The shift
What happens when product discovery starts inside an AI chat instead of a search results page? Over the last 90 days, we tracked 15 e-commerce client accounts to measure how AI shopping agents including Copilot, Google Shopping Graph, and Perplexity are changing the path to purchase.
The pattern was clear. Discovery is moving upstream into AI-assisted interfaces, and that changes what wins visibility. It is less about matching a keyword and more about giving agents clean product data, credible reviews, and semantically rich content they can interpret with confidence.
For brands selling health products, lifestyle goods, and direct-to-consumer essentials, we are already seeing this show up in analytics, conversion rate, and order value. So, you know, this is no longer a theory piece. It is happening in live accounts now.
Data point #1
How much discovery is already coming from AI chat interfaces? Across the 15 accounts we tracked, 23% of product discovery sessions originated from AI chat interfaces rather than traditional search results. That figure doubled quarter over quarter, which is the part we think most e-commerce teams are underestimating.
Actionable takeaway: break out AI-originated discovery in your analytics now. If we lump it into generic referral or organic buckets, we miss the trend until it is much larger.
Data point #2
What makes some products more visible to agents than others? Products with schema markup including Product schema, Review schema, and FAQ schema appeared in 3.7x more AI agent recommendations than products without that markup.
That makes sense when you think about how these systems retrieve and rank information. Agents need structured signals. When pricing, availability, reviews, and product attributes are clearly marked, your catalog becomes easier to interpret and easier to trust.
Actionable takeaway: audit your top-selling SKUs first. Make sure Product, Review, and FAQ schema are implemented correctly and consistently across every key page.
Data point #3
Does agent-originated traffic actually convert better? Yes, and by a wide margin. In our 90-day sample, agent-originated traffic converted at 4.1% versus 2.3% for traditional search traffic.
Why the gap? In most cases, the customer arrived more qualified because the agent had already filtered options based on price, product fit, reviews, and intent. We have seen this especially with higher-consideration categories where shoppers want fast comparison without opening 12 tabs.
Actionable takeaway: build landing pages and product pages for answerability, not just aesthetics. The clearer the information, the more likely the agent is to send purchase-ready traffic.
Data point #4
What about order value? Here is where it gets even more interesting. The average order value from agent-referred customers was 18% higher: $97 versus $82 from traditional search users.
Bear in mind, agents often optimize for best-fit recommendations rather than cheapest-click browsing. That can lead shoppers toward bundles, premium variants, and better-matched products, which lifts basket size without forcing aggressive discounting.
Actionable takeaway: enrich product descriptions with use cases, specs, compatibility details, and bundle logic so agents can confidently recommend higher-value options.
The takeaway
Is optimizing for agentic commerce optional now? We do not think so. If product discovery is shifting into AI interfaces, then brands need to adapt the inputs those agents rely on.
Start with the basics:
- add and validate schema markup
- make product descriptions semantically rich
- sync reviews so agents can access fresh trust signals
- keep pricing, availability, and attributes accurate
We are watching this change happen in real client accounts, not just industry headlines. Are you seeing AI-driven discovery inside your own analytics yet? And if not, what would it take for your catalog to become the easiest one for an agent to recommend?
