← All articles

AI Search by the Numbers: 12 Touchpoints, 8% Clicks — What Both Sides Get Wrong

By VonixCo Team · · 15 min read

An analyst's desk with a laptop showing AI search clicks data and charts of declining click-through rates

AI search clicks are falling for many publishers, even as Google says AI search is expanding discovery. The direct answer is that both claims can be true at the same time, because the two sides are measuring different things: Google points to shoppers visiting more touchpoints before they buy, while independent research measures how often people actually click through to a website when an AI summary is on the page. Pew found users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% without one, and only 1% of visits with AI summaries produced a click on a cited source. Google, meanwhile, cites its own commissioned survey in which AI-using shoppers visited 12 touchpoints, nearly three times as many as non-AI shoppers. Neither number alone settles the debate, and that is exactly what this article unpacks — the numbers, the disagreement, and what to measure instead.

The argument flared up publicly in November 2026. Google published an article on its Think with Google site in October, written by Brendon Kraham, vice president of Search and Commerce Global Ads Solutions, arguing that AI search expands consumer discovery. SEOs pushed back, citing declining clicks and limited reporting, and Search Engine Land covered the exchange in its report on Google’s AI search advice and the SEO backlash, written by Danny Goodwin and published on 20 November 2026. This article follows that report’s facts and figures, adds no new statistics of its own, and does not present either side’s numbers as proof that the other side is wrong.

What happened in the November 2026 AI search debate?

Google’s case is that AI features such as AI Overviews and AI Mode help people discover more, not less. In the Think with Google article, Kraham argued that AI search expands consumer discovery — that people using AI tools encounter more products, brands and ideas on their way to a purchase than they would have done with traditional search alone. The advice that followed was, in essence, to focus on business outcomes rather than on the click as the only measure of success.

That framing landed badly with many SEOs. Their objection was practical rather than philosophical. For publishers and the brands that rely on organic search, a click is not an abstract metric: it is the visit that becomes a reader, a subscriber, a lead or a sale. If AI summaries answer the question on the results page, the visit may never happen, however many touchpoints came before it.

Two further complaints ran through the response. The first was accuracy: recipe publisher Inspired Taste accused Google of generating inaccurate versions of its recipes in AI Overviews while using its brand name — a charge that goes to the heart of trust, because a wrong answer attributed to a real brand harms the brand and misleads the reader. The second was reporting: critics said Google was asking them to measure business outcomes while withholding the data needed to do so.

It helps to see this debate in context. VonixCo has covered earlier click data in our guide to Google AI features click data, which looked at a different study. This article is about the November backlash itself — the Pew, Seer Interactive and Ipsos numbers, and why they appear to point in opposite directions.

What do the numbers say about AI search clicks?

The three numbers most quoted in the debate come from three different sources, measuring three different things. Put side by side, they look contradictory. Read carefully, they describe different stages of the same journey: how many places a shopper touches, how often a searcher clicks, and how often a click goes to a cited source.

Source What it measured Headline number
Google-commissioned Ipsos survey (Dec 2025, online shoppers in 25 markets) Touchpoints visited by shoppers using AI, and whether AI Overviews or AI Mode helped them decide 12 touchpoints for AI-using shoppers, nearly 3× non-AI shoppers; 3 in 4 said AI helped them decide faster and more confidently
Pew Research (900 US adults, March 2025) Clicks on traditional results and on cited sources when an AI summary appeared 8% of visits clicked a traditional result with an AI summary, vs. 15% without; only 1% of visits produced a click on a cited source; sessions ended after 26% of searches with AI summaries, vs. 16% without
Seer Interactive (June 2024–Sept 2025) Organic click-through rate for informational queries with AI Overviews Organic CTR fell 61% over the period

Start with Google’s evidence. The Ipsos survey, commissioned by Google and carried out in December 2025 among online shoppers in 25 markets, found that shoppers who used AI visited 12 touchpoints on their journey, nearly three times as many as shoppers who did not use AI. Three in four of those AI-using shoppers said AI Overviews or AI Mode helped them decide faster and more confidently. On Google’s reading, that is discovery expanding: people are encountering more brands, products and ideas before they buy.

Now look at the counter-evidence. Pew Research studied 900 US adults in March 2025 and measured what happened on the results page itself. When an AI summary appeared, users clicked a traditional result in 8% of visits, compared with 15% when no AI summary appeared. Only 1% of visits with AI summaries produced a click on a cited source — the link inside the summary itself. And sessions ended after 26% of searches with AI summaries, against 16% without, meaning the searcher simply stopped there. If the click is the measure, AI summaries are associated with fewer visits to websites, not more.

Seer Interactive’s data points the same way, over a longer window. Tracking informational queries that triggered AI Overviews, Seer found organic click-through rate fell 61% between June 2024 and September 2025. That matches what many publishers have reported from their own analytics: informational content, the kind AI summaries answer most easily, has lost the most clicks.

None of these three measurements is wrong on its own terms. The mistake is treating any one of them as the whole story. A shopper can visit twelve touchpoints and still never click through to a publisher’s site, because the touchpoints may include the AI summary itself, a product listing or a video — and because a satisfied searcher may stop searching altogether, as Pew’s session-ending figure suggests. Equally, click-through rate can fall sharply while total discovery rises, if the AI summary sends people to more places but takes a smaller share of each visit.

An analyst reviewing search traffic charts on a monitor, tracking how AI search clicks have changed

Why do the numbers disagree?

The numbers disagree because the two sides are counting different things, using different methods, at different moments — and because each side’s evidence has limits its own supporters do not always mention.

Google is counting touchpoints, not clicks

A touchpoint is any point of contact on the way to a purchase: a search, a summary, a product page, a review, a video, a store visit. Google’s Ipsos figure says AI-using shoppers touched more of them. That is a claim about the length and richness of the journey, not about how many of those touchpoints were clicks to an independent website. A journey can grow longer while the number of clicks to publishers shrinks, if the extra touchpoints happen inside Google’s own results.

That is the criticism Mathias Flink made: the 12 touchpoints may all have happened on Google-owned properties, and Google’s summary gives no breakdown of where they occurred. Without that breakdown, the figure cannot answer the question publishers are actually asking, which is how much of the discovery happened on their sites. The number is real, but it does not prove what it is often quoted as proving.

The Ipsos survey was commissioned by Google and relies on self-reporting

Two cautions belong with the Ipsos figures. First, the survey was commissioned by Google, which has an interest in the result; that does not make the data false, but it is a reason to read it alongside independent measurements rather than alone. Second, it is based on self-reported shopping behaviour — what people say they did and how confident they felt — rather than on observed clicks.

Pew and Seer measured clicks directly, but on different things

Pew and Seer are independent measurements, and they measured behaviour rather than opinion: whether people clicked, and how often. But they measured different outcomes from each other, too. Pew looked at clicks on traditional results and on cited sources when an AI summary appeared, across a panel of 900 US adults in March 2025. Seer tracked organic click-through rate for informational queries with AI Overviews over fifteen months.

The timeframes matter as well. Pew’s data is from March 2025; Seer’s runs from June 2024 to September 2025; Ipsos surveyed shoppers in December 2025. AI features changed substantially across that period, and a number from one moment is not automatically a description of the next.

The reporting gap is the part both sides should agree on

The sharpest criticism in the backlash was not about any single number. It was about reporting. Daniel Foley Carter put the publisher experience bluntly, saying AI overviews have been a nightmare for SEOs. Joe Youngblood made the measurement point: “How are we supposed to ‘track real outcomes’ when you only give us impression data on a per page basis” — with no prompt-level or click data, and with Google Search Console not exposing every underlying search behind an AI response.

That complaint deserves attention because it is testable in a way the touchpoint debate is not. Google’s advice is to measure business outcomes, but the reporting it provides gives publishers impression data on a per-page basis, not the prompt-level or click-level data that would show which AI responses featured their content and what happened next. Search Engine Land’s bottom line was the same: Google says AI search creates discovery and wants business outcomes measured, but its reporting does not show how AI visibility turns into clicks, leads or sales.

A shopper using AI search on a phone while browsing products in a store

What should I measure instead?

If the click is shrinking and the touchpoint count cannot be verified, the sensible response is not to pick a side in the argument. It is to build a measurement routine that tells you what is actually happening to your own site, and to make your content the kind of source AI systems cite — because, as Pew’s 1% figure shows, being cited matters more than ever, and being the cited source is now the scarce position.

  1. Separate your traffic by query type. Track informational queries and commercial queries separately. Seer’s 61% decline was for informational queries with AI Overviews; your commercial and transactional pages may behave very differently. A blended average hides the split.
  2. Watch sessions, not just clicks. Pew found sessions ended after 26% of searches with AI summaries. Check whether your own analytics show the same pattern: are sessions from AI-triggered queries shorter, or do they end without a visit at all? That tells you whether the summary is satisfying the query or simply diverting it.
  3. Track brand searches and direct visits. If discovery is expanding but clicks are falling, some of the value may show up as people searching for your brand name later, or arriving directly. Brand search volume and direct traffic are imperfect proxies, but they are visible in your own data, unlike touchpoints inside Google’s results.
  4. Monitor which pages AI systems cite. Ask the assistants your buyers use the questions your pages answer, and note which sources they cite. Our guide to how Google rankings and AI recommendations overlap explains why a page can rank well and still not be the one an assistant recommends — and what to check when that happens.
  5. Measure what a visit is worth, not just how many there are. Fewer, better-qualified visits can be worth more than a larger number of casual ones. For a practical framework, see our guide to AI traffic quality benchmarks and measuring ROI, which sets out how to judge AI-referred traffic on conversions and revenue rather than on volume alone.
  6. Keep a simple weekly log. Record, for the same set of queries each week, whether an AI summary appears, whether your site is cited in it, and how many clicks and conversions those queries produced. A short, consistent log beats a one-off audit, because it shows direction of travel — and direction is what the headline numbers cannot give you.

Compare like with like: the same queries, the same pages, week after week. Separate what you can observe from what you are told: your analytics are observed, a vendor’s touchpoint figure is not. And do not let a single metric decide the strategy — a falling click-through rate on informational content is a real problem, but it is a different problem from a falling conversion rate, and the fix for each is different.

When should you hire a service like VonixCo?

The click squeeze changes what link building is for. When only 1% of visits with AI summaries produce a click on a cited source, the value of being cited goes up, not down — the cited source is the one that still gets the visit, the mention and the authority. Earning citations and mentions that survive the click squeeze means being the source an AI assistant trusts enough to name: original, specific, well-sourced content on real websites in your niche, linked to pages that answer the question properly.

That is work a service earns its place doing when you cannot sustain it yourself. VonixCo is a niche-relevant guest posting and SaaS link building service based in Lahore, Pakistan; services are SaaS guest posts, niche guest posts, link building and guest post content writing; plans are $500 / $950 / $1,800 per month with content included and a live link report; contact info@vonixco.com and WhatsApp +92 340 0096621. A service like VonixCo fits when you want guest posting handled as an ongoing operation — finding relevant sites, writing content that can be cited, and placing it where your buyers’ questions are actually answered. This is for SaaS brands and agencies that already have content worth citing and want it placed and linked consistently. Before hiring anyone, ask how they choose sites, whether they report live links, and how they measure whether placements earn citations and visits.

Frequently asked questions about AI search clicks

How many AI search clicks does a website get when an AI summary appears?

Pew found users clicked a traditional result in 8% of visits when an AI summary appeared, compared with 15% without one, and only 1% of visits with AI summaries produced a click on a cited source. Those are averages across the panel studied, not a prediction for any single website — your own rate will depend on your queries, your position and whether you are cited.

Why do Google’s numbers look better than Pew’s and Seer’s?

Google’s Ipsos survey measured touchpoints — how many points of contact shoppers visited — while Pew and Seer measured clicks. A shopper can visit more touchpoints and still click less, especially if the extra touchpoints happen inside Google’s own results. The numbers disagree because they measure different outcomes.

Did organic click-through rate really fall 61%?

Seer Interactive found organic click-through rate for informational queries with AI Overviews fell 61% between June 2024 and September 2025. That figure covers informational queries that triggered AI Overviews, not all searches, so it should not be quoted as a fall in all organic traffic — but for informational content it is a substantial decline.

What is the 12-touchpoint figure, and can I trust it?

The 12-touchpoint figure comes from a Google-commissioned Ipsos survey of online shoppers in 25 markets in December 2025, which found AI-using shoppers visited 12 touchpoints, nearly three times as many as non-AI shoppers. It is a real survey result, but it was commissioned by Google, based on self-reported behaviour, and gives no breakdown of where the touchpoints happened — so read it alongside independent click data, not instead of it.

How can I track AI search clicks if Google does not report them?

Google Search Console gives impression data on a per-page basis, not prompt-level or click-level data for AI responses, so you cannot track AI search clicks directly in the same way as ordinary clicks. The practical workaround is to separate informational and commercial queries, monitor brand searches and direct visits, log which of your pages AI assistants cite, and measure conversions from AI-referred traffic rather than volume alone.

Should I still invest in guest posting if AI search sends fewer clicks?

Yes, if the goal is citations and authority rather than raw click volume alone. When only 1% of visits with AI summaries produce a click on a cited source, being the cited source is the position that still earns visits, mentions and trust. Guest posts on relevant sites in your niche are one of the ways to earn those citations — provided the content is original, specific and worth citing.

Conclusion

The November 2026 debate is best read as a disagreement about measurement, not about whether AI search exists or matters. Google’s 12 touchpoints describe a longer journey; Pew’s 8% and 1% describe fewer clicks at the end of it; Seer’s 61% decline describes what happened to informational queries along the way. All three can be true, because they measure different things — and none of them, alone, tells you what is happening to your site. The publishers who come out ahead will be the ones who measure their own queries, track citations as well as clicks, and make their content the source an AI assistant names. Disclosure: Search Engine Land, the primary source for this article, is owned by Semrush.

Want niche-relevant guest post placements?

Tell VonixCo about your website and niche, and we will reply with suitable publishing options and pricing.

Keep reading