend.
Other questions worth answering
Do unlinked brand mentions in AI summaries deliver value without a click-through?
Yes. An unlinked mention still teaches the engine your brand belongs to the topic. The summary fires, your name appears, the reader sees it. Across thousands of impressions a quarter, brand recall builds.
The Pew Research July 2025 study tracked 68,879 Google searches. Eight percent of AI-summary visits ended in any traditional click. The remaining visits still read something, and your name being in that something does work.
How quickly is the click-through compression accelerating across informational queries?
Fast enough to measure in single calendar years. Ahrefs ran the same 300,000-keyword comparison in April 2025 and again in December 2025. The first run showed a 34.5 percent click drop on position-one results when an AI summary appeared. The second run showed roughly 58 percent.
In nine months the click loss nearly doubled. A click forecast from six months ago is already optimistic.
How do queries without AI summaries fare on click-through trends?
They still lose click share, even without an AI summary in the slot. Ahrefs measured informational keywords without any AI summary present in their 2025 data and still found click-through rates falling year over year. Featured Snippets, People Also Ask, and Local Pack boxes also eat screen space.
The honest framing is ‘less squeezed, not unsqueezed.’ Plan as if the slow leak applies everywhere on informational search.
Are there blind spots in how Pew Research measured the AI summary effect?
Two worth flagging.
First, Pew Research tracked behavior in March 2025, when AI summary trigger rates had not yet peaked at roughly 25 percent that July. The 8 percent click figure could shift at different trigger volumes.
Second, Google disputed the methodology in May 2025, arguing the study analyzed only the top three AI-summary cited links. Independent Ahrefs re-analyses pointed in the same direction despite magnitude debates.
What should you check before chasing rank one?
Two questions, asked in order, before you spend another month on a rank-climb campaign.
Is the page already cited inside the AI summary for the queries you care about? If the answer is yes, the rank you already have is doing real work — the citation is firing, the brand is being mentioned, the page is being read by the engine even when no human clicks. The next move is widening the cited surface to adjacent queries that fan out from the same topic, not climbing one more rung on a ladder whose top rung does less than it used to.
If the answer is no, the next move is making the page extractable enough to earn the citation, and then revisiting rank from there.
If you read this and still feel unsure whether your own ranks are doing the work you think they are, that is normal. Most small site owners are reading their Search Console chart through a 2018 lens. You can contact me here and we can open your Search Console together for one or two pages, run the two checks side by side, and read what the numbers actually say.
There is no charge for the call, and there is no pitch at the end. I will tell you which of your rank changes still matter, which ones have stopped working the way they used to, and one or two pages where a selection-side fix would beat another month of rank-chasing.