AOP's 2026 AI publisher study: Google referrals fell 7.1%, news interviews grew 63%

Summary. The Association of Online Publishers published its Artificial Intelligence Publisher Impact Study on 20 July 2026, built on 10.8 billion pageviews across 8 major UK publishing groups and 147 domains. Organic Google Search referrals fell 7.1% quarter on quarter, and the AOP projects that Google-delivered pageviews will halve by Q3 2027 if the rate holds. The averages hide the useful part. Technology content fell 28% while sports rose 37%. Guide and how-to content dropped by more than 20% while opinion and commentary gained 4%. News interviews grew 63%. Two details are missing from most coverage of this study: the traffic it measures is from Q1 to Q2 2025, not 2026, and a separate Seer Interactive dataset covering 5.47 million queries found the decline reversed in January and February 2026, with organic click-through rate on AI Overview queries climbing from 1.31% to 2.36%. Google's Search revenue grew 12% to $54.2 billion in the same quarter the referrals fell.

Both studies are real. They disagree because they are looking at different years.

What the AOP actually measured

The AOP commissioned DJB Strategies to run the traffic analysis and Ipsos to run a consumer survey on awareness and usage of AI tools. Richard Reeves, managing director at the AOP, set out the traffic findings in Press Gazette on 20 July 2026.

The traffic component covers six months of data across eight major UK publishing groups, totalling 10.8 billion pageviews. The sample accounts for roughly 80% of those groups' total traffic once long-tail URLs were removed. Traffic was segmented by topic and content type across 147 domains, which is what makes the study worth reading closely: almost every other public dataset on AI search reports at the query level, not the editorial-format level.

The headline number is a 7.1% fall in organic Google Search referrals between Q1 and Q2 2025 across all participants. Reeves is explicit about the extrapolation: "If this trajectory is sustained, pageviews delivered by Google will be cut in half by Q3 2027, and this is without accounting for compounding effects that habitual AI use may have on return traffic."

That is a conditional statement, and the condition is doing a lot of work.

The date problem

The study was published on 20 July 2026. The traffic it measures is the Q1-to-Q2 2025 window. Reeves names the quarter directly, noting that Q2 2025 was "the same quarter in which Google said its Search revenue grew by 12% and AI Overviews significantly expanded."

Alphabet's Q2 2025 results confirm the revenue side. The "Google Search & other" line reported $54,190 million for the quarter ended 30 June 2025, against $48,509 million a year earlier, a rise of about 12%. Consolidated revenues were up 14% to $96.4 billion. So the contrast Reeves draws is accurate. Google's search business grew in the quarter its referral traffic to these publishers fell.

But a projection to Q3 2027 built on a single 2025 quarter-over-quarter step is a straight line drawn through two points, published a year after those points were recorded. Anyone planning 2026 budgets from the headline is planning against data that is roughly twelve months old.

The format-level results are the actual finding

Strip out the extrapolation and the study becomes considerably more useful, because it answers a question most publishers cannot answer from their own analytics: which kinds of content lost, and which did not.

By segment, news publishers held up far better than the rest. News averaged a quarterly decline of around 5%, against 18% for the consumer and B2B segment. Reeves attributes this to "the tighter guardrails placed around news in AI Overviews compared to more easily aggregated evergreen content."

Topic	Quarterly change in Google referrals	Direction
Sports	+37%	Grew
Travel	+22%	Grew
Health and wellness	+7%	Grew
News	+2%	Grew
Politics	-7%	Fell
Finance and business	-23%	Fell
Technology	-28%	Fell

Technology was the worst-affected topic in the study. For a technology publisher that is an uncomfortable place to sit, and it is worth saying plainly rather than burying: this site publishes in the single category that lost the most.

The content-type cut is sharper still, because it separates what a piece is about from how it is written.

Content format	Quarterly change in Google referrals	Interpretation
Opinion and commentary	+4%	An explicit point of view resists substitution
Interviews	+1%	Primary-source material an AI cannot synthesise
Product reviews	-1%	Roughly flat
Reportage	-4%	Largest share of sessions, modest decline
Guide and how-to	Worse than -20%	Directly substitutable by an AI answer

Reeves's reading of the how-to collapse is that users "are likely to favour the speed and convenience of an AI answer over editorial expertise." The inverse holds for opinion: "content with an explicit point of view is naturally most resistant to AI substitution."

Overlaying topic and format produces the widest spreads in the dataset. News interviews grew 63% quarter on quarter. Health and wellness analysis grew 46%. In the other direction, finance and business how-tos fell 52% and food and drink how-tos fell 54%.

A 63% gain and a 54% loss inside the same 7.1% average is the reason format-level reporting matters. The average describes almost none of the individual cases.

One caveat the study makes itself: reportage accounted for more traffic than all other formats combined, and it fell only 4%. So the formats that moved most are not the formats that carry most of the volume. Reeves suggests publishers consider "how reportage can be injected with more of a human touch to make it more resilient to AI substitution" rather than abandoning it.

The trend line the AOP study could not see

Seer Interactive published the third version of its AI Overview click-through-rate study in April 2026, covering January 2025 through February 2026. The cohort is 53 brand accounts, 5,471,127 distinct queries, 2.43 billion organic impressions and 296.9 million paid impressions, filtered to queries with complete data in every month of the window.

Seer's September 2025 model predicted continued decline into 2026. That is the same shape as the AOP extrapolation, built independently, on a larger query sample. It did not hold.

Month	Organic CTR, AI Overview shown	Organic CTR, no AI Overview
January 2025	3.19%	2.75%
June 2025	2.06%	3.07%
September 2025	2.14%	2.99%
December 2025	1.31%	3.16%
January 2026	2.33%	3.88%
February 2026	2.36%	3.82%

Organic click-through rate on AI Overview queries bottomed at 1.31% in December 2025 and recovered to 2.36% by February 2026, an 85% rise off the floor in two months. Every informational and transactional organic segment beat Seer's projection in both January and February 2026.

Seer is careful about what this means, and the caution deserves repeating rather than paraphrasing away. The team writes: "We aren't suggesting you forecast growth or recovery based on only Jan-Feb 2026 data." They describe the pattern as a levelling out or a possible "new normal", not a recovery, and note they do not expect a return to pre-AI Overview rates.

Held together, the two studies say something more precise than either says alone. Referral traffic to publishers fell materially through 2025, which the AOP measured. The rate of decline stopped compounding in early 2026, which Seer measured. A projection built only on the 2025 slope will overstate the 2027 position.

Citation status moves the number more than AI Overview presence

Seer's most actionable finding is not about the trend at all. Across informational queries for the full year, average organic click-through rate ran 3.35% when no AI Overview appeared, 2.07% when one appeared and the brand was cited inside it, and 0.94% when one appeared and the brand was not cited.

Translated into clicks per million informational impressions:

AI Overview status	Organic clicks per 1M impressions	Relative to no AI Overview
No AI Overview present	~33,500	Baseline
AI Overview present, brand cited	~20,743	-38%
AI Overview present, brand not cited	~9,445	-72%

Being cited delivers about 120% more organic clicks per impression than not being cited on the same page of results. It does not restore the pre-AI Overview baseline. Both halves of that sentence matter, and most summaries of this study carry only one of them.

What our own numbers show

Aggregate benchmarks are a poor substitute for first-party data, so here is ours, taken from Google Search Console for ecorpit.com over the period we scaled publishing volume.

Month	Impressions	Clicks	CTR	Average position
December 2025	725	18	2.48%	8.2
February 2026	1,130	42	3.72%	6.4
April 2026	3,376	52	1.54%	6.9
May 2026	15,573	141	0.91%	7.1
June 2026	897,183	2,401	0.27%	7.5
1-20 July 2026	304,659	927	0.30%	6.8

Read as a CTR series, that looks like a catastrophe: 2.48% down to 0.27%, a fall of about 89% in six months. Read as a click series, clicks went from 18 to 2,401 a month over the same period while average position barely moved, sitting between 6.4 and 8.2 throughout.

This is the same arithmetic Seer flagged in its own cohort. When brand-cited impressions rose from 15.8 million in September 2025 to 33.1 million in October, clicks moved only from 398,798 to 400,271, and CTR halved from 2.52% to 1.21%. The denominator grew faster than the numerator. Seer's conclusion applies directly: "Always look at clicks and impressions separately before calling something a problem."

The honest reading of our own data is less flattering than the click growth suggests. A 0.27% CTR at an average position of 7.5 means we are earning impressions far faster than we are earning clicks, and that is exactly the pattern the AOP describes for technology content. Publishing volume bought reach. It did not buy proportional attention. That is a real cost, and it is the strongest argument in this dataset for changing what gets written rather than how much.

Which query formats trigger an AI Overview

If format is the variable that matters, it helps to know where AI Overviews actually appear. Seer analysed 49,353 distinct queries against SERP-feature intent classification.

Query type	AI Overview trigger rate	Organic impressions in sample
Comparison (X vs Y)	95.4%	79,436
Review queries	86.3%	21,221
Question (what/why/how/is/are)	85.9%	1,408,843
Price, cost or buy	83.4%	192,911
Best-of	81.3%	105,279
Near me	76.9%	188,606
Single word	27.3%	25,042,440

By intent, AI Overviews appear on roughly 36% of informational queries, 8% of commercial queries and 5% of transactional queries.

There is a tension here worth naming, because it cuts against the standard advice. The AOP data says opinion and interviews held up while how-tos collapsed. Seer's data says comparison queries trigger an AI Overview 95.4% of the time. Comparisons are the format most SEO guidance in 2026 recommends writing, including guidance we have published ourselves. Both can be true: a comparison page is very likely to sit under an AI Overview, and the click that survives is likelier to go to a cited source with a defensible point of view than to a neutral feature grid that the overview has already summarised.

The distinction is not topic. It is whether the page contains something the overview cannot restate: a benchmark you ran, a price you paid, a decision you made and regretted, a named person on the record.

What to change in the next two quarters

Neither study supports abandoning search as a channel, and neither supports carrying on unchanged.

Stop measuring format performance with a site-wide CTR number. The AOP spread runs from +63% to -54% inside a -7.1% average. A single site-level figure will hide every decision worth making. Segment your own Search Console data by content format before you segment it by anything else.

Treat clicks and impressions as separate series. Our own numbers and Seer's both show CTR collapsing while clicks rose. A falling CTR with rising clicks is a reach story. A falling CTR with falling clicks is a demand story. They need different responses.

Audit the how-to library first. Guide and how-to content fell more than 20% in the AOP sample, and finance and business how-tos fell 52%. That is where the exposure concentrates. Our own framework for choosing keywords that survive AI Overviews covers the selection side of this problem.

Chase the citation, not the ranking. Seer's +120% citation premium is the largest single lever in the dataset, and it operates on pages that already rank. The data on ranking versus AI Overview citation and the platform-level GEO playbook both sit on this point.

Add a point of view to reportage. This is the AOP's own recommendation and the hardest to execute, because it requires someone with standing to have an opinion and to sign it.

Do not solve a traffic problem by publishing more pages. Reeves is blunt about this failure mode: "Doubling down on declining content only accelerates losses, and publishers can no longer rely on traffic volume driven by audiences seeking basic information retrieval." Our own June and July figures are a working example of what expanding output does to CTR.

India-specific considerations

The AOP sample is entirely UK publishing groups, so the topic and format percentages should be treated as directional outside that market rather than transferable. Two differences matter for Indian publishers and Indian B2B marketing teams.

Query mix skews differently. A large share of Indian search volume runs through question-format and price-oriented queries, which Seer measured at 85.9% and 83.4% AI Overview trigger rates respectively. A content programme built mainly on "what is" and "cost of" pages carries more AI Overview exposure than the AOP's UK averages imply.

Language and regional coverage change the citation calculus. AI Overview coverage and citation behaviour are not uniform across languages, and a page in English competing for an India-targeted query sits in a different citation pool than the same page competing globally. Measure India-segmented Search Console data separately rather than inferring it from a global average.

On data handling, any first-party measurement programme that combines Search Console exports with CRM or user-level analytics needs to account for the Digital Personal Data Protection Act 2023. Search Console data itself is aggregated and anonymised, but the moment it is joined to identifiable user records for attribution, DPDP consent and purpose-limitation obligations apply to the joined dataset.

FAQ

What did the AOP AI publisher study actually measure?

The Association of Online Publishers analysed six months of traffic across eight major UK publishing groups, totalling 10.8 billion pageviews across 147 domains. DJB Strategies ran the traffic analysis and Ipsos ran a parallel consumer survey. The sample covered about 80% of participants' total traffic after long-tail URLs were removed.

Why does the date of the AOP data matter so much?

The study was published on 20 July 2026, but the traffic it measures is the Q1-to-Q2 2025 window. The projection that Google pageviews halve by Q3 2027 extrapolates a single 2025 quarterly step. Separate Seer Interactive data covering January 2025 to February 2026 shows the decline stopped compounding in early 2026.

Which content formats lost the most Google traffic?

Guide and how-to content fell by more than 20% quarter on quarter in the AOP sample, the worst-performing format. Finance and business how-tos fell 52% and food and drink how-tos fell 54%. By topic, technology fell 28% and finance and business fell 23%, making them the hardest-hit subject areas.

Which formats grew despite AI Overviews?

Opinion and commentary grew 4% and interviews grew 1%. Combining topic with format produced larger gains: news interviews rose 63% and health and wellness analysis rose 46%. By topic alone, sports grew 37%, travel 22%, health and wellness 7% and news 2% over the same quarter.

Did click-through rates actually recover in 2026?

Seer Interactive measured organic click-through rate on AI Overview queries rising from 1.31% in December 2025 to 2.36% in February 2026, an 85% increase off the floor. Seer explicitly declines to call this a recovery, describing it as a levelling out, and warns against forecasting from two months of data.

How much difference does an AI Overview citation make?

Across Seer's informational queries, being cited inside an AI Overview delivered about 120% more organic clicks per impression than appearing on the same results page uncited. Per million impressions that is roughly 20,743 clicks versus 9,445. Citation still underperforms a results page with no AI Overview by 38%.

Which query types are most likely to trigger an AI Overview?

Seer analysed 49,353 queries and found comparison queries trigger an AI Overview 95.4% of the time, review queries 86.3%, question-format queries 85.9%, and price or cost queries 83.4%. By intent, coverage runs at roughly 36% of informational queries, 8% of commercial and 5% of transactional.

Should a falling click-through rate trigger a strategy change?

Not on its own. Seer found brand-cited impressions doubling from 15.8 million to 33.1 million in a single month while clicks stayed flat near 400,000, which halved CTR without any click loss. Track clicks and impressions as separate series before treating a CTR decline as a demand problem.

How eCorpIT can help

eCorpIT is a CMMI Level 5 certified technology consulting organisation in Gurugram, and our senior engineering teams build the measurement layer that this kind of analysis depends on: Search Console and analytics pipelines segmented by content format, citation tracking across AI Overviews and assistant platforms, and reporting that separates clicks from impressions rather than collapsing both into a single CTR figure. We also run content and GEO programmes against those measurements for technology and B2B teams, which is the segment the AOP data shows to be worst affected. If you want your own format-level numbers before you rewrite a content roadmap, get in touch.

References

1. Richard Reeves, "Google search traffic to leading UK publishers set to halve by Q3 2027", Press Gazette, 20 July 2026.
2. Seer Interactive, "Search Engine Land and other search publications cite Seer's 2026 CTR AIO research" (announcement of the April 2026 study by Tracy McDonald, Hannah Cooley and Marketa Williams).
3. Seer Interactive, "Generative Engine Optimization".
4. Seer Interactive, "SeerSignals research and technology".
5. Alphabet Inc., "Consolidated Alphabet revenues in Q2 2025 increased 14%", SEC Exhibit 99.1, July 2025.
6. Alphabet Inc., "Alphabet Announces Second Quarter 2025 Results".
7. Association of Online Publishers, "Digital publishers firmly focused on growth as attitudes to AI mature".
8. InPublishing, "AOP CRUNCH: The post-search publishing playbook for finding and keeping audiences".
9. Search Engine Land, "Google AI Overviews CTR shows early signs of recovery: Study".
10. Search Engine Journal, "AI Overview CTR Fell 61%, But Clicks Didn't Collapse".
11. Press Gazette, "Publishers versus bots: How Trusted Reviews is fighting back against LLMs".
12. Google Search Console data for ecorpit.com, December 2025 to 20 July 2026, retrieved 21 July 2026.

Last updated: 21 July 2026.