If my store has good SEO, is it agent-ready?

Partly. Good SEO usually means sitemaps, canonical URLs, server-rendered pages and Product JSON-LD, which cover much of the product data and discovery categories. It says nothing about whether AI bots are blocked by your WAF, whether browsing agents can use your add-to-cart button, or whether agents can buy through a protocol like UCP.

Does Google use llms.txt or Markdown pages for AI search?

No. Google says Google Search doesn't use AI text files or Markdown versions and that you don't need to create them. They're aimed at other agents, and the scorer gives them Low weight.

What is GEO in ecommerce?

GEO (generative engine optimization) means making content easy for AI answer engines like ChatGPT, Perplexity and Google AI Mode to find, read and cite. For a store, that means allowing AI search bots, server-rendering product facts and publishing accurate structured data.

Is structured data still worth adding for AI agents?

Yes. Google says structured data isn't required for its AI features, but OpenAI says ChatGPT shopping uses structured metadata, and AI agents that don't run JavaScript can only read JSON-LD that is in the server HTML.

Agent readiness vs SEO: what's different for ecommerce

How is agent readiness different from SEO?

Agent readiness and SEO answer different questions. SEO asks whether a search engine can crawl, index and rank your pages. Agent readiness asks whether an AI agent can reach your store, read product facts, operate the page and, ideally, buy through an API on a shopper’s behalf. For ecommerce stores, the two share a foundation of crawlable, server-rendered pages with good structured data. Agent readiness then adds bot policy, page operability and commerce protocols that search engines don’t care about.

What do agent readiness and SEO have in common?

Ecommerce SEO and agent readiness overlap on four technical foundations. If your SEO is in good shape, you already pass much of the product data and discovery categories.

Sitemaps. Both search engines and AI agents use XML sitemaps to list your catalog. The sitemap check also confirms that product URLs in it return 200.
Structured data. Product and Offer JSON-LD powers merchant listings in Google. OpenAI says ChatGPT shopping uses “structured metadata from first-party and third-party providers” (OpenAI). See Product JSON-LD, variants and return and shipping data.
Server-side rendering. Googlebot renders JavaScript, but a Vercel log study found that GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot did not (Vercel). SSR matters more for agents than for Google. See product facts in server HTML.
Canonicals and status codes. Canonical URLs and real 404s instead of soft 404s stop both search engines and agents from treating dead products as live. See canonical and status codes.

What does agent readiness add on top of SEO?

Agent readiness adds requirements that an SEO audit doesn’t check, because search engines don’t shop.

Robots and WAF policy for AI bots

SEO usually only checks Googlebot and Bingbot. Agent readiness checks more than 20 AI user agents and treats them differently: blocking training crawlers is fine, blocking AI search bots and user-initiated agents is not. WAFs and CDNs also block AI user agents even when robots.txt allows them. See AI crawler and agent user agents, robots.txt and agent access.

Operability for browsing agents

Browsing agents click your buttons and fill your forms. Google’s Lighthouse notes that “Agents rely on the accessibility tree as their primary data model” (Chrome). SEO doesn’t care whether add to cart is a <button> or a <div>, but an agent does. See add to cart, variant controls, accessible names and no CAPTCHA.

Commerce protocols (UCP and MCP)

The Universal Commerce Protocol lets a store publish its capabilities at /.well-known/ucp, usually with an MCP endpoint agents can call to search, build carts and start checkout. Search engines don’t use these. See UCP profile, live agent endpoint and agentic commerce protocols.

Markdown pages and llms.txt

Markdown for Agents serves a Markdown version of a page for Accept: text/markdown, and llms.txt is a curated Markdown map of a site. Both reduce tokens for agents. Google says Google Search doesn’t use them (Google), so they’re Low weight here. See Markdown alternative and llms.txt.

Checkout permalinks

A checkout permalink is a URL that builds a cart and redirects to checkout, for example /buy/{variant_id}:{qty} in the UCP format. An agent can hand it to the shopper to finish payment. No search engine needs this. See checkout permalinks.

How do agent readiness and SEO compare side by side?

Signal	Ecommerce SEO	Agent readiness
XML sitemap with product URLs	Yes	Yes, declared in robots.txt
Product JSON-LD (price, availability)	Yes, for rich results	Yes, plus a match with the visible price
`ProductGroup` variants	Yes	Yes
Return policy and shipping structured data	Recommended	Yes
Server-side rendering	Helpful (Googlebot renders JS)	Required (most AI bots don’t run JS)
Canonical URLs, no soft 404s	Yes	Yes
robots.txt rules	Googlebot, Bingbot	20+ AI search, user and training tokens
WAF and bot-challenge behavior	Rarely checked	Core check
CAPTCHA on shopping pages	Not relevant	Hard stop for agents
Real buttons, labelled inputs, variant controls	Not ranked	Core check
GET search URL	Not relevant	Core check
Markdown pages, `llms.txt`	Not used by Google Search	Low weight
UCP profile, MCP endpoint, OAuth discovery	Not relevant	Agentic commerce category
Checkout permalinks	Not relevant	Agentic commerce category

Where does GEO fit in?

GEO, or generative engine optimization, sits between SEO and agent readiness. GEO is about getting your pages read and cited by AI answer engines such as ChatGPT search, Perplexity, Claude and Google AI Mode. For an ecommerce store, the GEO requirements are mostly agent-readiness basics:

Let AI search bots in. OAI-SearchBot, Claude-SearchBot, PerplexityBot and similar bots must reach product pages, in robots.txt and at the WAF.
Put answers in the HTML. Prices, availability, sizes and policies should be in server-rendered text and structured data, not only in JavaScript.
Keep facts consistent. If JSON-LD and the visible page disagree on price, an AI answer may quote the wrong one.
Publish policies. AI answers to “does this store offer free returns?” can only cite a returns page that agents can find. See policy links.

Google’s guidance says you don’t need new machine-readable files, AI text files or Markdown for its AI features (Google). Other assistants run their own crawlers, which is why the robots and access checks matter beyond Google.

Should I prioritize SEO or agent readiness?

Prioritize the overlap first, since it pays off for both: SSR product facts, complete Product JSON-LD, a sitemap and real 404s. Then fix the agent-only issues that cost the most: unblock AI search bots and user agents, remove CAPTCHAs from shopping pages, and make add to cart and variant selectors real controls. Protocols like UCP come last. The order and weights are in agent-ready ecommerce and the scoring methodology.

Scan your store to see where it stands, or browse the glossary for terms like UCP, MCP and soft 404.