Gen Z shoppers are starting product research in ChatGPT rather than Google, according to reporting from Retail Gazette — a behavioral shift that breaks the assumption behind most discovery funnels product teams have optimized for the past two decades.

The headline claim is simple, but its implications for product managers are not. For twenty years, discovery meant SEO: ranking on Google, winning the featured snippet, capturing high-intent keyword traffic. If a growing cohort of buyers asks ChatGPT "what's the best budget running shoe" instead of typing it into Google, the visibility surface moves from a ranked list of blue links to a single conversational answer. There is no page two. There is often no cited list the user scrolls past. One model, one response, one winner.

That changes what "being discoverable" means. In the search paradigm, ten competitors could share page one and split the traffic. In the assistant paradigm, the model typically names a small number of options — sometimes one — and everything else gets zero impressions. Product managers running growth and discovery programs should treat this as channel concentration risk: a single algorithmic gatekeeper, but one whose selection criteria are considerably less transparent than Google's published ranking signals.

There are real caveats before anyone reallocates budget. The Retail Gazette report describes a directional shift among Gen Z, not a completed migration; Google still processes billions of queries daily across all demographics, and purchase behavior varies sharply by category — high-consideration purchases behave differently from impulse buys. Conversational assistants also carry their own failure mode: users trust the answer more because it reads as tailored advice, but models can hallucinate product specifications, outdated pricing, or discontinued items. A Gen Z shopper who gets a confident, wrong recommendation from an assistant has no SERP of competing results to correct it.

For PMs, three practical questions follow.

First, can the model find accurate information about your product? If your specs, pricing, and reviews live in JavaScript-heavy pages the crawler can't render, or in walled-garden content, the assistant may be describing your product badly — or not at all. An audit of what major assistants actually say about your category is now a quarterly task, not a curiosity.

Second, are you measuring the right channel? Standard analytics tools attribute traffic to search, social, and direct. Assistant-mediated referrals often arrive as direct or dark traffic with generic prompts as the true upstream source, which means dashboards can undercount the shift even while it's happening. Treating LLM-driven discovery as its own line item — even a rough one — beats pretending the old funnel taxonomy still holds.

Third, does your positioning survive summarization? Assistant answers compress. A value proposition built on rich comparison pages, video, and mid-funnel content gets flattened into one sentence alongside competitors. Products that win in this environment tend to have one clear, statable claim — the best price, the best battery life, the strongest review signal — rather than a diffuse bundle of benefits.

The honest limitation of any framework here: nobody yet has durable data on how assistant-driven discovery converts versus search. Recommendation engines have historically driven high cart values but low repeat rates because they reward the cheapest listing, not the best product. Assistants could inherit that dynamic, or escape it. PMs should run small experiments — tracking what assistants say, correcting spec errors where possible, testing concise positioning — rather than betting the roadmap on the shift.

The direction, though, is clear enough to act on. Discovery is fragmenting from ten blue links into a handful of conversational answers, and the product teams that learn to measure and influence those answers early will hold a structural advantage as Gen Z's habits harden into the default buying behavior of the next decade.