Research brief · AI shopping discoverability
How to get products found in ChatGPT and AI shopping agents
AI shopping is becoming a real acquisition channel, but the useful optimization target is not “mentions by an LLM.” It is whether a high-intent shopper or buying agent can resolve the right product, understand the differences, trust the merchant-owned facts and move cleanly toward purchase.
Published 11 September 2026 · sources listed below
+197%
Shopify AI-referred sessions, Q2 2026 year over year
~2×
conversion in spec-led categories versus organic referral, Shopify Q2 data
2×
conversion when AI used structured Shopify Catalog data versus scraped/third-party feeds
Why this matters now
AI referrals are still smaller than organic search, but they are increasingly high-intent
Shopify's Q2 2026 commerce analysis found AI-referred sessions to Shopify storefronts grew 197% year over year while organic search sessions grew 12% from a much larger base. When AI-referred shoppers landed on product pages, Shopify reported roughly 80% higher conversion than organic referral overall, with about 2× conversion in research-heavy, specification-led categories.
The most useful signal is data quality. Shopify reported that sessions where AI used structured Shopify Catalog product data converted at about twice the rate of AI sessions relying on scraped or third-party feeds. That does not prove the same uplift for every merchant or platform, but it does support a practical acquisition hypothesis: clean first-party product facts can improve the quality of AI-mediated discovery.
PayU's Agent HQ launch at Global Fintech Fest on 10 September adds another market signal. PayU is explicitly offering Indian SMB merchants tooling to make products, services and businesses discoverable and transactable on AI platforms including ChatGPT and Claude. That is infrastructure being sold against a concrete merchant problem, not just an SEO thought experiment.
Merchant checklist
Eight things an AI shopping system should not have to infer
Canonical product identity
Use one stable product URL and stable variant identifiers. Avoid making an AI system reconcile several contradictory names for the same SKU.
Structured product facts
Expose title, variant, specifications, compatibility, price, currency, availability and image data in structured markup or a first-party catalog feed.
Decision-useful detail
Answer the questions a high-intent shopper asks before buying: who it is for, exact differences between variants, constraints, delivery timing, returns and warranty.
Authoritative merchant source
Keep merchant-owned pages and feeds fresher than scraped aggregators. Update changed prices, stock and product attributes promptly.
Machine-readable policies
Make shipping, returns, geographic restrictions, subscriptions and other purchase conditions explicit enough that an agent does not have to guess.
Channel-specific integrations
Use supported catalog, merchant or agentic-commerce integrations when they exist. Do not assume a robots file or llms.txt alone makes a product purchasable.
Trust and provenance
Keep reviews, certifications, manufacturer details and source claims attributable. Avoid synthetic reviews, copied claims and unsupported superlatives.
Measure qualified referrals
Track AI-referred landing pages, conversion, new-customer rate, repeat visits and assisted sales. Do not optimize for referral counts that never progress toward a purchase.
What to publish
Human product pages and machine-readable commerce data should agree
A strong canonical product page should expose complete human-readable facts and appropriate structured data such as Product and Offer. A catalog feed can provide normalized product identity, variants, inventory and pricing. A commerce protocol or merchant integration can add transactional capability. These layers are complementary when they describe the same current product.
Do not create thin “AI SEO” copies of every product page. Duplicated pages split authority and create another place for price, stock or specifications to drift. Strengthen the source of truth and expose structured derivatives from it.
For API and machine-service sellers, the equivalent pattern is OpenAPI, MCP, agent metadata and truthful payment/discovery contracts. zFinia's separate agent-ready API guide covers that technical service case.
Measurement
Measure qualified AI acquisition, not just the referral label
Segment known AI referrers where analytics exposes them, then compare landing pages, product-page conversion, average order value, new-customer rate and repeat behavior. Keep assisted conversion in mind: some AI-mediated discovery will not preserve an obvious AI referrer through checkout.
Shopify notes that Google AI Overviews can be classified as organic in standard analytics, so “AI referral traffic” is an incomplete denominator. Treat attribution as evidence with gaps, not as a perfect market-size measure.