AI Search in 2026: The Data Behind the Case for Client Content

Most of this page exists so you can steal from it. If you are building the slide that explains to a client why their content programme needs to change shape, these are the figures worth citing and the ones worth being careful with. Every number below carries a footnote to the page it came from; if a figure has no footnote, we could not find its source and we took it out.

The three numbers that carry the argument

94% of business buyers used AI during their most recent purchase, up from 89% the year before.¹ The buyers have already moved.

−61% is the drop in organic click-through rate on Google queries where an AI Overview appears: from 1.76% to 0.61% between June 2024 and September 2025, with paid click-through down 68%.² Google’s own summaries are absorbing the clicks that used to reach Google’s own listings.

87% of US businesses that rank on Google’s first page do not appear in AI-generated answers, in one vendor’s December 2025 benchmark.³ Treat the exact figure with care, it is a single vendor’s sample, but the pattern it describes is consistent with everything else on this page: traditional search visibility does not transfer.

The third figure is the commercially interesting one for an agency. The first two describe a problem every client can feel in their analytics. The third says that ranking well is not the same as being named, which is the gap a retainer fits into.

What is happening to the traffic line

Google’s global search share fell below 90% in late 2024 for the first time since 2015, and has stayed there for most of 2025.⁴ Searches per US user fell by nearly a fifth year on year.⁵

The click story is sharper than the query story. In the first four months of 2026, 68% of US Google searches ended without a click to the open web, up from 60% in 2024.⁶ Where an AI summary appears, Pew found users clicked a result on 8% of visits, against 15% without one.⁷

Publishers show the effect at full scale. Similarweb tracked organic traffic to news sites falling from a peak above 2.3 billion visits a month in mid-2024 to under 1.7 billion by mid-2025, while the share of news-related searches ending without a click rose from 56% to 69% over the same year.⁸ If your client publishes to earn sessions, that is the trend line they are on.

Meanwhile AI assistants sent 1.13 billion referral visits to the world’s top 1,000 websites in June 2025, up 357% on the year before, against 191 billion from Google search in the same month.⁹ Both facts are true and both belong in the slide.

Where buyers now start

McKinsey’s October 2025 research is the most quotable source for a considered-purchase client. Half of US consumers surveyed now deliberately use AI-powered search, 44% of those users call it their primary source of insight, ahead of traditional search at 31%, and McKinsey projects $750 billion in US revenue flowing through AI-powered search by 2028.¹⁰

Treat the revenue projection as market context and say so out loud. It is a category forecast, not a number any programme can promise a client, and per-client revenue attribution depends entirely on their own analytics, CRM or commerce platform.

The behavioural finding is the more useful half. Buyers are arriving at a shortlist before they arrive at a website. If your client is not in the shortlist, the website’s conversion rate is irrelevant to that buyer.

What AI actually reads

Yext analysed 6.8 million AI citations across 1.6 million queries per model on ChatGPT, Gemini and Perplexity between July and August 2025, and concluded:¹¹

“86% of AI citations come from sources that businesses already control - their own websites, their own business listings, and their own profile pages.”

Yext, October 2025¹¹

For an agency this is the single most encouraging number in the whole dataset. The material the assistants synthesise is, overwhelmingly, material the client can commission. It is not locked behind a platform relationship or an ad budget. It is writing that either exists or does not.

Yext’s one-line summary of the differences between platforms is worth memorising for client meetings: “Gemini trusts what your brand says. ChatGPT trusts what the internet agrees on. Perplexity trusts industry experts and customer reviews.”¹¹

What the academic research says works

The foundational study on Generative Engine Optimisation, from Princeton, Georgia Tech, the Allen Institute and IIT Delhi and published at ACM SIGKDD 2024, tested content techniques and measured their effect on visibility in AI-generated answers.¹² It remains the most rigorous thing anyone has published on the question, and every finding in it is about editorial craft rather than technical configuration.

Technique	Visibility against baseline12
Adding quotations from sources	+41%
Adding statistics	+33%
Citing sources	+28%
Traditional keyword stuffing	Below baseline (−9%)

All four figures are from the paper’s own results table.¹²

The finding that should shape how you pitch: these gains land hardest on brands that are not already winning traditional search. In the same study, citing sources produced a 115% visibility increase for a site ranked fifth in conventional results.¹² For a challenger client who has spent five years failing to outrank an incumbent, that is a genuinely different proposition.

Concentration is already setting in

Citation behaviour has a winner-take-most shape. Across more than three million US queries, YouTube alone accounts for 22.9% of the pages Google’s AI Overviews cite, and social platforms hold four of the top five slots.¹³ The sources assistants lean on are consolidating while most brands are not looking.

That is the window, and it is described honestly as a window rather than a guarantee: early, structured, regularly refreshed content gives assistants better source material and improves the odds a client is understood and referenced. It does not promise a position.

The traffic AI does send

AI-referred visitors convert at materially higher rates than organic search visitors in every published study we have seen. One analysis of 12.3 million visits across 347 businesses between January and September 2025 found 14.2% against 2.8% for Google organic.¹⁴ Semrush’s June 2025 research puts the average AI-referred visitor at 4.4x a traditional organic visitor.¹⁵

The mechanism is obvious once stated: a visitor arriving from an AI recommendation has already described their need, received a curated answer, and clicked through to evaluate a specific suggestion. They arrive at the vendor stage, not the awareness stage.

These are market benchmarks and should be labelled as such in front of a client. What can be tracked directly on published work is narrower and more honest: how the articles perform, how many readers click through to the client’s own site, and when the assistants last fetched the page. Full conversion and revenue attribution needs the client’s own analytics stack.

The honest caveats

Leave these in. An agency that presents only the alarming half of this dataset gets caught.

Google handled more than five trillion searches in 2024, about 14 billion a day, and grew by around a fifth that year.¹⁶ AI assistants still account for under 1% of website traffic against organic search’s 48.5%, in SE Ranking’s study of nearly 64,000 sites.¹⁷ ChatGPT reached 900 million weekly users in February 2026, up from 800 million the previous October, and it is still a fraction of Google’s volume.¹⁸

None of that changes the direction. It changes the claim you are entitled to make. The defensible version is: a growing and disproportionately valuable share of buying research now happens inside assistants, ranking on Google does not put a client into those answers, and the material those assistants read is material a client can commission.

What to do with this

The data supports a specific shape of work rather than a general urgency.

Find out which buying questions in the client’s category the assistants currently answer with somebody else’s name. That list is the brief - it is written by the market, not by a strategist guessing. Write against it: comparison sections, direct answers to specific evaluation questions, fact-dense material with cited sources, in the formats the Princeton study found actually move citations. Publish under the client’s brand. Then re-ask the same questions, see what moved, look at what the published articles earned in readers and click-throughs to the client’s site, and let both of those steer the next month’s brief.

That is a loop with a named editor’s sign-off in it - yours or your client’s - not an engine that runs unattended. But it is the difference between publishing on instinct and publishing against evidence.

Found by AI monitors how leading AI assistants answer buying questions, produces the content that fills the gaps, and reports on what it earns - sold only through white-label agency partners, under their brand, never in front of their client. See how the partner programme works.

Sources

Every figure above links to the page it was taken from. Checked on 10 September 2026.

Footnotes

1. Forrester, “B2B Buyers Make Zero-Click Buying Number One”, Buyers’ Journey Survey, January 2026. ↩

2. Seer Interactive, “AIO Impact on Google CTR: September 2025 Update”, 3,119 queries across 42 organisations, June 2024 to September 2025. ↩

3. NEWMEDIA.COM, “New RankOS Benchmark Finds 87% of U.S. Businesses Are Invisible in AI Search Despite Ranking on Google”, press release, 22 December 2025. A vendor benchmark. ↩

4. StatCounter, Search Engine Market Share Worldwide; reported by Search Engine Land, “Google’s search market share drops below 90% for first time since 2015”, January 2025. ↩

5. Datos and SparkToro, reported by Search Engine Land, “Google searches per U.S. user fell nearly 20% YoY”, 2025. ↩

6. SparkToro, “In 2026, Less than One Third of Google Searches Still Send a Click”, US clickstream data, January to April 2026. ↩

7. Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, browsing data of 900 US adults, March 2025, published 22 July 2025. ↩

8. Similarweb, “Zero-Click Searches”, news-related Google searches, May 2024 to May 2025. ↩

9. Similarweb, “AI Referral Traffic Winners By Industry”, June 2025, reported by TechCrunch, “AI referrals to top websites were up 357% year-over-year in June, reaching 1.13B”, 25 July 2025. ↩

10. McKinsey & Company, “New front door to the internet: Winning in the age of AI search”, AI Discovery Survey of a representative US consumer panel, August 2025, published October 2025. ↩

11. Yext, “AI Doesn’t Rank, It Cites. And 86% of Its Sources Are Brand-Managed”, 6.8 million citations, 1.6 million queries per model, July to August 2025, published October 2025. ↩ ↩² ↩³

12. Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande, “GEO: Generative Engine Optimization”, Proceedings of the 30th ACM SIGKDD Conference, 2024. Table 1 (Position-Adjusted Word Count: baseline 19.5; Quotation Addition 27.8; Statistics Addition 25.9; Cite Sources 24.9; Keyword Stuffing 17.8) and Table 2 (rank-5 sites, Cite Sources, +115.1%). ↩ ↩² ↩³ ↩⁴

13. Ahrefs, “The 50 Most-Cited Websites in Google AI Overviews”, more than 3 million US queries, September 2026. ↩

14. Superprompt, “AI Search Traffic Converts 5x Better Than Google: 2025 Conversion Data from 12M Visits”, 12.3 million visits across 347 businesses, January to September 2025. A vendor’s own data. ↩

15. Semrush, “AI Search Traffic Study”, 9 June 2025. ↩

16. SparkToro, “New Research: Google Search Grew 20%+ in 2024; receives ~373X more searches than ChatGPT”, 2025, drawing on Google’s own disclosure of more than 5 trillion searches in 2024. ↩

17. SE Ranking, “How Does Social Media Traffic Compare to Organic and AI in 2025?”, 63,987 websites, 2025. ↩

18. OpenAI’s announcement, reported by TechCrunch, “ChatGPT reaches 900M weekly active users”, 27 February 2026. ↩