How ChatGPT Recommends Products — AI Shopping Visibility
Learn how ChatGPT discovers products via Bing, merchant feeds, and web crawl. What data signals matter, why products fail to appear, and how structured product data gets you visibility.
How ChatGPT Discovers and Recommends Products
This page is for technical SEO leads, ecommerce engineers, and product data owners who need a precise understanding of how ChatGPT discovers and recommends products, and how to make a catalog reliably visible.
Primary web layer: Browse with Bing. ChatGPT leverages Microsoft's index to discover and cite sources, including retail and product pages. Shopping data: Merchant feeds flow into Bing via Microsoft Merchant Center (product feeds, pricing, availability). Direct browsing: For user-initiated browsing, ChatGPT-User fetches pages; robots directives for OAI-SearchBot and GPTBot manage search and training access. Blocked sources: Some marketplaces (e.g., Amazon) restrict OpenAI bots in robots.txt.
If your products are not represented in Bing's index or merchant feeds, or if robots policies block relevant agents, ChatGPT visibility suffers. Structured, verifiable product truth, especially price and availability, improves the chance of being cited and recommended.
Brand systems through eLLMo to AI surfaces and checkout.
Data Signals ChatGPT Relies On
Structured identity
name, brand, sku, mpn, gtin for canonical matching and de-duplication.
Live offers
offers.price, priceCurrency, availability, shippingDetails. Reliable recommendations need current price and stock.
Specs and attributes
model, size, color, material, ingredients, compatibility for matching natural-language queries.
Trust and provenance
Clear publisher, policy URLs, authoritativeness for confidence in inclusion and citation.
Source diversity
Multiple consistent sources (site plus feeds) reduce ambiguity and improve ranking confidence.
Schema quality
Product, Offer, FAQPage, BreadcrumbList for machine-readable assertions and FAQs.
Recency and freshness
Lastmod in sitemaps, feed refresh cadence to avoid stale recommendations.
Structured Product Data: What ChatGPT Needs
Required: Product identity (name, sku, gtin/mpn, brand); Offers (price, priceCurrency, availability, url); Primary image, canonical url. Recommended: AggregateRating/Review, shippingDetails, itemCondition, size/color/material; Variant modeling (sameAs/isVariantOf); FAQPage linked from PDP.
Keep strict consistency between PDP, schema.org JSON-LD, and merchant feed fields. Verify current OpenAI crawler behaviors in the official docs.
Common Reasons Products Fail to Appear
Your roadmap to AI-first commerce
Robots or paywalls block critical agents
Symptom: PDPs not indexed or not fetchable. Fix: Allow-list bingbot and OAI-SearchBot; expose essential content without script-gated paywalls.
Thin or inconsistent product truth
Symptom: Price/availability mismatch across PDP, schema, and feed. Fix: Single source of truth feeding all surfaces in lockstep.
Missing identifiers and attributes
Symptom: Low match confidence for long-tail queries. Fix: Add sku, mpn, and gtin where available; normalize attributes.
Poor performance or reachability
Symptom: Crawl timeouts; missing mobile versions. Fix: Optimize TTFB/LCP; ensure 2xx on canonical PDP; keep sitemaps fresh.
Broken canonical or duplicate variants
Symptom: Diluted signals across near-duplicate PDPs. Fix: Canonicalize variant strategy; use isVariantOf in schema.
Outdated feeds
Symptom: Stale prices/stock in Bing Shopping. Fix: Establish frequent feed refresh via Microsoft Merchant Center.
How eLLMo Helps
eLLMo normalizes and verifies your catalog into a structured, agent-ready truth layer (Product Intelligence) with two-tier extraction and confidence scores. Product Catalog keeps inventory current with real-time sync and audit trails.
Your catalog becomes queryable across UCP, ACP, MCP, and A2A and distributes to ChatGPT, Google AI, and Perplexity without replatforming. SOAV dashboards measure and improve your AI presence with prompt-level analytics and competitor benchmarking.
Implementation Workflow
Connect your sitemap and/or feeds
Takes about 30 minutes. eLLMo auto-discovers product URLs.
Extract, normalize, and verify product truth
Takes 1 to 2 hours. Two-tier verification with confidence scores.
Deploy via protocols
UCP, ACP, MCP, A2A to AI surfaces including ChatGPT. No replatforming.
Monitor SOAV and iterate
Track citations and share of voice; fix highest-impact pages first.
Feed Field Alignment Checklist
Your roadmap to AI-first commerce
sku/mpn/gtin in feed and JSON-LD
Keep identity fields consistent across surfaces.
offers.price and availability synced with PDP
Single source of truth for pricing.
Category taxonomy mapped (Microsoft/Google)
Align to platform taxonomies.
shippingDetails for primary regions
Machine-readable shipping info.
Return policy and warranty URLs linked
Trust signals for agents.
Images meet size/ratio guidelines
1200px+ preferred for product hero.
Titles and descriptions match PDP
No keyword stuffing; aligned copy.
Frequently Asked Questions
Does ChatGPT use Bing for product discovery?
Yes. ChatGPT's browsing and search rely on Bing's index. Merchant feeds submitted through Microsoft Merchant Center can impact surfacing.
What's the difference between GPTBot and OAI-SearchBot?
GPTBot is OpenAI's training crawler. OAI-SearchBot supports search and indexing. Robots directives can manage each independently.
If I block GPTBot, will I disappear from ChatGPT?
Blocking GPTBot restricts training use. Visibility via search/browsing can still come from Bing index and OAI-SearchBot (if allowed).
Why do my prices in ChatGPT differ from my PDPs?
Stale feeds or mismatched PDP/JSON-LD/feed fields. Use a single truth layer and refresh feeds frequently.
How do I prioritize schema fields?
Must-haves: name, brand, sku/mpn/gtin, offers.price, priceCurrency, availability, url, image. Recommended: shippingDetails, aggregateRating, return policy, variant relations.
How does eLLMo distribute to ChatGPT?
eLLMo normalizes product truth and serves it via UCP, ACP, MCP, and A2A so agents can discover, cite, and transact without you replatforming.
What are the top crawl issues to fix first?
Allow-list bingbot and OAI-SearchBot, ensure 2xx on canonical PDPs, add complete schema.org Product and FAQPage, align PDP and Merchant Center feed data.