How AI Shopping Agents Discover Products

Something significant has been happening in the way people discover products to buy — and it's been easy to miss if you're heads-down managing campaigns.

A growing share of product discovery is now happening through AI interfaces: people asking ChatGPT for product recommendations, using Google's AI Overviews to research purchases, or querying Perplexity for comparisons. These aren't replacing traditional shopping journeys yet — but they're becoming a meaningful part of the funnel, particularly for considered purchases.

Understanding how these AI agents actually find and surface products is becoming as important as understanding how Google Shopping worked in 2019.

The New Discovery Landscape

For most of the last decade, the main pathways for online product discovery were fairly predictable: Google Search, Google Shopping, social media ads, and word of mouth. Each had established mechanics — optimise your feed, bid well, show great creative.

In 2026, add these to the mix:

Each of these works differently — but they share a common characteristic: they're pulling from structured data sources, review signals, and web content to generate recommendations. And they're doing it without showing you the traditional ten blue links.

How These Agents Actually Find Products

Think of an AI shopping agent like a very well-read personal shopper. It's read millions of product reviews, blog posts, retailer websites, and comparison guides. When you ask for a recommendation, it draws on all of that accumulated reading to give you its best answer.

But here's the key: it can only recommend products it "knows about." And it learns about products through:

1. Structured web data (schema markup)

AI systems crawl the web and are particularly good at reading structured data — the machine-readable code (schema markup) that tells them: "this is a product, it has these attributes, it costs this much, it has these reviews." Products with rich, accurate schema markup are much more likely to be well-understood and correctly represented by AI agents.

If your product pages are missing Product schema or have incomplete attributes, AI agents may not understand your product well enough to recommend it — even if you sell exactly what someone is looking for.

2. Google Merchant Center

For Google's AI Overviews and Shopping surfaces, the most direct connection is through Google Merchant Center. Products in a clean, verified Merchant Center feed have a direct line to Google's AI product recommendation systems.

This is a well-established signal Google's AI uses heavily — making Merchant Center participation arguably more important than ever for ecommerce businesses that want AI visibility on Google's surfaces.

3. Review signals

Reviews are a major input into AI product recommendations. An AI system asked "what's the best standing desk under $500?" will draw heavily on review data — both structured (schema markup reviews) and unstructured (review articles, comparison posts, forum discussions).

Products with high review volumes, high ratings, and reviews that discuss specific use cases tend to appear more frequently in AI recommendations. This is because AI agents can match "this product has reviews that mention [use case]" to the buyer's stated need.

4. Authority and trust signals

AI agents trained to surface trustworthy recommendations pay attention to authority signals: Is this product sold by a recognised brand? Does it have a verified presence on Google (a Business Profile, a Merchant Center account, a Google Knowledge Panel)? Has it been reviewed by authoritative third-party publications?

These signals are harder to build quickly but matter significantly for competitive categories where many products exist and the AI needs to make a quality judgement.

What This Means for Your Marketing Strategy

The key insight is that the fundamentals of AI product discovery overlap significantly with good SEO and Merchant Center hygiene. If you're already doing these well, you're better positioned than you might think.

• Clean, complete Merchant Center feed — the most direct connection to Google's AI surfaces
• Product schema markup on every page — makes your products machine-readable for all AI agents, not just Google
• Rich, specific product descriptions — helps AI agents match your products to the right queries
• Strong review volume and ratings — a trust and relevance signal for AI recommendations
• Earned media and third-party coverage — reviews on external platforms, mentions in comparison articles, coverage in industry publications

What We Don't Know Yet

Honesty matters here: AI product discovery is genuinely new, and the signals that drive it are still being understood. What we know is based on documented API specifications, published research, and observed behaviour — not access to the black boxes inside these AI systems.

What practitioners and researchers at Search Engine Land and elsewhere have observed is that the overlap between "good SEO fundamentals" and "good AI visibility" is significant. Businesses that have invested in structured data, quality content, and legitimate review building are better positioned than those who haven't.

The space will evolve quickly. Watching how Google, OpenAI, and Perplexity develop their commerce features over the next 12–18 months will be important for any business that relies on product discovery.

Practical Next Steps

If you want to improve your AI product discovery visibility, start with the things that have the most established impact:

1. Verify and improve your Google Merchant Center feed — clean data, no disapprovals, complete attributes
2. Check your product schema markup using Google's Rich Results Test and fix any missing or incomplete properties
3. Review your product descriptions for specificity — can an AI answer "is this good for [use case]?" from your description alone?
4. Audit your review collection and syndication — are you actively collecting reviews and getting them into schema markup?
5. Ensure you have a verified Google Business Profile and a Merchant Center account with verified domain