Position one means less than it did, because there may be nothing at position one that anyone looks at. Here is what the data actually shows, and which numbers to watch instead.
Summary
Ranking reports measure position within organic results, but organic results are no longer what searchers encounter first. AI Overviews, paid ads, local packs and video carousels push listings down the page, and personalisation means no two people see the same results. Studies show organic click-through rates on AI Overview queries fell from 1.76% to 0.61% before partially recovering. Impressions, click-through rate and conversions describe reality more accurately.
Key Takeaways
- Rankings measure position in organic results, not visibility on the page
- Pew found clicks fell from 15% to 8% when an AI summary appeared
- One large study found AI Overview CTR fell 1.76% to 0.61% before rebounding to ~2.4%
- AI Overview prevalence estimates range from 21% to 65% depending on the source
- Personalisation means no single set of rankings exists for most queries
- Search Console’s Generative AI report has no clicks or query data
- The general Performance report includes AI Mode queries, unlabelled
- Impressions, CTR and conversions describe outcomes; rankings describe inputs
The Problem in One Sentence
You can rank first for a term and receive nothing, because the searcher never scrolled to where you are.
A modern results page can stack four paid ads, product listings, an AI Overview, a local pack and a “People also ask” block above the first organic listing. Your position one is genuine. It is also below the fold on a laptop and several swipes down on a phone.
Ranking tools report where you sit in a list. They do not report whether anyone reaches the list.
What the Data Actually Says
These figures come from published research rather than agency assertion.
| Finding | Source |
| Clicks fell from 15% to 8% of searches when an AI summary appeared | Pew Research Center |
| Organic CTR on AI Overview queries fell 1.76% → 0.61%, then rebounded to ~2.4% | Study across 2.43bn impressions, 53 brands, 5.47m queries |
| Publisher organic search traffic down 42% | Define Media Group |
| Roughly 70% of news-related queries end without leaving Google | Similarweb |
| Sessions ended without any click 26% of the time with an AI summary, versus 16% without | Pew Research Center |
The rebound in that second row deserves attention. CTR ended higher than it started. Nobody has established why — user behaviour adapting, Google adjusting the feature, or a change in which queries trigger AI Overviews are all candidates.
Which is itself the point. A metric that fell 65% and then overshot its starting position is not something a ranking report would have told you anything about.
How Often This Applies
Estimates of AI Overview prevalence vary enormously, and that variance is itself informative.
| Source | Share of searches with AI Overviews |
| Conservative industry benchmarks | 21–25% |
| BrightEdge | 48% across major industries |
| Some US trackers | 60–65% |
Anyone quoting a single figure without a source is guessing. Prevalence differs sharply by industry, query type and country which means the gap between your rankings and your visibility is not uniform either.
Check your own verticals rather than accepting an industry average.
Personalisation Makes “The Rankings” a Fiction
There is no single set of results for most queries.
What appears varies by location, device, search history, signed-in status and language. The results your chief executive sees on her phone in one city differ from what you see on a desktop in another, and both differ from what your customer sees.
A ranking tool reports one synthetic position — typically a fixed location, a fixed device, no personalisation. That number is a useful trend line. It is not a description of anyone’s actual experience.
Most ranking tools also cannot report which SERP features appeared above you, which is the information that would explain whether the position was worth anything.
The Reporting Gap Nobody Mentions
Google’s own tools split the relevant data across two reports, and neither is complete.
Its Generative AI report in Search Console shows impressions from AI Overviews and AI Mode — but no clicks, no click-through rate and no query data.
The general Performance report has clicks and queries, but blends AI-driven traffic in with everything else. John Mueller has confirmed that AI Overviews and AI Mode information appears there.
So the AI-specific view tells you how often you were seen and not whether anyone came. The complete view does not tell you which was which.
Practical response: export both for the same URLs and date ranges, and pair them with analytics sessions. Visibility and value have to be measured separately because Google reports them separately.
What to Measure Instead
Rankings are an input. These are outcomes.
- Impressions — how often you appeared, across all surfaces
- Click-through rate — whether appearing produced anything
- Sessions from organic search — verified in analytics, not inferred
- Conversions and revenue from organic — the only number that survives a board meeting
- Branded versus non-branded query split — separating demand you created from demand you captured
- Share of visibility across SERP features — featured snippets, People also ask, image and video results, local packs
The useful diagnostic is the relationship between impressions and clicks. Rising impressions with flat clicks means you are being seen and passed over — a very different problem from falling impressions, and one no ranking report distinguishes.
What Rankings Are Still Good For
They have not become useless, and overcorrecting is its own error.
Rankings remain a legitimate directional indicator. Sustained movement across a keyword set after a substantial change still tells you something real about how a page is being assessed.
They are useful for competitive tracking, for spotting sudden losses, and for diagnosing whether a traffic decline is a visibility problem or a conversion problem.
The error is treating position as the outcome rather than as one input to it. A ranking report answers “where am I in the list.” It cannot answer “did the list matter.”
A Practical Test
Take your top ten commercial queries and search each one manually. On a phone, in an incognito window, from the location your customers are actually in.
Then count what appears before your listing. Ads, AI Overview, local pack, People also ask, video carousel, shopping results.
Compare that against what your ranking tool says. The gap between the two numbers is the gap between what you are measuring and what your customers see.
It takes twenty minutes and it will change what you report on.
Frequently Asked Questions
Are SEO rankings still useful?
Yes, as a directional indicator and for competitive tracking. They show how a page is being assessed over time. They do not show whether anyone saw or clicked the listing, which requires impressions, click-through rate and conversion data.
Why don’t rankings reflect what customers see?
Because search results now include paid ads, AI Overviews, local packs, People also ask and video carousels that push organic listings down the page, and because results are personalised by location, device and search history.
How much have AI Overviews affected click-through rates?
Pew Research found clicks fell from 15% to 8% of searches when an AI summary appeared. A separate study across 2.43 billion impressions found organic CTR on those queries fell from 1.76% to 0.61% before rebounding to around 2.4%.
How often do AI Overviews appear?
Estimates vary widely 21% to 25% by conservative benchmarks, 48% according to BrightEdge across major industries, and 60% to 65% by some US trackers. Prevalence differs substantially by industry and query type.
Which metrics should I report instead?
Impressions, click-through rate, organic sessions, conversions and revenue from organic search, the branded versus non-branded query split, and visibility across SERP features rather than organic position alone.
Can I see AI Overview performance in Search Console?
Partially. The Generative AI report shows impressions but no clicks, click-through rate or query data. The general Performance report includes AI Mode data with clicks and queries, but does not label which entries came from AI features.
What does rising impressions with flat clicks mean?
That you are appearing but not being chosen. That is a relevance, title or SERP-feature problem rather than a visibility problem, and it requires a different response from falling impressions.
How do I check what my customers actually see?
Search your key commercial queries manually on a phone, in an incognito window, from your customers’ location, and count how many features appear before your listing. Compare that with what your ranking tool reports.
Conclusion
Rankings are not wrong. They are answering a narrower question than the one most businesses think they are asking.
Position one is a fact about a list. Whether that list appears above an AI Overview, below four ads and a local pack, on a phone screen, for a signed-in user in a different city those are facts about a customer, and no ranking report contains them.
The twenty-minute test settles it faster than any argument. Search your own top queries the way your customers do, and count what stands between them and you.
