Growth Through Content · Research · October 2026
Forty-seven surveys and 252,123 responses from three years of fieldwork, all on how ChatGPT, Gemini, Claude and other AI tools figure in what people buy: whether they use them, trust them and act on what they say. Survey after survey points the same way.
Why this report exists
GEO is not a distraction
Marketers keep being told that optimising for AI answers is a sideshow. AI sends little traffic, hardly anyone buys inside a chatbot, so why bother?
We went looking for the evidence. We gathered the published surveys from the last three years that asked consumers or business buyers whether AI tools such as ChatGPT, Gemini and Claude affect what they buy: 47 surveys with 252,123 responses between them.
They point the same way. AI is now one of the most influential sources in the buying journey. It shapes which brands make the shortlist and which get ruled out, often before anyone clicks a link. That influence is substantial, and referral analytics alone cannot measure it.
Key findings
Consumers keep telling researchers the same thing
- 69%Seven in ten AI shoppers remember AI changing a purchase
Among US adults who had used AI to shop, 68.64% can definitely remember AI influencing them to buy something they would not otherwise have bought, and 36.89% say it has happened many times. That is about 53% of all 1,009 people surveyed (Exploding Topics).
Exploding Topics · report ↗ - 41%Four in ten bought what AI recommended
Klaviyo surveyed 7,998 consumers in eight countries: 41% had bought an AI-recommended product in the previous six months.
Klaviyo · press release ↗ - 62%AI is switching brands among the people who use it
Among US shoppers who used a chatbot to shop, 62% bought a different product or brand from their usual choice because of AI (Attentive, 3,054 adults).
Attentive · report ↗ - ~40%People say AI led them to buys they would not have considered
Nearly 40% of Americans say they bought something they would not normally have considered because of an AI tool (Visa and Morning Consult, 2,000 adults).
Visa · press release ↗ - 57.5%AI vetoes as often as it recommends
57.5% of AI users have been talked out of a purchase by a chatbot, almost exactly the share who bought on a chatbot’s advice. The two groups overlap: one person can be talked into one purchase and out of another (Semrush, 2,338 adults).
Semrush, Jul 2026 · report ↗ - 69%Business buyers feel it most
In G2’s survey of 1,076 software buyers, 69% said an AI chatbot changed the vendor they finally chose, and one in three bought from a vendor they had never heard of.
G2, Mar 2026 · press release ↗
The headline number
About 1 in 3 survey responses report AI shaping a purchase
Thirteen surveys publish both a sample size and the share a figure applies to, so responses can be counted. Across their 63,626 respondents, 20,757 reported AI shaping a purchase, a brand choice or what they discovered and went on to consider: 32.6%. The surveys ask different questions of different groups, so this is a pooled descriptive figure, not an estimate for the whole population. Each row shows what was measured.
| Survey | Sample | Rate* | People |
|---|---|---|---|
| Total | 63,626 | 32.6% | 20,757 |
*Rate is the share of the whole sample. Where a survey reports a figure only among AI users, it is multiplied by that survey’s own share of AI users, so the rate is never inflated. The Accenture row combines two figures from the same report and is an estimate; without it the rate is 31.3% (14,200 of 45,412). Exploding Topics is counted at 53.3% of its full sample: 68.64% of the 77.6% who had shopped with AI. McKinsey’s measure is the strictest here: AI as the single biggest influence.
Every direct-influence figure
What each survey found, side by side
Surveys ask different questions of different groups, so these bars are not one scale. Solid bars are shares of everyone surveyed. Striped bars are shares of people who already use AI.
The veto effect
AI talks people out of buying, too
Influence runs both ways. A brand that AI describes badly, or never mentions, loses buyers before they ever search for it, and that loss shows up in no dashboard.
Semrush and Exploding Topics · 2,338 US adults · Jul 2026
Share of AI users. Among people who use chatbots to research purchases, 80.93% have dropped a purchase on AI advice.
Semrush, Jul 2026 · report ↗Attest 2027 Consumer Trends · 4,000 US and UK adults · Sep 2026
Of shoppers who used AI on a first-time purchase, the share who ruled out a product on its advice without checking anywhere else. Across both countries the figure is 56%.
Attest · press release ↗ Attest · full report ↗Skyword and Dynata · 1,000 US adults · Apr 2026
Took significant action on AI-generated information, avoided a purchase because of it, or switched brand because of it.
Skyword · report ↗Kantar’s UK barometer adds a quieter version: 15% of consumers assume a brand AI does not suggest is not for them. Kantar · press coverage ↗
Discovery
AI is introducing people to brands they had never heard of
Before a purchase there is a shortlist. These surveys measured how often AI puts an unfamiliar brand on it.
Why attribution misses it
AI shapes the shortlist. Another channel gets the credit.
Few people go straight from an AI answer to the checkout. Most check first, on Google, on review sites or on the brand’s own website, and analytics usually credit whichever of those channels they touched last.
Verification channels: Semrush, Dec 2025 · report ↗
These are the figures sceptics cite against GEO. Read with the rest of the report, they describe the mechanism that hides AI’s influence from last-click reporting rather than an absence of influence.
The evidence is piling up
Consumers surveyed, by year of fieldwork
Research on AI and purchasing barely existed in 2023. By 2025 the big consultancies, payment networks and retail platforms were all asking.
Consumer surveys only (43 studies, 227,499 responses). Dated by when fieldwork ran, not when results were published. This chart shows how much research is being done, not how fast AI influence is growing.
Business buyers
In B2B, the chatbot writes the shortlist
The full evidence
All 47 surveys
Open any survey for its method, who was asked and the source. Direct influence means the survey asked whether AI changed a purchase, a choice or a decision. In the journey means it measured AI use or trust while shopping.
Conclusion
What the evidence says, plainly
- AI influences purchases at scale.Across the 13 surveys that can be counted, about 1 in 3 responses report AI shaping a purchase or a choice. Among AI shoppers in McKinsey’s 11-market study (5,095 people describing a recent purchase), AI chat and search was the most-cited single influence: 29%, against 9% for search engines.McKinsey ↗
- It works before the click.93% of AI-assisted shoppers looked for information outside the chatbot before buying. When they then arrive through Google, a review site or your own website, AI’s part in the decision does not show up in referral reports. Low AI referral traffic is not the same as low AI influence.Attentive ↗
- It cuts both ways.57.5% of AI users have been talked out of a purchase by a chatbot. A brand that AI leaves out, or describes badly, can lose buyers without ever seeing them in its data.Semrush ↗
- Adoption is accelerating.The share of consumers who have replaced search engines with generative AI for product recommendations rose from 25% in 2023 to 58% in 2024. In the US, the share of AI users who have bought something through an AI tool rose from 10% to 17% in six months.Capgemini ↗McKinsey ↗
- AI changes who wins demand, even when it does not create more of it.If someone was always going to buy a laptop and AI recommends Brand B over Brand A, B wins the sale and total demand is unchanged. That redistribution is exactly what GEO competes for. These surveys show that AI influences buying decisions; they do not measure how much new demand AI creates, or how reliably any specific GEO tactic changes what AI recommends. Those need testing brand by brand.
GEO is not a distraction. Judging it by clicks is.
Judge it by whether AI recommends you when real buyers are choosing, not by the referral traffic it sends.
Method and caveats
How this report was built
What was included
Published surveys of consumers or business buyers, fieldwork from October 2023 to September 2026, that report a sample size and ask about AI tools such as ChatGPT, Gemini, Claude, Copilot or Perplexity in shopping or buying. Each figure was checked against the publisher’s own release or report where one was reachable. This is a curated evidence review, not a systematic review or meta-analysis.
Totals are summed samples
252,123 is the sum of the 47 sample sizes. Panels overlap and some publishers ran more than one wave, so it is a count of responses rather than unique people. Exploding Topics is owned by Semrush; their two surveys use different samples (1,009 and 2,338) but are not fully independent. B2B surveys (4 studies, 24,624 responses) are kept separate from every consumer figure. Samples stated as “more than” or “nearly” are counted at the stated figure.
Questions differ
“Used AI to shop” and “bought because of AI” are different claims, so the report keeps them apart. The 1 in 3 figure uses only the 13 surveys where people can be counted, and converts every rate back to the full sample.
Self-reported, often commissioned
Every figure is what respondents said, and many studies come from companies that sell marketing or commerce tools. Independent research firms and consultancies (McKinsey, BCG, Forrester, Deloitte, Kantar, IBM) carry the most weight and point the same way. None of these surveys tests whether a specific GEO tactic changes what AI recommends; they show that AI’s recommendations matter to buyers.
Every figure is linked
Each number on this page links to its source. 37 of the 47 surveys link to the publisher’s own report or press release; the other 10 have no publisher release we could find online and link to independent press coverage, marked as such. All links were checked on 30 October 2026. Studies that published a striking figure without a sample size (a16z, Morgan Stanley, Akeneo, Bloomreach, Statista) are left out of every total.
Counter-evidence is included
Low trust in AI checkout, heavy verification and HSBC’s finding that only 7% of US investors say AI drove their last major decision are reported, not hidden. They describe how AI influences decisions, not whether it does.
Use this research
Cite this report
Holland, A. (2026). The AI Influence Report: 47 surveys and 252,123 survey responses on how AI influences buying decisions. Growth Through Content, 30 October 2026. https://growththroughcontent.com/ai-influence-report/
Questions people ask
How many consumers say AI influences what they buy?
Across 13 surveys that allow a head count, 20,757 of 63,626 responses (32.6%) reported AI shaping a purchase or a choice. The surveys measure slightly different things, so this is a pooled descriptive figure. Single surveys put it higher: 41% bought an AI-recommended product in six months (Klaviyo), and nearly two-thirds of Americans say AI influenced a recent purchase decision (LDWW).
Is AI search a real marketing channel if it sends little traffic?
The surveys say yes. AI mostly works before the click: it builds the shortlist, and shoppers then verify on Google, review sites and brand websites before buying. 93% of AI-assisted shoppers looked outside the chatbot before purchasing, so last-click analytics credit those later steps rather than the AI that started the journey.
Which AI tools influence shopping most?
ChatGPT leads in most surveys that name tools, followed by Google Gemini, though domestic platforms lead in China and ChatGPT and Gemini are level in Japan (McKinsey). Semrush found 64% of AI users use ChatGPT monthly and 49% use Gemini; Claude and Perplexity reach smaller, more professional audiences.
Can AI stop people buying from a brand?
Yes. 57.5% of AI users have been talked out of a purchase by a chatbot (Semrush), and 62% of US shoppers using AI on a first-time purchase ruled out a product on its advice without checking elsewhere (Attest).
Do people let AI complete purchases for them?
Rarely, so far. Only 2.18% of AI shoppers completed a whole purchase inside an AI tool (Exploding Topics), and OpenAI moved away from in-chat checkout in March 2026. The influence happens in research and selection, not at the till.