AI Overviews Cut #1 Clicks by 58%.
Ahrefs' December 2025 update makes the cost of an answer-first result page measurable.
Ahrefs, 300,000 keywordsPosition-one CTR: 7.3% → 1.6%
The short answer: rank still matters. The click does not arrive the way it used to.
Ahrefs measured the effect twice using 300,000 keywords and aggregated Search Console data. In April 2025, AI Overviews correlated with a 34.5% reduction in clicks to the top-ranking page. In its February 2026 update, using December 2025 data and the same methodology, the loss was 58%.
This is not a reason to declare rankings dead. The first organic result is still a powerful source of credibility, retrieval, and possible citation. It is a reason to stop using a high rank as shorthand for traffic. An SEO dashboard can remain green while the actual visit, the thing many acquisition models are built to earn, quietly shrinks.
That distinction changes the planning conversation. The question is no longer merely, “Can we win this query?” It is, “If we win it, what does that visibility still return—and what else must this page do after the answer has already been summarized?”
April 2025
34.5%
position-one click loss
December 2025 data
58%
position-one click loss
Position one CTR on AI Overview queries
7.3% → 1.6%
Measured twice, with a control group
The update matters because it is not a single dramatic screenshot or a claim about one unlucky publisher. Ahrefs compared 150,000 informational keywords that triggered an AI Overview with 150,000 that did not. The control group also lost clicks over the period—a reminder that search has been moving toward zero-click behavior for years. But the Overview group fell much farther: average position-one CTR moved from 7.3% to 1.6%.
That gap is how the study arrived at the 58% effect. It is also why a top ranking can look stable in a rank tracker while the commercial outcome worsens. A team that relies on rank alone will read the result as success. A team that compares click-through rate, answer presence, and conversion quality can see the new bargain: visibility may remain, but the visit is no longer guaranteed.
The number should not be used as a magical multiplier. Query intent, device, brand strength, the quality of the answer, and the product page all change the realized impact. But it is a much better planning input than an old generic CTR curve. Treat the curve as a risk signal, then calculate your own exposure with the queries and outcomes that actually pay your team.
No position escapes the loss.
The Ahrefs update removed the easiest excuse: that AI Overviews only hurt the first organic result. Every position from one through ten declined. The curve is graduated, but it is not selective.
Position #1
−58.0%
Position #2
−50.8%
Position #3
−46.4%
Position #4
−38.8%
Position #5
−32.6%
Position #6
−30.5%
Position #7
−29.7%
Position #8
−28.8%
Position #9
−29.7%
Position #10
−19.4%
The result is most consequential at the top because those are the positions businesses spend years and budget earning. Roughly 9% of Overviews also appeared below position one in the update. The feature is not simply taking the first result's real estate; it is changing the click economics of the whole results page. A number-one label still means something, but it can no longer be translated mechanically into a historical share of sessions.
The real loss is the gap between being present and being chosen.
A result page with an AI Overview can still put a brand in front of a buyer. It may even cite the source that supplied part of the answer. But a citation is not a visit, a visit is not a qualified conversation, and an impression is not revenue. That sounds obvious until the reporting system rolls all three ideas into a single “visibility” success story.
The pages most exposed are the ones whose full value fits inside a compact answer: basic definitions, simple comparisons, easy checklists, and procedural questions with no meaningful next step. Those pages can still serve a brand, but their old job—reliably sending a large volume of search referrals—has become less dependable.
The response is not to withhold useful information. It is to make the page worth choosing after the overview does its job. Original research, a specific point of view, an interactive tool, a live inventory, implementation detail, first-party examples, or a decision framework gives a reader a reason to continue. That is content strategy and product strategy meeting in the same URL.
The corroboration stack is the point.
One vendor study is an input. Several independent measurements in the same range are a planning signal. Seer Interactive reported organic CTR declines in the 49.4% to 65.2% range, while Authoritas research reported by Press Gazette found a 47.5% publisher loss in its analysis. These are not identical experiments, and they should not be treated as one pooled number. They do point in the same operational direction.
The methods differ: Search Console aggregates tell one part of the story; observational behavior research tells another. The direction does not. When the result page answers the question, a material share of historical clicks does not happen. That does not make every AI Overview query useless. It does mean SEO reporting cannot treat an impression, a rank, and a visit as interchangeable evidence.
If the program never separates those layers, it will keep celebrating traffic forecasts that no longer materialize. If it does separate them, the loss becomes tractable: which queries lost clicks, which pages earn mentions, which routes still create demand, and where a different distribution surface is worth funding.
For the related strategic question, see how citation authority changes the work. For the analytics problem, see the visibility-trap analysis.
A practical reset for the SEO dashboard
01
Map exposure before changing the plan
Export the queries where you rank, tag the ones that trigger an AI Overview, and segment by intent and commercial value. Model the exposure before assigning blame to a page or writer.
02
Replace the old CTR curve
Build a baseline that distinguishes classic results from AI Overview results. Report actual clicks and conversions, not a traffic estimate inherited from an earlier era of SERPs.
03
Measure answer presence separately
Track whether the overview cites or accurately represents you. That tells you about visibility and retrieval, but keep it separate from referral traffic and business value.
04
Give the click a reason to exist
Prioritize pages that contain the evidence, tools, depth, specificity, and next action an overview cannot fully deliver. The goal is a better visit, not a longer article for its own sake.
This is not “stop doing SEO.” It is a move from ranking as the final metric to ranking as one signal in an evidence chain. Add citation and answer metrics beside clicks; add AI-referred conversions beside conventional organic conversions; and use brand demand to check whether being present in a result page is creating preference even when it does not create an immediate session. The team needs all of these facts to decide where the next unit of content effort belongs.
What a rank tracker cannot tell you
A rank tracker is still useful. It tells a team whether Google can retrieve and place a page for a query. It can show a competitor moving, a technical problem, a content refresh working, or an entire category becoming more competitive. What it cannot do is reveal the changing deal between the result and the person searching. A number-one result and a number-one result that earns a historical share of clicks are now different observations.
That matters because standard dashboards often promote rankings into the executive summary and leave CTR deeper in a search console report. The sequence should run the other way on exposed queries. Start with the business outcome, then inspect clicks, CTR, answer presence, rank, and the page's role in the journey. A stable rank with a falling click-through rate is not proof that the program is healthy. It is a prompt to ask whether the overview has taken over the job that the landing page used to perform.
The distinction also keeps a team from making the wrong optimization. If the page has lost rank, an editorial or technical fix may be appropriate. If the page has held rank but the Overview has absorbed the basic answer, rewriting the opening paragraph for the tenth time may not restore the old referral curve. The more useful move might be to add an original calculation, a product-specific workflow, a live comparison, or a next-step decision that creates a genuine reason to leave the answer layer.
In other words, rank describes eligibility. CTR describes the handoff. Conversion describes whether the handoff was worth winning. Keeping those three measures distinct is a simpler and more honest way to manage organic search in an answer-first results page.
Find the exposure before you change the content calendar
Start with the query set that already matters to the business: high-intent pages, category terms, comparisons, product questions, and topics tied to a known conversion path. Export current rank, clicks, impressions, CTR, landing page, and conversion data for that set. Then identify the queries where an AI Overview is present. The objective is not a perfect census on day one. It is to make the risk visible enough that the team stops averaging unlike kinds of search together.
Next, split the set by intent. A broad definition query might deserve a different goal from a comparison query, a pricing question, or a troubleshooting search. For each group, establish the observed CTR rather than borrowing a generic industry curve. Compare a recent period with a useful historical period, but annotate changes in seasonality, device mix, paid coverage, brand demand, and SERP features. The point is to understand the local pattern, not force every fluctuation into the 58% headline.
Then tie the query group to a commercial measure. A query with a low click rate may still matter if it produces unusually qualified visitors, if the brand is cited in the answer, or if it is a repeated source of assisted conversions. Conversely, a large top-of-funnel page can look impressive in an impression report while returning little business value after the Overview has answered the reader's first question. Exposure becomes a decision tool only when it is connected to the outcome the program exists to improve.
This work should produce a short operating list: preserve, improve, redesign for post-answer value, or de-prioritize. It is a better input to a content calendar than a list of keywords with a high old search-volume estimate, because it reflects the actual environment in which the page will have to earn attention.
The goal is a better evidence chain, not a prettier report
An answer-first search environment makes weak measurement more visible. If an executive only sees impressions rising, the program can be celebrated while the referral pipeline thins. If a team only sees traffic falling, it can conclude that the work has failed even when citation presence, qualified visits, or downstream demand have improved. Neither conclusion is safe without the chain between the result page and the commercial outcome.
Build that chain explicitly. Record whether the query triggered an Overview at the time it was observed. Record rank and the landing page. Record clicks and CTR. Record citation or brand mention when it can be measured responsibly. Finally, record the outcome that matters: an engaged session, a signup, a qualified lead, a purchase, a renewal, or a durable increase in branded search. Each signal has a different meaning. Putting them on one scorecard is helpful; collapsing them into one score is not.
The same discipline is useful when content succeeds. A page that is cited but does not receive many clicks may be doing category education. A page that receives fewer clicks but converts dramatically better may be serving a narrower, more ready buyer. A page that loses both clicks and downstream value may need to be rethought. The measurement model should help a team tell those stories apart so budget follows evidence rather than anxiety.
That is why the Ahrefs result is valuable beyond its headline. It is a warning against an old shortcut: assuming that top placement means the old distribution outcome. Once that assumption is removed, the next step is practical. Measure what the page is actually earning, decide what role it should play, and build the next piece of content for that role instead of for a vanished CTR curve.
Review that evidence on a regular cadence, not only after a traffic shock. A monthly query-class review can show whether the affected set is growing, whether the loss is concentrated in an intent group, and whether pages redesigned for post-answer value are creating stronger outcomes. That turns AI Overview exposure from a one-time panic metric into a normal operating condition: observed, explained, and managed with the same discipline as any other change in distribution.
FAQs
How much do AI Overviews reduce clicks?+
Ahrefs found a 58% lower position-one click-through rate on AI Overview queries in its February 2026 update, up from 34.5% in its April 2025 study. The figure is a category-level measurement, not a promise that every query loses exactly the same share.
Does ranking number one still matter?+
Yes, but its value has changed. The first result remains eligible for discovery and citations. On the December 2025 data, though, the first organic position received 1.6% CTR on AI Overview queries, down from 7.3% at the start of the measured period.
What should SEO teams measure now?+
Track AI Overview presence, click-through rate by query class, citations or mentions, AI-referred conversions, and brand demand alongside ranking. The point is to distinguish a visibility gain from a visit, and a visit from a commercial outcome.
A top ranking is still valuable. It is no longer a traffic guarantee.
Measure the path that remains after the answer is displayed, then invest in the work a buyer still has a reason to choose.