Start with the product record
A shopper needs to know whether a product fits their needs, what it costs and whether it can be delivered. Start with a clear product identity, the correct variant, useful specifications, current price and availability, and accurate delivery and returns information.
For electronics, that might mean compatible devices and connectors. For pet products, species and sizing. For food and drink, ingredients and pack size. Furniture dimensions, clothing fit and beauty shades answer different questions. Use verified product facts; leave unknown information for review rather than guessing.
Keep those facts consistent across the product page, its structured data and any feeds you submit. Google accepts product information through page markup, Merchant Center feeds or both; requirements depend on the experience. An accepted format does not guarantee a product will appear. Google's product-data guidance.
Give each file or protocol a clear job
robots.txt gives participating crawlers instructions about which URLs they may crawl. It is not a security barrier for private information. Google's robots.txt guidance.
llms.txt is a proposed format for a concise guide to a site's content, with links to useful detail. Treat it as supporting documentation, not proof that an assistant has found or used your products. The llms.txt proposal.
agents.txt appears in different proposals for describing agent capabilities or policies. Check which format the intended service supports before relying on it. A published declaration is not proof that a service follows it. Capability-discovery draft, Earlier policy draft.
MCP connects AI applications to external data and tools. Its usefulness depends on what the connected service actually provides. For OptAEO's supported connection and requirements, see the MCP setup guide. MCP documentation.
UCP and ACP support commerce integrations. UCP covers capabilities across product discovery, shopping and checkout; ACP specifies how agents and merchants handle supported purchases. Both require working integrations. Adding a discovery file does not create a checkout connection. UCP documentation, ACP documentation.
For Google AI Overviews and AI Mode, no special AI text file or special schema is required. Google's existing search requirements still apply, and eligibility does not guarantee inclusion. Google's AI guidance.
Check the route, then test the result
Start with representative products and variants. Compare the visible page with its markup and submitted feed: do identity, price, stock and delivery terms agree? Check that intended public pages are accessible to the relevant crawler. Use the product-data checklist to review the underlying record.
For an integration, confirm the provider's current requirements and whether your store has actually connected. Where checkout is supported, test the purchase flow separately. A file being present or a readiness check passing is not evidence that the whole journey works.
Then test relevant shopping questions and record the service, date, market, product mention and source link. Keep this evidence separate from your readiness score. The AI shopping agents guide explains how to interpret those checks.
OptAEO helps you scan product data and review gaps and proposed fixes. Applying fixes requires a paid plan. Its readiness checks do not activate an external checkout integration or guarantee an AI recommendation.