How to prepare product feeds for AI search in ChatGPT, Gemini and Perplexity
Fashion catalogues surface more accurately in generative AI shopping agents when merchants enrich Google Shopping feeds and implement complete Product schema markup with live pricing and stock.
KEY TAKEAWAYS Summary by the editors
- Shopping queries on generative AI platforms grew 4,700 per cent between 2024 and 2025, according to the State of Fashion 2026 report.
- ChatGPT returned structured product cards on 87 per cent of shopping prompts and Google AI Mode on 91 per cent, whilst Perplexity and Gemini answered in prose.
- Complete Product schema in JSON-LD requires name, image, price, priceCurrency, availability and GTIN or MPN for cross-reference matching.
- For apparel, Google Shopping feeds require colour, size, gender and age group attributes or products face disapproval.
Fashion merchants can increase visibility in generative AI search by optimising Google Shopping feeds and implementing complete schema.org Product markup. Shopping-related searches on generative AI platforms grew 4,700 per cent between 2024 and 2025, and Google's AI Shopping features now pull from Merchant Center feeds, making product data quality the primary lever for AI agent recommendations.
Which AI shopping agents read product feeds?
ChatGPT returned a structured product card on 87 per cent of shopping prompts and Google AI Mode on 91 per cent, making them the two primary surfaces for commerce in 2026. Perplexity showed a card on 1 per cent and Gemini on 0 per cent, instead answering with sourced citations. Google Shopping Feed, schema.org Product markup and an OpenAI Shopping Feed form the baseline and simultaneously reach ChatGPT, Gemini, Perplexity and Copilot.
During the 2025 holiday season, traffic to retail websites from AI sources like ChatGPT, Gemini and Perplexity increased by nearly 700 per cent year over year. AI-referred traffic to US retail sites grew roughly 393 per cent year over year in the first quarter of 2026, according to Adobe Analytics.
What does Product schema markup require?
Schema.org Product markup is the standard that all major AI systems and search engines understand. The markup should be written in JSON-LD format and embedded in the page HTML. Google's Product schema requires name, image, offers, price and priceCurrency.
Essential fields for AI visibility include:
- name: the product name exactly as it appears to shoppers
- image: a URL to a product photograph, minimum 250 by 250 pixels for apparel
- brand: one of the three unique product identifiers
- sku: internal stock-keeping unit
- gtin or mpn: add GTIN or MPN to every product for cross-reference matching
- offers: a nested Offer object containing price, priceCurrency, availability and URL
- availability: valid values include https://schema.org/InStock, https://schema.org/OutOfStock, https://schema.org/PreOrder, https://schema.org/LimitedAvailability and https://schema.org/Discontinued
- AggregateRating and Review: include AggregateRating and Review markup with real review data only
MerchantReturnPolicy is required by Google for Merchant Listings since 2025. Google writes that no special markup is needed for its generative AI features, whilst Microsoft says schema helps its systems, Copilot included, understand content.
How do fashion feeds differ from other categories?
Google Shopping requires colour, size, gender and age_group attributes for apparel and fashion products. Missing any of these causes immediate product disapproval. The colour attribute must match the colour name visible on the landing page. Use up to 3 colours separated by slash for multi-coloured items.
Apparel products sold in multiple sizes and colours require item_group_id to link variants of the same style. The connection happens in two places at once: in the feed through item_group_id and on the site through ProductGroup structured data. A shirt in 6 sizes and 4 colours creates 24 unique SKUs for one style, and each variant needs its own feed row with size and colour specified.
Google's apparel guidelines recommend Brand plus Product Type plus Important Attributes, with the key details in the first 70 characters. A worked example: "Nike Women's Running Shoes Black Size 8 Mesh".
Which feed attributes matter most for AI agents?
AI agents query structured product data, and if attributes are incomplete, descriptions are vague or compatibility details are missing, the AI recommends a competitor with better data. The highest-impact attributes for AI Overview carousels are product_highlight, product_detail and description length and structure.
| Attribute | Why it matters for AI | Format |
|---|---|---|
| product_highlight | Scannable feature bullets that answer common questions | 4 to 6 short phrases |
| product_detail | Structured name and value pairs for specifications | e.g. "Material: Merino wool" |
| description | Natural-language answers to conversational queries | Detailed paragraph with attributes in context |
| gtin | Cross-reference matching across platforms | 13-digit Global Trade Item Number |
| material | Primary fabric or composition for apparel | Single value, not size or colour |
| colour | Must match landing page exactly | Up to 3 colours with slash separator |
product_detail uses structured name and value pairs for measurable specifications like battery life, dimensions and materials, which powers comparison queries in AI Mode. AI agents answer natural-language questions like show me lightweight running shoes for wide feet under 150 pounds, so descriptions need to contain those attributes in natural language, not just structured fields.
How can merchants keep feeds accurate?
AI agents verify data accuracy before recommending, and a price mismatch between feed and site flags data as unreliable, causing the agent to move to a competitor. Real-time inventory sync is critical. Connect schema to a live Merchant Center feed so Google can cross-validate structured data against product inventory in real time.
Merchants can test visibility by querying AI assistants directly. Ask AI assistants about your product category and type target queries into ChatGPT, Perplexity and Gemini. For Perplexity, try product category queries; for Gemini, search on Google with questions like best product for use case and look at the AI Overview and Gemini Shopping results.
Validation tools include:
- Google Rich Results Test for rich-result eligibility
- Schema Markup Validator for general schema.org validity
- Google Merchant Center feed diagnostics for attribute completeness
- Manual queries on ChatGPT, Perplexity and Gemini for actual visibility
What are the most common implementation mistakes?
Declaring a price, rating or stock that does not match what the user sees on the page leads to Google penalising this gap harshly, up to removing all rich results. A Product entry without price, stock status or a valid image reference often gets ignored entirely by AI parsers, not partially used.
Other frequent errors include:
- Using plain text like "InStock" instead of the full schema.org URI https://schema.org/InStock
- Leaving brand, GTIN or MPN fields empty on branded products
- Mixing schema types, such as marking up a blog post as a Product
- Adding review markup for scraped third-party reviews rather than genuine customer reviews
- Forgetting to update priceValidUntil dates, which signals expired data
- Missing colour, size, gender or age_group on apparel, causing immediate disapproval
AI-generated content that goes straight into feeds can trigger Merchant Center policy flags like misrepresentation or website-needs-improvement. Merchants using AI to draft descriptions should review every line before upload.
Frequently asked questions
Do I need separate feeds for ChatGPT and Google?
A Google Shopping feed enriched with complete attributes reaches both platforms. ChatGPT reads public product pages with schema.org markup and can access Google-compatible feeds through approved partners, whilst Google AI Mode pulls directly from Merchant Center. One well-structured feed serves both.
How often should product feeds update?
AI agents verify pricing and stock in real time before recommending products. A price mismatch between the feed and landing page flags data as unreliable. Sync feeds whenever price or availability changes, and use automated scheduled uploads at least daily for high-turnover catalogues.
What happens if colour or size is missing on apparel?
Google disapproves apparel products immediately if colour, size, gender or age_group attributes are missing. Disapproved products do not appear in Shopping ads, free listings, Performance Max or AI Mode recommendations. Every variant needs its own feed row with these fields populated.