There's a purchase happening right now that started in ChatGPT, not on Google. The shopper didn't open a browser tab for your category. They didn't scroll an Instagram feed or type keywords into a search bar. They asked a question, got a shortlist, and started comparing from there.
You weren't in that shortlist. Or maybe you were. The difference between those two outcomes is the subject of this piece.
The funnel used to start with a search
For twenty years, the top of the purchase funnel was a search engine. Someone wanted a standing desk, a natural deodorant, or a gift for their sister. They typed keywords. They clicked blue links. They landed on product pages, read reviews, maybe bounced to three other tabs.
Merchants understood this. They invested in ranking. They wrote for keywords. They built link profiles and optimized title tags because the click was the gateway to every sale.
That model is not obsolete, but it is no longer complete. A new entry point has opened at the very top of the funnel, and most merchants don't yet know how to show up there.
Where the first question now gets asked
In 2025, AI referrals from ChatGPT and Perplexity to ecommerce brands spiked 752% year-over-year. That number is worth sitting with. It is not a rounding error. It reflects a structural shift in where consumers start product research.
The new pattern looks like this: a shopper types a natural language question into ChatGPT, Perplexity, or Gemini. "What's a good yoga mat for hardwood floors?" or "I need a birthday gift for someone who cooks but isn't a serious chef." The AI generates a response that mentions specific products or brands. The shopper uses that response as the basis for their next move.
They may click through to validate. They may return to a search engine with more specific terms. They may head directly to a marketplace. But the framing, the shortlist, and the emotional framing of the problem have already been shaped by an AI response they never had to browse for.
According to research from Profound, 79.7% of buyers rely on AI answer engines for at least half of their decision-making process. The decision architecture is forming inside these tools, not on product pages.
This is structural, not cyclical
It would be easy to frame this as a trend, a peak of inflated expectations, something that will settle back to normal. That is not what the data suggests.
Consider what has actually changed. Google built AI Overviews into search results. OpenAI struck commerce partnerships with Target, Instacart, and DoorDash to enable purchasing inside ChatGPT. Perplexity launched a shopping experience that lets users research and buy without leaving the app. Google added agentic checkout, where the AI can execute a price-triggered purchase on your behalf.
These are not experiments. They are infrastructure decisions by the largest platforms in the world.
The three gaps most merchants don't know they have
Gap one: machine-readable product attributes. AI systems need to understand not just what a product is called, but who it is for, in what context it works best, and how it compares to alternatives.
Gap two: trust signals that exist outside your own site. AI recommendations draw heavily on third-party evidence: verified reviews, press mentions, expert roundups, comparison articles.
Gap three: the absence of context about use cases. Traditional SEO content targets keywords. AI discovery requires content that addresses problems, scenarios, and decision contexts.
