Free AI Agent Readiness Checker
See your site the way ChatGPT, Perplexity, Claude and Gemini see it. We fetch your page without a browser and without running JavaScript, run 23 checks across crawler access, readable content and machine-readable meaning, and give you a 0–100 score with the exact fixes — including a prompt you can paste straight into a coding agent.
Check your AI agent readiness
Enter any public URL. We fetch it the way ChatGPT, Perplexity and Claude do — no JavaScript, no browser — and score what they can actually see.
Enter your website to track its AI visibility across ChatGPT, Gemini, Claude, and Perplexity — and turn chatbot mentions into traffic.
- Set up in minutes
- Cancel anytime
Why AI Agent Readiness Is Not Just SEO
Google has rendered JavaScript for years. Most AI crawlers do not. When ChatGPT opens your link to answer a question, or Perplexity indexes your page to cite it, what they get is the raw HTML your server returned — no hydration, no client-side fetches, no framework. A React app that ships an empty <div id="root"> and paints the content in the browser ranks fine in Google and reads as a blank page to an answer engine.
The same split runs through the rest of the stack. Bot protection that lets Googlebot through by allowlist will challenge ChatGPT-User, because it is not on the list. A soft 404 that returns 200 with your app shell is a minor SEO annoyance and a serious agent problem, because an agent probing for /llms.txt or /pricing concludes every URL it invents exists. This tool checks the things that differ, not the things you have already fixed for Google.
What This Tool Checks
Nine essential checks worth 70 points, seven recommended checks worth 30, and seven bonus signals worth up to 5 more.
Essential — can an agent reach and read you (70 pts)
- AI crawlers allowed in robots.txt (14 pts) — eleven AI user-agents tested individually against your rules, with answer-engine crawlers weighted above training crawlers.
- Reachable by an AI agent user-agent (10 pts) — we request your page as
ChatGPT-Userand compare it to the browser response, which is how bot walls get caught. - Content readable without JavaScript (12 pts) — text length, H1 presence and text-to-HTML ratio in the unrendered response.
- Missing pages return a real 404 (8 pts) — we request a path that cannot exist and check you do not answer 200.
- Redirects an agent can follow (6 pts) — chain length, plus meta-refresh stubs that non-JavaScript crawlers never get past.
- Headings form a usable outline (6 pts) — one H1, starting at H1, no skipped levels, because agents chunk long pages on heading boundaries.
- Title and description (6 pts) — the two strings most often carried into a citation.
- HTTPS and a trustworthy canonical (4 pts) — including a check for a leftover
noindex, which several AI crawlers honour. - Sitemap an agent can discover (4 pts) — present, parseable, and declared in robots.txt rather than guessed at.
Recommended — can an agent understand you (30 pts)
- llms.txt (6 pts) — present, and actually a markdown file rather than your catch-all route returning HTML with a 200.
- JSON-LD structured data (6 pts) — blocks that parse, and how many types they declare.
- Organization schema (5 pts) — scored on field completeness (name, url, logo, description), not merely on being present.
- Trust pages linked from the homepage (4 pts) — about, contact, pricing, privacy, terms.
- sameAs entity links (3 pts) — whether anything ties your site to a Wikipedia, Wikidata, LinkedIn, Crunchbase or GitHub entity.
- Open Graph metadata (3 pts) — what renders when an assistant hands a user your link.
- Page token budget (3 pts) — whether the page fits in one agent pass or gets truncated mid-document.
Bonus — emerging conventions (up to +5)
/llms-full.txt, markdown content negotiation, agent discovery files under/.well-known/,AGENTS.md, an RSS or Atom feed,security.txt, and<lastmod>dates in your sitemap.- These are early standards, so they can only ever add points. A site that ignores all seven can still score 100.
How the Score Works
Essential and recommended checks award partial credit — a sitemap that exists but is not declared in robots.txt earns most of its points, not none. Bonus signals are capped at 5 points and never subtract, because penalising a site for skipping a convention that is eighteen months old would say more about the convention than about the site.
Three checks behave differently. AI crawler access, agent reachability and content without JavaScript are blocking: if any of them scores zero, no AI system can read the page, and every point further down the report is theoretical. Rather than subtracting their weight and printing a comfortable number, the score is capped at 45 and the report names the check responsible. A single-page app with perfect schema markup that no crawler can read is not an 87 out of 100.
How to Use This Tool
- Enter your homepage URL — the scan fetches your page, robots.txt, llms.txt, sitemap and several well-known paths, which takes a few seconds.
- Read the score band before the number — the label tells you whether you have a blocking problem or a polish problem, and they need very different responses.
- Work the essential tier top to bottom — it is ordered by weight, and the items at the top gate everything below them.
- Expand any check for the fix — each one includes the evidence we observed and the specific change to make.
- Copy the fix prompt — it packages every failing and partial check into one brief you can paste into Cursor, Claude Code or a ticket.
- Re-run after deploying — nothing here is cached, so a re-scan reflects the change immediately.
The Fixes That Move the Score Most
- Unblock answer-engine crawlers. A blanket
Disallow: /underUser-agent: *catches every AI crawler as collateral. Name the bots you actually want out instead. - Server-render the content. If your homepage ships under 500 characters of text before hydration, that is the whole page as far as an answer engine is concerned.
- Check your bot protection. Cloudflare, Akamai and similar products block
ChatGPT-Userunder their stricter presets while browsers sail through, so the problem is invisible until you test with the right user-agent. - Return real 404s. Soft 404s teach agents that every URL they guess is real, which is how you end up cited with links that go nowhere.
- Add Organization schema with sameAs. This is what turns your site from a web page into an entity a model can recognise and name.
Frequently Asked Questions
What is AI agent readiness?
AI agent readiness is how easily an AI system — ChatGPT, Perplexity, Claude, Gemini or an autonomous agent — can fetch your site, read it without running JavaScript, and understand what your business does. It overlaps with SEO but is not the same thing: a page can rank well in Google, which renders JavaScript, and still be a blank document to an AI crawler that does not.
How is the AI agent readiness score calculated?
The score is out of 100. Essential checks are worth 70 points and cover whether a crawler can reach and read the page at all: AI crawler access in robots.txt, reachability under an AI agent user-agent, content present without JavaScript, real 404s, redirect hygiene, heading structure, title and description, HTTPS and canonical, and a discoverable sitemap. Recommended checks are worth 30 points and cover whether an agent can understand and attribute you: llms.txt, JSON-LD, Organization schema, sameAs entity links, trust pages, Open Graph tags and page token budget. Seven bonus signals can add up to 5 more points but never subtract.
Why is my score capped even though most checks passed?
Three checks are treated as blocking rather than additive: AI crawler access in robots.txt, reachability under an AI agent user-agent, and content readable without JavaScript. If any of them scores zero, no AI system can read the page, so every point further down the report is theoretical. When that happens the score is capped at 45 and the report names the check responsible. Fix the blocking check and the rest of your points become real.
Which AI crawlers does the tool check for?
Eleven: OAI-SearchBot, ChatGPT-User and GPTBot from OpenAI, ClaudeBot and Claude-User from Anthropic, PerplexityBot and Perplexity-User from Perplexity, Google-Extended, Applebot-Extended, meta-externalagent and CCBot. The six that answer live user questions — OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-User and Google-Extended — are weighted more heavily, because blocking one of those removes you from answers people are reading right now rather than from a training corpus.
Is blocking GPTBot bad for my AI visibility?
Not necessarily. GPTBot and CCBot crawl for model training, and opting out of training is a legitimate business decision that this tool does not penalise heavily. Blocking OAI-SearchBot, ChatGPT-User, PerplexityBot or Google-Extended is a different matter — those crawlers build and fetch the sources that answer engines cite, so blocking them means you cannot be recommended even to someone asking for exactly what you sell.
What is llms.txt and do I need one?
llms.txt is a markdown file at the root of your domain that points AI systems at the pages you most want read: a title, a one-line summary, then sections of links. It is a convention rather than a standard, and support is still uneven, but it costs an afternoon to write and it is the only mechanism that lets you choose what an agent reads first instead of leaving it to your navigation. It is worth 6 of the 30 recommended points here.
Enter your website to track its AI visibility across ChatGPT, Gemini, Claude, and Perplexity — and turn chatbot mentions into traffic.
- Set up in minutes
- Cancel anytime
From the blog
The 8 Best AI Visibility Tracking Tools for ChatGPT, Claude, Gemini & Perplexity in 2026
Compare the 8 best AI visibility tracking tools for multi-engine coverage across ChatGPT, Claude, Gemini, and Perplexity. Pricing, engine support, GSC integration, sentiment, and who each one fits — SMB vs enterprise.
What Gets Cited by ChatGPT, Claude, Gemini, and Perplexity (2026 Data)
ChatGPT, Claude, Gemini, and Perplexity cite wildly different sources — only 14% of top domains overlap. A data-led look at what each AI chatbot actually cites, based on 30M+ citations.
Related Tools
Check your robots.txt rules for syntax errors and accidental blocks.
AI Crawler Access CheckerSee which AI crawlers your robots.txt lets through, with the exact rule behind every verdict.
Structured Data ValidatorValidate your JSON-LD schema markup and find missing required fields.