Crazy Egg surveyed 49 shoppers on how they actually feel about buying through AI and social platforms versus going directly to brand websites. The results are counterintuitive — and strategically important. Most brands have been worrying about the wrong problem.
What the data shows
Awareness of in-platform shopping is nearly universal: 81% of respondents knew it existed. Adoption tells a different story — only 27% have ever actually used it. And when asked where they prefer to buy, 80% of shoppers chose brand websites over platforms. That preference, the survey notes, has not weakened over the past year.
The resistance to in-platform purchasing is not coming from users who don't understand the technology. It is coming disproportionately from those who understand it best. One participant articulated the pattern clearly: "I honestly think, because I'm so deeply familiar with how these platforms work, that I'm less likely to shop there. I don't want the data sharing, I don't want the influence from algorithms guiding my purchases."
ChatGPT is the hardest platform to shift sentiment on. 70% of respondents feel negatively about shopping through it — the highest of any platform tested. Only 4% are somewhat positive. Awareness is lowest there too: only 13 of 49 respondents even knew it was an option. The resistance is informed, structural, and unlikely to dissolve through UX improvements or better price points alone.
The wrong conclusion to draw
The tempting interpretation is that in-platform AI commerce is overhyped, that brand websites are safe, and that the existential threat to direct-to-consumer commerce has been postponed. This is the wrong conclusion — and drawing it would be a strategic mistake.
The Crazy Egg data tells us something specific: consumers do not want to purchase through AI platforms. It tells us nothing about whether they want to discover and evaluate through them. And the evidence on discovery and evaluation points in a completely different direction.
The Growth Memo research, published the same week, tracked 48 participants through 147 major-purchase tasks in Google AI Mode. 64% never clicked anything. They read the AI's shortlist, identified their finalists, and stopped. The participants who did click through went to brand websites — to confirm a decision that the AI had already shaped, not to begin a discovery process they would conduct themselves.
These two datasets describe the same consumer, at two different moments in the same journey. The Crazy Egg data captures the conversion moment: they want to buy from the brand, not from the platform. The Growth Memo data captures the evaluation moment: they are letting AI do the shortlisting work before they ever reach a brand site.
What this means for brand website strategy
The brand website is not dying. But its role in the purchase journey has changed structurally. It is no longer where discovery happens for a significant and growing share of high-intent consumers. It is where confirmation happens — after the AI has already determined which brands are worth considering.
This changes what the brand website needs to do, and it changes where the marketing investment needs to go upstream of the website.
A brand that focuses exclusively on its own site — optimising conversion flows, improving product pages, refining UX — is optimising the confirmation layer while neglecting the selection layer. The users it is optimising for are users who have already been shortlisted by AI. The users it is losing are the ones who never make it to the brand site at all, because the AI did not include the brand in its response.
Citation Presence: the degree to which a brand appears in AI-generated responses across the surfaces where high-intent consumers evaluate options before visiting any website.
The discovery gap
The gap between brand website trust (80% prefer it for purchasing) and AI shortlist inclusion is the most important strategic gap in modern marketing. Consumers want to buy from brand websites. But they are increasingly deciding which brand websites to visit based on what AI tells them.
If AI describes your brand accurately and compellingly in response to a relevant query, the user's preference for brand websites works in your favour. They read the AI's recommendation, they navigate to your site, they convert. The trust in the brand website is an asset — but only for brands that have already made it into the AI's shortlist.
If AI does not include your brand, or describes it inaccurately, the user's preference for brand websites is irrelevant. They will go to a competitor's brand website — the one the AI recommended — with the same preference intact.
This is the discovery gap. It is not visible in conversion analytics, because the users who never reach your site do not appear in your data. It is visible only in AI visibility measurement — in the Shortlist Rate, in the Citability Score, in the AI Visibility Index tracking of how often and how accurately your brand appears in relevant AI responses.
What trust-resistant consumers tell us about citability
The Crazy Egg survey contains one finding that is particularly relevant for citability strategy: the most resistant consumers are the most informed ones. The people least likely to buy through ChatGPT are the ones who understand best how these platforms work.
This is a signal about what these consumers trust. They do not trust platform-native commerce. They do trust brand websites. But they are using AI to decide which brand websites to trust in the first place. Their resistance to in-platform purchasing coexists with their reliance on AI-generated recommendations — and that coexistence is not a contradiction. It is the precise pattern that makes citability strategy urgent.
A brand that is accurately cited by AI, describes itself consistently across authoritative sources, and presents structured, machine-readable information about its value proposition will earn the recommendation of the most discerning AI users — the ones who will then navigate directly to the brand website to complete the purchase on their own terms.
The brand website is safe. The discovery layer is not. And the path between them runs through the AI shortlist.
Source: Crazy Egg — Aware, Uncomfortable, and Still Not Buying: How Consumers Really Feel About Shopping in ChatGPT and Google (April 2026). Survey of 49 shoppers on in-platform purchasing behaviour.