Seven credible studies answer the same question about AI Overviews with figures ranging from a 15.49% click loss to 78.4%. The two largest datasets published in 2026 disagree about the direction of the trend. Two respected sources measure the overlap between rankings and AI citations 59 percentage points apart.
This report separates what the data measures from what the citation chain has turned it into.
Every figure verified against the primary source. No aggregators, no secondary summaries, no AI fact-checkers. Derived percentages are shown with their inputs.
The headline number does not exist
Search for how much AI Overviews cut clicks and the answer comes back as a single confident figure. Fifty-eight percent. Sometimes 34.5%. Sometimes 61%, 79%, or 15.49%.
Every one of those numbers is real. Every one is published by a credible source. Every one is measuring something different.
The 2026 US evidence base on AI’s impact on search rankings does not converge on a number. It fractures across at least six incompatible measurement decisions, and in the most consequential case, the two largest datasets published in 2026 disagree about the direction of the trend, not just its size.
That matters for budget decisions. A B2B marketing leader who reallocates spend based on the claim that AI Overviews cost 58% of clicks is acting on a modelled counterfactual derived from desktop-only, informational-intent keywords in a 300,000-keyword sample. A leader who reads the same period through a different dataset sees organic click-through rate on AI Overview queries rising 80% in two months.
Both readings are defensible. Neither is the number.
The reported range: one question, seven answers
Every figure below purports to answer the same question: what happens to organic click-through rate when Google places an AI Overview above the results?
| Source | Reported impact | What was actually measured |
|---|---|---|
| Amsive | -15.49% | Average CTR change across 700,000 keywords, all positions, all intents |
| Ahrefs, April 2025 | -34.5% | Modelled position-1 CTR loss, desktop, informational keywords only |
| Seer Interactive, Sept 2025 | -41% | CTR decline on queries with NO AI Overview present |
| Pew Research Center | -47% relative | Real-user click rate per search page visit, all query types |
| Ahrefs, February 2026 | -58.0% | Modelled position-1 CTR loss, desktop, informational keywords only |
| Seer Interactive, Sept 2025 | -61% | Observed 15-month organic CTR trajectory on AIO-present queries |
| Ahrefs, February 2026 | -78.4% | Raw observed position-1 CTR change, unadjusted for baseline erosion |
The spread runs from 15.49% to 78.4%. The third row is the one that should stop a budget conversation cold: a 41% CTR decline measured on queries where Google served no AI Overview at all.
Figure 1. Reported organic CTR impact of Google AI Overviews, by study and measurement method.
Fracture 1: The same 300,000 keywords produce three answers
Ahrefs’ February 2026 update is the most-cited AI Overview statistic in circulation. The methodology is published in full, and the published numbers contain three distinct answers to the same question.
Ahrefs compared 150,000 keywords that trigger an AI Overview against 150,000 informational keywords that do not, using aggregated Google Search Console desktop data from December 2023 (pre-rollout) and December 2025.
In December 2023, average position-one CTR for informational keywords was 0.076. By December 2025 it had fallen to 0.039. Over the same period, position-one CTR for AI Overview keywords fell from 0.073 to 0.016.
Three legitimate calculations follow from those four numbers:
| Calculation | Result | What it means |
|---|---|---|
| Raw change on AIO keywords (0.073 to 0.016) | -78.4% | What actually happened to those pages |
| Change on non-AIO informational keywords (0.076 to 0.039) | -48.6% | What happened without any AI Overview involved |
| Modelled AIO-attributable loss (0.016 vs forecast 0.037) | -58.0% | The portion attributed to AI Overviews specifically |
Ahrefs derives the forecast by applying the non-AIO decline rate to the AIO cohort’s 2023 baseline, producing a predicted CTR of 0.037 against an actual of 0.016.
The 58% figure is not an observation. It is a counterfactual estimate of what would have happened in a world where AI Overviews were never launched. It is a defensible method and Ahrefs documents it transparently. It is also routinely quoted as though a measurement instrument recorded it.
The number that gets almost no circulation is the middle one. Keywords that never triggered an AI Overview lost roughly half their position-one click-through rate over the same two years. Whatever is compressing organic CTR in the US, AI Overviews are not the whole of it.
Figure 2. Three defensible readings of one Ahrefs dataset, December 2023 against December 2025.
The position-by-position breakdown
| Organic position | AIO-attributed CTR impact |
|---|---|
| 1 | -58.0% |
| 2 | -50.8% |
| 3 | -46.4% |
| 4 | -38.8% |
| 5 | -32.6% |
| 6 | -30.5% |
| 7 | -29.7% |
| 8 | -28.8% |
| 9 | -29.7% |
| 10 | -19.4% |
The damage concentrates at the top. Position one absorbs three times the impact of position ten. For B2B sites that fought for the top spot on high-value informational terms, the asset that was most expensive to acquire is the one losing the most value.
Fracture 2: AI Overview queries were already low-click
Seer Interactive’s September 2025 study introduced a finding that undermines the causal reading of every CTR study in this space, including its own.
Seer analysed 3,119 informational and educational search terms across 42 client organisations, covering 25.1 million organic impressions and 1.1 million paid impressions from June 2024 through September 2025. The cohort was classified by whether a query showed an AI Overview in September 2025, then that label was applied retrospectively across the full 15 months.
In June 2024, before AI Overviews had meaningfully rolled out, queries that would later trigger an AI Overview already showed an organic CTR of 1.76%, against 2.74% for queries that never triggered one.
Queries destined to receive an AI Overview were converting attention 36% worse than their peers before the feature arrived.
Seer’s own reading is that AI Overviews are not randomly appearing on high-performing queries and destroying their CTR. They are surfacing on query types that were already generating fewer clicks.
This is selection, not causation. Google deploys AI Overviews on queries where users historically resolved intent without clicking, which is the same pattern featured snippets established years earlier. Part of the measured AI Overview effect is the pre-existing character of the queries Google chose to put them on.
Seer states the limit directly in its methodology: the study is a retrospective cohort analysis, not a month-by-month correlation study, and causation cannot be claimed.
Fracture 3: A 52% CTR collapse where clicks did not fall
The most instructive single data point in the 2026 evidence base comes from Seer’s April 2026 update, which flagged October 2025 as the largest month-over-month organic CTR drop in its entire dataset. Informational queries where the brand was cited in the AI Overview fell 52.1% in a single month.
| Month | Brand-cited impressions | Brand-cited clicks | CTR |
|---|---|---|---|
| September 2025 | 15,797,234 | 398,798 | 2.52% |
| October 2025 | 33,100,739 | 400,271 | 1.21% |
| November 2025 | 39,500,144 | 301,783 | 0.76% |
Clicks moved from 398,798 to 400,271 in October. Impressions more than doubled. CTR compressed because the denominator grew far faster than the numerator.
A 52% CTR collapse occurred in a month where click volume was flat and rising. Any dashboard reporting CTR alone would have shown a catastrophe. Any dashboard reporting clicks would have shown stability.
Seer’s own interpretation is that the impression surge could reflect brands earning citations on more queries, potentially from new content or stronger SEO performance, or it could reflect changes in how the AI Overview algorithm selects sources. The two cannot be separated without account-level investigation.
The operational consequence is straightforward. CTR is a ratio, and in AI search both of its terms are moving independently. Reporting the ratio without reporting its components produces false alarms and false victories.
Fracture 4: The largest 2026 datasets disagree on direction
Ahrefs published its update on 4 February 2026, using December 2025 data, concluding that the impact of AI Overviews had worsened from 34.5% to 58%.
Seer published its update on 24 April 2026, using data through February 2026, concluding the opposite.
Seer tracked 53 brands, 5,471,127 distinct queries, 2.43 billion organic impressions and 296.9 million paid impressions from January 2025 through February 2026. Organic CTR on AI Overview-present queries climbed from a December 2025 floor of 1.31% to 2.36% in February 2026.
Seer headlines that recovery as an 85% increase, derived from rounded values of 1.3% and 2.4%. The precise figures in its own published data table produce 80%. The difference is immaterial to the conclusion and material to the principle: rounding applied before a percentage calculation moves the published number by five points.
Every informational and transactional organic segment beat Seer’s forecast model in both January and February 2026. The model had projected continued decline.
Figure 3. Seer Interactive monthly organic CTR, January 2025 to February 2026, against the month Ahrefs measured.
Ahrefs measured December 2025, which is the floor of Seer’s series. The two studies do not necessarily contradict each other on the facts. They contradict each other on the story, because one stopped measuring at the bottom of a trough and the other kept going.
Anyone building a 2026 budget on the February 2026 publication cycle inherited a trajectory that had already reversed by the time the article was published.
Seer’s own caution applies: two months is not a trend, and the recovery may prove seasonal. The point is not that recovery is confirmed. The point is that the most-cited decline figure in the market was measured at a single point that later data revised.
Fracture 5: Does ranking still drive AI citation?
For B2B teams, this is the fracture with the largest strategic consequence, because it determines whether existing ranking investment carries over into AI visibility. Two credible studies answer it with a 59-percentage-point gap.
Ahrefs studied 1.9 million citations from 1 million AI Overviews, analysing the top three most visible citations in each response. It found that 76.10% of AI Overview-cited pages rank in the top 10, 9.50% rank between positions 11 and 100, and 14.40% do not rank in the top 100 at all.
BrightEdge, tracking February 2025 through February 2026, reported that only about 17% of sources cited in AI Overviews also rank in the organic top 10, and that top-10 overlap has remained largely flat while top-100 overlap has increased modestly.
Same feature. Same year. Opposite strategic implication.
Figure 4. Ranking and citation overlap measured two ways, and AI Overview exposure by query format.
The reconciliation is almost certainly the unit of analysis. Ahrefs examined the three most prominent citations per Overview, which are the ones most likely to be drawn from strong organic performers. BrightEdge appears to measure across the full citation set, including the long tail of sources referenced deeper in the answer. BrightEdge attributes the divergence to Google’s FastSearch retrieval method and query fan-out, where one conversational query triggers multiple sub-queries, each returning its own organic results.
Neither study is wrong. They answer different questions. The market quotes them as though they answer the same one.
Ahrefs’ own supporting detail complicates the optimistic reading further. Median organic ranking for the top-cited URL in an AI Overview is position 2. Second-place citations typically rank 4th, third-place citations 5th. Ahrefs’ Si Quan Ong summarised the correlation between ranking first and being cited as a coin flip at best.
The defensible conclusion for 2026: high rankings improve citation odds substantially, and do not secure them. Ranking and citation are correlated, overlapping, and separately measurable. Treating them as one metric produces reporting that cannot explain its own results.
Fracture 6: The conversion premium, from 23x to zero
The counter-narrative to click loss is that AI-referred visitors convert far better, so smaller traffic still produces more pipeline. The evidence for that claim spans the full range from dramatic to null.
Ahrefs, June 2025: sample size of one
AI search traffic accounted for 0.5% of Ahrefs’ traffic over 30 days while driving 12.1% of signups, a 23x conversion differential, with the majority arriving from ChatGPT. This is one SaaS company’s own analytics over one month.
Semrush, June 2025: one vertical
Semrush found the average AI search visitor 4.4 times as valuable as the average traditional organic visitor by conversion rate, and projects AI search visitors to surpass traditional search visitors in early 2028 with economic parity by the end of 2027. The study is based on over 500 high-value digital marketing and SEO topics, and Semrush notes its traffic and value projections are extrapolations of historical data and adoption rates.
The 4.4x figure is repeatedly republished as an across-industries average. It was derived from the digital marketing and SEO vertical, which is the category with the highest AI tool adoption in the economy. That is the least representative sample available for generalising to B2B SaaS, biotech, or pharma.
Amsive, September 2025: statistically tested
Amsive ran the test the other studies did not. Across 54 websites with manually audited GA4 conversion tracking over six months, organic traffic converted at 4.60% and LLM referrals at 4.87%. The mean site-level difference was +0.27 percentage points, the median +0.09, with a standard deviation of 7.53%. A paired t-test returned p = 0.794.
Not significant. Restricting to 33 higher-volume sites widened the gap to +1.24 percentage points and still failed significance at p = 0.376.
For B2B specifically, which is the segment that matters most here:
| Segment (thresholded sample) | Organic CVR | LLM CVR | Difference | p-value |
|---|---|---|---|---|
| B2B | 1.68% | 2.03% | +0.35 pp | 0.705 |
| B2C | 8.50% | 10.31% | +1.81 pp | 0.423 |
| B2B vs B2C (Welch’s test) | n/a | n/a | n/a | 0.546 |
56% of sites saw LLM traffic convert above their site average, 41% below, and 4% the same. That near-even split is the finding. The conversion premium is real on some sites and negative on others, and the variance swamps the average.
The scale context is the part most often omitted. Organic search accounted for 31.9% of sessions and 33.8% of conversions across Amsive’s sample. LLM traffic accounted for 0.24% of sessions and 0.42% of conversions. Nearly 90% of sites had LLM traffic below 0.6% of total traffic.
A 23x multiplier on 0.24% of sessions does not replace a third of the funnel.
What holds up across every dataset
Six things survive the disagreements. These are the load-bearing findings.
- Real-user click behaviour changes when an AI Overview appears. Pew Research Center analysed 68,879 unique Google searches from 900 US adults who shared browsing data in March 2025. Users clicked a traditional search result on 8% of visits to pages with an AI summary, against 15% of visits to pages without one. This is observed behaviour, not modelled, and it is the cleanest causal evidence available.
- Citation links inside AI Overviews are almost never clicked. Pew recorded clicks on a source link inside the summary on 1% of visits to pages containing one. Any strategy justified by traffic arriving through the citation footer is built on a 1% event rate.
- Sessions end more often. 26% of visits to search pages with an AI summary ended the browsing session, against 16% for pages with only traditional results.
- Being cited beats not being cited, and both lose to no AI Overview. Seer’s per-million-impression translation for informational queries: no AI Overview yields roughly 33,500 organic clicks, brand cited yields roughly 20,743, and AI Overview present without citation yields roughly 9,445. Citation delivers 120% more organic clicks per impression than non-citation, and still underperforms a no-AIO SERP by 38%.
- Exposure is a function of intent, not of industry alone. Across 53 accounts, informational queries showed an AI Overview 36% of the time, commercial queries 8%, and transactional queries 5%, a sevenfold gap between informational and transactional.
- Traditional results still dominate more than half of queries. BrightEdge found AI Overview coverage grew 58% year over year to approach half of all tracked queries, with classic results still appearing on 52% of queries.
What this means for B2B specifically
Aggregate statistics understate the exposure for B2B technology companies. The vertical data is worse than the average, and the query-format data is worse still.
BrightEdge tracked B2B technology AI Overview coverage rising from 36% to 82% of queries between February 2025 and February 2026. Healthcare reached 88%, education 83%, and restaurants 78%.
A B2B technology company is operating in a vertical where four out of five tracked queries now return an AI-generated answer above the organic results, against a cross-industry average near half.
The query formats that B2B content programmes are built on are the most exposed formats in the dataset:
| Query type | AI Overview trigger rate |
|---|---|
| Comparison (X vs Y) | 95.4% |
| Review queries | 86.3% |
| Question format (what / why / how / is / are) | 85.9% |
| Price / cost / buy | 83.4% |
| Best of | 81.3% |
| Near me | 76.9% |
| Single word | 27.3% |
Comparison pages and question-format content are the backbone of most B2B SaaS content programmes. They are also, per this data, the two formats least likely to deliver a click and most likely to require citation to deliver anything at all.
The screen-space finding compounds it. BrightEdge measured the average AI Overview at more than 1,200 pixels in height against a standard desktop viewport of roughly 900 pixels, placing the first organic result entirely below the fold.
Position one on an AI Overview SERP is not above the fold. It is not visible without scrolling. The ranking is intact and the visibility is gone.
Two counterweights apply. Transactional queries showed AI Overviews only 5% of the time, and organic CTR on queries with no AI Overview improved across 2025, rising from 2.93% in January to 3.97% in December for informational queries. Seer’s hypothesis for that improvement is a selection effect: as AI Overviews absorb answer-seeking, low-engagement queries, the remaining no-AIO pool skews toward higher-intent users who are going to click something.
That is the strategic opening. Queries Google has not chosen to summarise are increasingly populated by people who intend to act.
Why the numbers diverge: seven measurement decisions
Every study in this report made a defensible choice at each of these seven points. The choices are not disclosed in the headline, and they compound.
- Unit of measurement. Position-one CTR (Ahrefs) is not average CTR across all positions (Amsive) is not click rate per search page visit (Pew) is not clicks per million impressions (Seer). Four different denominators.
- Observed versus modelled. Ahrefs reports a counterfactual estimate. Pew reports observed behaviour. Both are labelled as percentage declines.
- Query intent scope. Ahrefs restricted its analysis to informational keywords because 99.2% of AI Overview-triggering keywords are informational in intent. Seer’s 2026 update expanded across informational, transactional and commercial. Amsive’s CTR study covered all intents. The narrower the informational focus, the worse the number.
- Device. Ahrefs used desktop CTR only. Pew tracked whatever device panellists used. Mobile and desktop click behaviour differ materially.
- Cohort labelling. Seer applies a static AI Overview label from the most recent SERP snapshot retrospectively across all months, which the methodology acknowledges as a known limitation. Queries are classified by their current state, not their state at the time of measurement.
- Time window. Ahrefs compared December 2023 against December 2025, a two-year gap, because December 2024 already fell after the rollout. Pew captured a single month. Seer ran 14 months. Different windows capture different phases of the same curve.
- Sample composition. Seer discloses that its informational AI Overview non-cited aggregate of 0.94% CTR is materially influenced by one account representing 47% of that segment’s impressions at 0.20% CTR, and that excluding it raises the aggregate to approximately 1.61%. A single account moved a published benchmark by 71%.
That last disclosure is the standard the rest of the market should be held to.
How to measure this in a specific account
Aggregate benchmarks in this space are directional at best and misleading at worst. The measurement work that produces defensible answers is account-specific.
- Segment queries by AI Overview status before reading any CTR number. Three buckets: no AI Overview, AI Overview present with the domain cited, AI Overview present without citation. Blended CTR across these three buckets is uninterpretable.
- Report clicks and impressions separately, always. The October 2025 case establishes why. A CTR drop with flat clicks and doubled impressions is expanded citation coverage, not lost traffic. A CTR drop with falling clicks is lost traffic. The ratio alone cannot distinguish them.
- Classify by intent, then by query format. Informational exposure runs seven times transactional exposure. Comparison and question-format pages carry near-total AI Overview exposure. An audit that starts with those two formats finds the largest concentration of risk fastest.
- Benchmark against a no-AIO control within the same account. Non-AIO informational keywords in Ahrefs’ sample lost 48.6% of position-one CTR over two years. Without an internal control group, that baseline erosion gets attributed to AI Overviews.
- Shift the reported KPI from rank to citation share. Rankings and citations are correlated and not interchangeable. Both need tracking, separately, with the divergence between them treated as a finding rather than a reporting error.
- Instrument self-reported attribution. Amsive’s study used last-touch attribution and flags it as a limitation, noting that form fills and demo requests do not necessarily become paying clients. A how-did-you-hear-about-us field on lead forms captures the AI-influenced pipeline that referrer data structurally cannot see.
SERPsculpt builds this measurement layer into B2B SaaS SEO and content programmes before making budget recommendations, because the alternative is reallocating spend against a benchmark derived from someone else’s sample. Related analysis is available in the SERPsculpt research library, including B2B customer retention statistics and sales and marketing strategy for B2B SaaS.
The 2026 position
AI Overviews reduce clicks to top-ranking pages. That is established across every dataset examined, using every methodology, in every direction of the debate.
The magnitude is not established. The trend direction as of Q1 2026 is contested between the two largest published datasets. The relationship between ranking and AI citation is measured at 76% overlap by one credible source and 17% by another. The conversion premium fails statistical significance in the only study that tested it, including for B2B.
None of that argues for inaction. It argues against acting on a single number.
The findings that survive scrutiny point in one direction: exposure is determined by query intent and format rather than by industry averages, citation status now separates outcomes more than AI Overview presence does, and CTR has stopped being a reliable standalone metric because both of its terms move independently.
The teams that will be right about 2026 are the ones measuring their own accounts against their own controls. Everyone else is quoting a counterfactual.
Methodology
This report is built exclusively on primary sources retrieved directly from the publishing organisation. No figure was sourced from an aggregator, a secondary summary, or an AI-generated fact-check. Where a figure appears in secondary circulation with a different framing than its source provides, the source framing is used and the discrepancy is noted.
Every percentage derived rather than quoted is shown with its inputs so the calculation can be reproduced.
Primary sources retrieved
- Pew Research Center, Do people click on links in Google AI summaries? (22 July 2025)
- Ahrefs, AI Overviews Reduce Clicks by 34.5% (17 April 2025)
- Ahrefs, Update: AI Overviews Reduce Clicks by 58% (4 February 2026)
- Ahrefs, 76% of AI Overview Citations Pull From the Top 10 (21 July 2025)
- Ahrefs, Does AI Search Traffic Convert Better Than Traditional Search? (June 2025)
- Seer Interactive, AIO Impact on Google CTR: September 2025 Update (4 November 2025)
- Seer Interactive, AIO Impact on Google CTR: 2026 Update (24 April 2026)
- Amsive, Does LLM Traffic Convert Better Than Organic? (3 September 2025)
- Semrush, We Studied the Impact of AI Search on SEO Traffic (21 July 2025)
- Search Engine Journal on BrightEdge’s twelve-month AI Overview analysis (1 March 2026)
- Search Engine Land on the Ahrefs and Amsive CTR studies
Geographic scope is United States unless a source states otherwise. Ahrefs’ CTR studies use US-weighted Google Search Console data on desktop. Pew’s panel is US adults. Seer’s cohort is client accounts, predominantly North American. BrightEdge’s tracking is industry keyword sets.