It's easy to assume "structured data for AI" is one project. It isn't — it's at least three separate systems, each consumed differently, and each worth a different amount of your team's time depending on which AI shopping surfaces actually matter to your business.

The three systems, and how they differ

1. schema.org Product markup (JSON-LD on your own pages). This is the same structured-data vocabulary that's powered Google's traditional rich results (star ratings, price, availability shown directly in search) for over a decade. It's embedded in your page's HTML and read by any crawler that chooses to parse it — Googlebot, but also, increasingly, the crawlers behind AI systems that build their own web indexes (Perplexity's crawler, OpenAI's search crawler, and others). It's free, it's a W3C-adjacent open standard, and it requires no application or approval from anyone — you publish it, crawlers either find and parse it or they don't.

2. Platform ad/shopping feeds (Google Merchant Center, and equivalents). This is a separate, direct submission channel — an XML, TSV, or content-API feed you actively push to a platform's own system, not something a crawler discovers by visiting your page. Google Merchant Center has existed for years to power Shopping ads and free product listings; as of 2026, Google has added feed attributes specifically for its AI Mode/Gemini shopping surfaces, distinct from the older ad-focused fields. According to reporting on Google's own feed specification, the AI-Mode-specific attributes include: Product Highlight (short benefit-focused phrases, 1-150 characters), Product Detail (structured spec tables), Variant Option (explicit, machine-readable attribute/value pairs for non-obvious variants), Item Group Title (a shared parent label across variants), Related Product (up to 30 linked items across six relationship types, powering "what else do I need" answers), Question and Answer (up to 30 Q&A pairs explicitly built for conversational surfaces), Document Link (linked PDFs — manuals, spec sheets — an AI system can pull detail from), and Popularity Rank (a 0-100 sales-signal field). None of these were required for traditional Shopping ads; they exist specifically because a conversational, comparison-oriented surface needs different information than a static ad unit does.

3. AI-assistant-specific product feeds (ChatGPT's direct feed access, Perplexity's Merchant Program, Shopify Catalog). These are yet another submission channel, separate from Google Merchant Center, built to feed a specific assistant's product index directly rather than being crawled or repurposed from an ad feed. OpenAI's help documentation describes ranking as drawing on "structured metadata from first-party and third-party providers," and offers merchants a direct-feed application path distinct from Merchant Center; Perplexity's Merchant Program is its own free enrollment with its own dashboard. A Shopify merchant may get partial coverage automatically (Shopify Catalog syncs to both ChatGPT and, per Perplexity's own materials, Perplexity), but a feed submitted to Google Merchant Center does not automatically reach either of these — they're genuinely separate systems requiring separate setup.

Why this differs from traditional on-page SEO markup

Traditional on-page SEO treats schema.org markup as a hint to a crawler that already has your page's full text and visual layout to work with — the markup mainly earns you a richer display (stars, price) in results you'd likely have earned anyway based on your page content and authority. A merchant feed, by contrast, is often the only representation of your product a system has, especially for a surface like Google's Shopping tab or an AI assistant's product carousel that doesn't render your actual page — if a required or AI-relevant attribute is missing from the feed, no amount of on-page schema.org markup compensates, because the feed and the page are different data paths into different destinations. Practically, this means catalog teams need to audit both: on-page schema.org completeness (usually a template-level fix, done once per page type) and feed attribute completeness (an ongoing per-SKU content task, closer to A+ content or a listing description than a one-time technical setup).

A practical setup order

  1. Confirm baseline schema.org Product markup exists and validates (price, availability, aggregateRating, brand at minimum) across your product page templates — see Setting Up Google Merchant Center for the account-level setup this pairs with.
  2. Audit your existing Merchant Center feed against the newer AI-Mode-specific attributes (Product Highlight, Q&A, Related Product, Document Link, Popularity Rank) — these are commonly missing entirely from feeds built years ago for Shopping ads alone, since they weren't required for that purpose.
  3. Evaluate which AI-assistant-specific programs are worth separate enrollment — Shopify Catalog if you're on Shopify (largely automatic), ChatGPT's direct feed application and the Perplexity Merchant Program if a meaningful share of your traffic or category relevance suggests it. Don't assume Merchant Center coverage extends to these; check each platform's own enrollment status for your store.
  4. Re-audit quarterly. This is explicitly a moving target — Google has already added attributes to its feed spec specifically for AI surfaces once in 2026, and there's no reason to assume that's the last change.

Common mistakes

  • Assuming a complete Google Merchant Center feed automatically reaches ChatGPT Shopping or Perplexity — it doesn't; each requires its own submission or integration.
  • Treating on-page schema.org markup and merchant feed data as duplicates of the same effort, when they're consumed through entirely different paths and often need different content (a feed's Product Highlight field, for instance, has no equivalent in standard schema.org Product markup).
  • Leaving new AI-Mode-specific Merchant Center attributes (Product Highlight, Q&A, Related Product) blank because they weren't required fields for Shopping ads, the feed's original purpose.
  • Filling structured-data fields with keyword-stuffed or exaggerated content, which risks disapproval under a platform's content policies and works against a system trying to synthesize an accurate, trustworthy answer.

Best practices

  • Maintain schema.org Product markup at the page-template level, validated at least quarterly, independent of any ad feed work.
  • Audit your product feed specifically against each platform's current attribute list rather than assuming an older, ads-only feed configuration is complete.
  • Treat Q&A and Product Highlight feed fields as content-writing tasks, not technical fields to leave blank or auto-generate carelessly.
  • Track which of your AI-facing enrollments (Shopify Catalog, ChatGPT direct feed, Perplexity Merchant Program) are actually active, since it's easy to assume coverage that was never actually set up.

FAQ

Do I need schema.org markup if I already have a Google Merchant Center feed? Yes — they serve different purposes. Markup helps any crawler (including AI-system crawlers building their own web index) understand your actual page; the feed is a direct data submission a platform uses instead of, not in addition to, reading your page for surfaces like Shopping ads or an AI shopping carousel.

Is there one universal feed format that works across Google, ChatGPT, and Perplexity? Not currently, based on each platform's own documentation — Google Merchant Center, OpenAI's direct-feed program, and Perplexity's Merchant Program are separate systems with their own submission paths, even though the underlying product data overlaps substantially. Some third-party feed-management tools claim to distribute one base feed to multiple destinations; verify what they actually support before relying on that claim.

How much of this is required versus optional? Baseline schema.org markup and a working Merchant Center feed are close to table stakes for any serious online seller at this point. The newer AI-Mode-specific Merchant Center attributes and the assistant-specific merchant programs (ChatGPT, Perplexity) are optional and additive — worth doing if AI-assistant-driven discovery is plausible for your category, lower priority if it isn't yet.