Skip to content
Enjoy 20% off for 12 months on all plans. Use promo code: HYPERGROWTHClaim it

Check your agent readiness

A separate score out of 100 for whether AI agents can reach, read and trust your site, with a fix for each failing check.

Open in the appUpdated 2026-10-12

Overview

Agent readiness answers one question: when an AI assistant such as ChatGPT, Claude or Perplexity visits your site for someone, can it reach the pages, read them and trust what they say? Every SEO audit measures it and shows the result as Agent readiness in the audit report.

It is its own score out of 100. It never changes your SEO grade.

When to use this

Read it after every SEO audit, and act on it when people find businesses like yours by asking an AI assistant rather than searching. The failing checks are usually small fixes: a robots.txt rule, a missing page, a placeholder phone number left in a template.

How it works

While the SEO audit crawls, agent readiness requests your homepage, a page that does not exist, robots.txt, llms.txt and your sitemap the way AI agents do. It also reads up to 50 of the pages the audit crawled. Each check passes, partly passes, fails, or is not measured, for example:

  • AI agents can reach the homepage, and robots.txt lets AI search agents in (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User). Training crawlers such as GPTBot are your decision, so they are reported but never scored.
  • Homepage content loads without JavaScript, Missing pages answer 404, and Contact details are real, with no template placeholders.
  • llms.txt guides AI readers, Homepage states what you are in structured data, Prices an agent can find, and Sitemap lists your pages.
  • Homepage not hidden from search: no noindex in its robots meta tags or X-Robots-Tag header. AI assistants find most sites by searching, so a noindex left on after launch takes the site out of their answers.
  • Profile links in structured data still work: each sameAs link on the homepage business entity is opened. A page that is gone (404, 410 or no answer) fails; social networks that turn away scripted requests are not measured.
  • Google sees the page visitors see: the homepage is requested as Googlebot and compared with the visitor's copy. A hacked site often shows Google gambling or pharmacy spam, or Japanese keyword pages, while the owner sees nothing wrong. If your bot protection refuses a request claiming to be Googlebot, the check reads not measured, because the real Googlebot comes from Google's own network.

Checks come in three tiers. Essential checks share 80 points and recommended checks share 20, with partial credit in proportion; a check that does not apply leaves the pool. Each passing bonus signal adds 0.25 points, up to 5. A software product is also checked for an API description and an MCP server; a business site is not graded on those.

When the homepage was built with WordPress, Wix, Shopify, Squarespace or Webflow, each failing check also gets click-by-click steps for that builder. When llms.txt is missing or incomplete, Draft my llms.txt under that check writes one from your strategy (your services, target keywords and the pages that answer), ready to copy or download and upload.

Steps

To check your agent readiness:

  1. Run an SEO audit on the website, or open its latest one.
  2. Select Agent readiness in the report's header, or scroll to the Agent readiness section. The same score has its own tab under Deep audits, beside AI Deep-Dive.
  3. Read the open checks at the top. Passing checks fold away under Passing, and the rest under Optional and not measured.

Understand the results

The score ring shows the score and grade: A+ (97 and up), A (90), B (80), C (70), D (60), F. Beside it, Essential and Recommended show how many checks pass, and Bonus shows the points added. Each open check states what was found and What to do.

The agent test drive goes one step further: an AI agent tries four things a buyer asks (Find the price, Find how to get in touch, Understand what they offer, Decide whether to trust them) using only your site's own pages, and each task comes back Done, Partly done or Not done, with the pages it read. A site that only offers quotes, with no prices shown, reads Partly done on the price. Trust is Done only when the site names who runs the business or works there, and shows something a buyer can check, such as reviews or a licence.

Sample output

This is an example on a made-up website, not a real report:

Agent readiness 69/100 (D)
Essential 7/10 pass · Recommended 6/14 pass · Bonus +0.5

Fail     Contact details are real
         Template placeholders still on the site: 555-0100.
Partial  robots.txt lets AI search agents in
         robots.txt blocks OAI-SearchBot, PerplexityBot from the homepage.
         What to do: Remove the Disallow rules for these agents.
Partial  Missing pages answer 404
         HTTP 404, but no markdown body when an agent asks for one.

Agent test drive: 3 of 4 tasks done
Find the price ............... Not done

What to do with the results

To raise your agent readiness:

  1. Start with the failing essential checks at the top of the list.
  2. Follow each check's What to do or its builder steps, or select Copy fix prompt and paste the prompt into your AI coding tool or send it to your developer. The prompt lists every open check in priority order with its evidence.
  3. When the fix is live, select Check again on the card. It reads your site again for free in about 20 seconds.
  4. Compare: the card shows the change since the audit, for example "62 to 75", and names the checks that were fixed, got better or got worse. Run the agent test drive again to see whether a task now completes.

Limits and usage

  • Agent readiness comes with every SEO audit at no extra charge, and Check again is free.
  • The agent test drive costs 2 credits a run and takes about 15 seconds. See plans and credits.
  • Check again sends your site about 30 requests, so it is limited to a few runs a minute.
  • A check that could not be measured, because a page did not load, is listed as Not measured and never counts against you.

Troubleshooting

Every check says not measured

Your homepage did not load for the probe, usually because a firewall refused it. Allow SearchBlueprint's crawler, then select Check again.

A check still fails after the fix

The page may still be cached. Clear your site's cache or CDN, wait a few minutes, and select Check again.

Was this page helpful?

Still stuck?

Email us and the team replies from the same address. If you're signed in, Ask anything in the app answers from these guides.