ChatGPT shopping is now product discovery, not just search

A growing share of shopping now starts with a question to an AI, and the assistant answers by naming specific products and businesses. ChatGPT, Perplexity and Google's AI Mode have all turned product discovery into a built-in feature over the past year. Here is the part that matters for a small business owner. When the assistant recommends someone, it is reading from feeds, profiles and reviews. It is not browsing your website. No data, no mention.

What changed

Three big assistants made shopping a core feature, and they all answer the same way. From structured data and trusted third-party sources.

• ChatGPT rolled out shopping research powered by a model tuned for product queries, with sourced details and side-by-side comparisons. OpenAI launched in-chat Instant Checkout in September 2025, then in March 2026 shifted commerce toward merchant-run apps and the retailer's own flow rather than owning every purchase. The discovery stayed. The buy step moved closer to you.
• Perplexity built a shopping flow with a free merchant program that lets retailers share their catalog, specs and reviews, plus in-chat purchase through PayPal. It leans heavily on reviews when it answers.
• Google AI Mode runs on Gemini plus Google's Shopping Graph, which Google says holds more than 50 billion product listings. In January 2026 Google added new Merchant Center attributes built for conversational shopping queries.

The common thread runs through all three. They do not form an opinion about your homepage. They build an answer out of feeds, business profiles, reviews and mentions, then put a name in front of the buyer.

Where the recommendations come from

ChatGPTproduct-tuned research, buy step moving to merchants

Perplexityfree merchant program, references reviews heavily

Google AI ModeGemini plus a 50B-listing Shopping Graph

This is a real shift. Two years ago a shopper typed a keyword, scanned ten blue links, and clicked around. Now the assistant reads the underlying data and hands back a short list with reasons. Fewer clicks. Higher intent. A much narrower set of names ever reach the buyer.

Why it matters

The traffic is still small, but it behaves differently from anything else hitting your site.

Adobe's numbers tell the story. AI traffic to US retail sites rose 393% year over year in the first quarter of 2026, and Adobe reports that traffic now converts about 42% better than non-AI visitors. By the time an assistant sends someone your way, they are half-sold. The AI already filtered, compared and recommended. The visitor is closer to a decision than a cold search click ever is.

Most owners cannot see this yet, because their analytics do not tag AI referrals as their own channel. That makes it easy to ignore. Do not. The cheap time to win these answers is while the channel is still quiet.

It comes down to two signals doing most of the work.

• Your data feeds. Sell products and you are not in Google Merchant Center, or not in Perplexity's merchant program, and you are mostly invisible to their shopping answers. The feed is the on-ramp. Reporting on ChatGPT shopping suggests it draws a large share of product data from Google Shopping, so a clean Merchant Center feed quietly feeds more than one assistant at once.
• Your reviews and citations. Perplexity references reviews in close to every product answer, and every assistant leans on third-party sources it trusts. Google reviews and local press for local businesses. Comparison pages and community threads for software. A shop with fifteen stale reviews loses to one with two hundred fresh ones, every time.

It is the same lesson from our briefing on AI Overviews for small businesses. Recognition is the currency. The model quotes what it has seen structured, reviewed and mentioned. If the open web barely mentions you, there is little for the assistant to repeat. Our piece on how AI decides which business to recommend walks through the same plumbing in more detail.

Your move

You do not need to chase every platform. You need to be clean, fed and reviewed. Five steps, in order.

To see which third-party pages and reviews already rank for your category, a research tool like Semrush maps them and spots the gaps. The point is not the tool. The point is to be present in the exact data these assistants read from, then catch the visit when it lands.

Worth watching

Two questions will decide how big this gets.

Where the checkout settles. OpenAI tried owning the transaction, then pulled it closer to merchants. Google and Perplexity are still pushing in-chat purchase. If buying inside the assistant becomes normal, the assistant is the storefront and your product feed outranks your website. If it stays on your site, your job is to be the recommendation that earns the click and converts it. A page with no clear next step loses the ready buyer an assistant sends. The same scarce-but-valuable-traffic logic runs through our Meta Advantage+ briefing, and through how customers now find you by name in our brand search piece.

Measurement. Most owners still cannot see AI referrals in their analytics, which keeps the channel easy to dismiss. As tracking catches up, expect more competition for these answers. The same follow-up discipline matters once those leads arrive, which is why our email deliverability guide pairs well with any feed work you do.