sitemap, metadata and JSON-LD: how an agent learns who you are
Before recommending a product, an agent works out what it is, who makes it and where the docs are. Some of those answers come from markup, not text.
Why it matters
- JSON-LD with the Organization or SoftwareApplication type gives facts in machine-readable form: name, site, contacts. Without it the agent guesses from the text.
- A sitemap.xml listed in robots.txt is a map the agent uses to find the right pages without crawling everything.
- A homepage link to API docs is the shortest path for an agent that came to integrate.
What to do
- 1
JSON-LD on the homepage
Add Organization (or SoftwareApplication for a product) with name, url, description and contacts.
<script type="application/ld+json"> {"@context":"https://schema.org","@type":"Organization", "name":"Company","url":"https://example.com", "description":"What we do","contactPoint":{"@type":"ContactPoint","email":"sales@example.com"}} </script> - 2
sitemap in robots.txt
Add the sitemap URL with a Sitemap: line in robots.txt.
Sitemap: https://example.com/sitemap.xml - 3
Metadata and a docs link
The homepage needs a title, meta description, canonical and lang on <html>. Put a link to the API docs in the header or footer if you have them.
What the scanner checks
A sitemap exists
sitemap.xml is the list of pages. An agent uses it to find sections it did not reach by links.
Homepage metadata is filled in
Title, description, canonical URL and language. They tell an agent what the site is.
Company or product markup is present
JSON-LD is company data in a machine format: name, site URL, product type.
The homepage links to API docs
When asked to set up an integration, an agent looks for docs starting from the homepage.
Page preview data is filled in
Title, description, image and page type (Open Graph). An assistant uses them when it shows a link to you.
Company markup has a description and contacts
The description, logo and contacts in the markup let an assistant confirm this is a real company before recommending it.
Markup links to official profiles
sameAs links to GitHub, social networks and app stores. They tell an assistant that mentions elsewhere are about you.
About, contact and policy pages are linked from the homepage
An agent checks these pages to make sure the site can be trusted.
The sitemap carries update dates
Dates tell an agent and search which pages have changed and are worth rereading.
Prices are in the markup
Price and currency as machine data. An assistant takes them from there exactly instead of reading them out of page text.
Questions and answers are marked up
An FAQ block in the markup gives ready question-and-answer pairs an assistant can quote.
Check your own site
Paste the address and get the same breakdown in half a minute.
Sources
Other guides
- An agent can open your site: no CAPTCHA, no JavaScript wall, an honest 404
- robots.txt for AI: YandexAdditional, GPTBot, ClaudeBot and Content-Signal
- llms.txt and llms-full.txt: what they are and whether you need them
- Markdown for AI agents: .md page versions and Accept: text/markdown
- API and MCP for agents: OpenAPI, MCP Server Card, OAuth discovery