--- language: en title: "AI readiness checklist" description: "46 checks covering access, content, and operations to investigate how AI agents and systems find and read your site. Separate search fundamentals from optional protocols." canonical: "https://tropk.ai/en/resources/checklists/ai-readiness-checklist" last_modified: "2026-10-12T16:14:27.844Z" --- # AI readiness checklist Reviewed on 2026-10-06. ## Find out what an agent can access and understand AI readiness is a way to organize the investigation of access, content, and agent interactions. The first question is concrete: does the public URL deliver accurate and sufficient information to the system you intend to serve? This checklist is for content, [SEO](/en/resources/glossary/seo), and development teams. It combines search fundamentals with operational checks and emerging proposals. Each item states when it applies, so an optional protocol does not become an obligation for every site. ## What’s in the spreadsheet The 46 checks include a method and acceptance criteria. Status, evidence, and plan fields are blank. Scanner criteria are used to guide questions, with human review and no automated readiness score. - Public content, [HTML](/en/resources/glossary/html-the-web-s-markup-language), [canonical](/en/resources/glossary/canonical), [indexing](/en/resources/glossary/indexing), links, and [sitemap](/en/resources/glossary/sitemap). - Bot policies by purpose, differences between search and training, firewalls, and agent identity. - Clear answers, context, sources, authorship, dates, tables, original evidence, and accurate [structured data](/en/resources/glossary/structured-data). - Markdown when adopted, Accept negotiation, format equivalence, caching, and protection of private content. - APIs, authentication, agent cards, actions, and commerce only when there is an approved use case; measurement and retesting. ## How to start with access Define the agents and URLs of interest. Test without a session, check the status and body, and compare the initial HTML with the rendered content. Inspect robots, headers, and the firewall separately: a permission rule does not prove that access actually succeeded. For OpenAI, the documentation distinguishes [OAI-SearchBot](/en/resources/glossary/oai-searchbot) for search, [GPTBot](/en/resources/glossary/gptbot) for training, and [ChatGPT-User](/en/resources/glossary/chatgpt-user) for user-initiated requests. Read the current documentation for the agents in scope and record a policy for each purpose. Do not authenticate a bot based only on its User-Agent name. ## Markdown and protocols need a reason If you offer Markdown, test Accept header weights, parity with HTML, and [cache](/en/resources/glossary/cache) separation. Private content, drafts, and downloads requiring registration must remain protected in every format. Byte size does not automatically equal token count. [llms.txt](/en/resources/glossary/llms-txt) is an optional proposal. API catalogs, [MCP](/en/resources/glossary/model-context-protocol), [A2A](/en/resources/glossary/agent2agent), DNS-AID, [WebMCP](/en/resources/glossary/webmcp), and commerce protocols depend on the product and public service offered. Do not expose a privileged interface or add payment actions just to raise a score. ## Readiness is no guarantee of citations Google says its AI features retain SEO fundamentals: a page must be indexed and eligible for snippets to appear as a supporting link, and there is no universal additional technical requirement. Passing checks does not guarantee inclusion. Record observations of responses with the question, product, date, region, and method. Distinguish mentions, recommendations, and citations; distinguish bot access, human visits, and conversions. An observed change alone does not prove causation. ## Download and use Open the XLSX in Excel or import it into Google Sheets. Start with a template and a target agent, record evidence, and mark items that do not apply to your service as Not applicable. After making corrections, repeat the test and check the affected formats and pages. Email is the only required field to submit the request. Name and website are optional; the file can be downloaded immediately after submission. ## Baixar o checklist O material editável está disponível mediante cadastro na [this checklist page](). O e-mail é required; nome e site da empresa são opcionais. ## Sources and references - [AI features and your website]() — Google Search Central - [Creating helpful, reliable, people-first content]() — Google Search Central - [Check documentation]() — geo.new - [Agent Readiness]() — IsItAgentReady - [OpenAI Crawlers]() — OpenAI - [Markdown for Agents]() — Cloudflare - [The llms.txt proposal]() — llmstxt.org ## Continue a leitura - [Technical SEO Checklist]() - [E-E-A-T]() - [AI HTML Inspector]() - [JSON-LD Validator]() - [AI robots.txt checker]() - [AI Presence Audit]() - [AI response monitoring]() ## About the author [Vinicius Pacheco]() Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content. --- [All pages in Markdown]()