Where AI Shoppers Actually Go: The Real Journey Behind the ChatGPT Recommendation

This is the second chapter of the same dataset. In our first study we showed that AI platforms expand search behavior rather than replace it. Here we go deeper into the same panel of real, permissioned consumers and cut it three new ways: the order in which platforms get touched first, an AI Index that scores who is winning the recommendation, and the comparison density of the answers themselves. The picture that emerges is the question every AI-visibility tool leaves on the table: once the AI names your brand, where does the shopper actually go?

AI discovery is a small but accelerating new first touch

The first-touch curve sat flat at roughly 1–2% through early 2024, then accelerated sharply from April 2025: 3.0% in March, 7.3% by June. Google's first-touch share held at 96.9% across the full window and 92.7% in the latest month — and because the panel's most recent periods are still filling in, that 7.3% is a floor, not a ceiling. The trajectory is the signal. AI is taking an early sliver of the discovery moment while Google keeps growing on verification and transaction. The two are running in parallel, not in a zero-sum trade.

Look inside the users who are active on both platforms in the same month and the complementary pattern sharpens. Google still comes first 84.9% of the time; ChatGPT leads the session 15.1% of the time. That within-session first-mover order is only visible in Measure's behavioral data — no analyst firm can stitch which platform a real person actually touched first. And the AI touches that do happen are disproportionately valuable: within classified ChatGPT prompts, Retail and Shopping carries the highest purchase-intent score of any topic at 0.81 out of 1.0. The AI channel is small, but every visit to it skews high-intent.

Owning Google search no longer guarantees you win the AI recommendation

To measure who is winning the AI moment, we built an AI Index: each brand's share of AI recommendations relative to its branded Google search share, where 100 means parity (Jan–Oct 2025). The new first touch is reshuffling the winners. Apple is the only major brand near parity at 86, taking 14.3% of AI recommendations against 16.6% of Google search. Amazon scores just 28 — 5.8% of AI recommendations against 21.1% of Google search. Samsung sits worst in the set at 16. Search-volume dominance and AI recommendation share are now two different races.

Amazon owns roughly one in five branded product searches on Google — but captures only 28 cents of AI recommendation share for every dollar of search mindshare. Apple is the only major brand that wins in AI at close to the rate it wins in search.

The demand and supply sides tell opposite stories for retail platforms. Amazon is the single most-named brand in user prompts at 6.5% of branded commercial queries — people ask ChatGPT about Amazon to compare their options — yet the assistant rarely recommends it back. Meanwhile a new class of AI-native destination is surfacing above its commercial search weight: TikTok appears in 3.7% of AI responses, Roblox 2.3%, YouTube 2.2%. Brands like Starbucks (AI Index 73) and Etsy (63) hold up best relative to their search share without yet reaching parity. The brands most exposed are the ones that assumed search leadership would carry forward automatically.

Every AI answer is a comparison engine that reopens the consideration set

More than one in five AI conversations (roughly 22%) carry a commercial signal — about 29% once you combine high and medium intent. The mechanism underneath that demand is what should keep brand teams up at night. In 7 of the 9 top-intent industries, 58% or more of high-intent conversations surface three or more competing brands, even when the user never asked for alternatives. Fintech answers average 6.8 brands per response, with 77.5% returning three or more. Ask a single shopping question and the AI hands back a slate.

Two switching patterns run through the data. Explicit comparison shows up where shoppers arrive in evaluation mode and the AI expands the set to four or five brands — Auto Parts answers are explicitly comparative 82% of the time, Software 63%. The harder pattern for brands is implicit comparison: in Telecom and Airlines, a shopper asks a plain logistics question and still gets a full competitive landscape back. They are being compared in conversations that do not look anything like comparison shopping. Food Delivery is the most switchable category in the data — thin loyalty, urgency, and a wide returned set combine into a high intent-to-switch score. The AI is not reinforcing loyalty; it restructures the consideration set on every turn.

Why AI-visibility tools cannot see any of this

AEO and GEO tools answer one question: is your brand mentioned? They scrape AI responses and report a share-of-voice score. That tells you whether you appeared in the slate — not whether the shopper who saw it touched AI first, which of the named rivals they actually visited, or what they bought within the week. The mention is the start of the story, not the end of it. When 60% of searches end without a click, the only way to recover the rest of the journey is to observe real people, not model AI outputs.

That is the difference between visibility and behavior. The first-touch order, the AI Index, and the comparison density above all come from permissioned, observed cross-platform behavior from real consumers — not scraped pages, not synthetic personas, not a survey of stated intent. Measure Predict lets a brand team ask a direct question of that behavior and get a grounded answer with cited evidence in seconds, not weeks. Where a visibility tool stops at the mention, Predict shows the journey on both sides of it.

See your own consumer journeys in Predict — ask which brands AI recommends in your category and where those shoppers actually went next.

What this means for brand teams in 2026

Three findings reinforce each other. Discovery is shifting: AI's slice of first touch is small but climbing fast and skews high-intent. That shift reshuffles winners: search leadership no longer carries into AI recommendation share, leaving brands like Amazon and Samsung exposed and Apple insulated. And the mechanism is comparison: AI answers high-intent questions with multi-brand option sets, so any brand absent from the slate is invisible at the exact moment of decision. The strategic move is not to chase a higher mention score in isolation — it is to learn what real shoppers do after the recommendation, and to win the brands of the journey rather than the brands of the response.

These figures came straight from Predict's behavioral data engine, grounded in cited evidence. Discovery-trend figures cover June 2024 to June 2025; brand-visibility and category-intent figures cover January to October 2025. You can run the same questions for your own category in minutes.

Try Predict Now and see the real consumer journey behind every AI recommendation in your market.