Google can read your site.
ChatGPT might not.
We request your homepage as the actual AI crawlers, compare raw HTML against a full JavaScript render, and show you exactly what ChatGPT, Claude and Perplexity receive. These failures never appear in Google Search Console.
No signup. No credit card. Results are a shareable link.
Three ways sites go invisible
Each one is silent. Your analytics look normal, your Google rankings hold, and you lose the answer box anyway.
Your CDN blocks them before they reach your content
Cloudflare Bot Fight Mode and similar bot-management rules return 403 to AI crawlers even when robots.txt allows them. Every audit tool that only reads robots.txt reports you as wide open. The request never gets that far.
Your content only exists after JavaScript runs
No major AI crawler executes JavaScript — GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot and PerplexityBot all read raw HTML only. Google renders JS, so a client-side app can rank normally in Google while being close to blank in ChatGPT.
robots.txt blocks retrieval bots along with training bots
Blocking GPTBot is a reasonable choice about training data. Blocking OAI-SearchBot and PerplexityBot is different — those are the crawlers that fetch sources for live answers. Most sites that block one block both without realising.
We test the request, not the config
Most “AI visibility” tools ask a language model what it thinks of your brand and chart the answer. That is downstream guesswork.
Crawlable measures the layer underneath: whether the crawler can complete a request at all, and whether your content is in the response it gets. Every finding is a reproducible HTTP fact you can verify yourself with curl.
- Live requests sent as GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot
- WAF, CDN and bot-management interception (403s, challenge pages)
- robots.txt policy for 14 AI crawlers, retrieval bots scored separately from training bots
- Raw HTML vs fully rendered content, word for word
- Empty SPA mount points and client-only sections
- JSON-LD entity coverage and parse errors
- Sitemap, llms.txt, canonical, headings, meta
- Question-shaped content and FAQ schema
- Server response time against crawler timeouts
Diagnosis is free. Fixing and staying fixed is paid.
Scan anything you like at no cost. Subscribe when you want the generated fixes and an alert the next time a deploy quietly breaks it.
Free
One-off diagnostic, no account needed.
- Unlimited one-off homepage scans
- Full findings with evidence
- Shareable public report
Pro
For a business keeping its own site readable.
- Monitor up to 3 sites
- Weekly automatic re-scans
- Email alerts when a deploy breaks AI readability
- Generated fixes: robots.txt patch, JSON-LD, llms.txt
Agency
For agencies reporting to clients every month.
- Monitor up to 25 sites
- White-label reports with your logo and colours
- Client-ready PDF/print export
- Weekly re-scans and regression alerts
Agency Plus
High-volume agencies and platforms.
- Monitor up to 100 sites
- Everything in Agency
- API access for your own dashboards
- Daily re-scan option
Questions
Can ChatGPT read my website?
Only if three things hold. Your robots.txt must allow the retrieval crawlers (OAI-SearchBot, ChatGPT-User); your server and CDN must actually answer their requests rather than returning 403; and your content must be present in the raw HTML, because no major AI crawler executes JavaScript. Crawlable tests all three with real requests.
Do AI crawlers run JavaScript?
No. As of 2026, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot and PerplexityBot all read raw HTML only. Googlebot does render JavaScript, which is why a client-side rendered site can rank normally in Google while being close to blank in ChatGPT.
Why would my robots.txt allow a crawler but the crawler still get blocked?
Because robots.txt is a policy file and bot management is a network control. Cloudflare Bot Fight Mode, "Block AI Scrapers", and similar WAF rules reject the request at the edge before robots.txt is ever consulted. Tools that only read robots.txt report you as open while the real request is refused.
How is this different from an AI visibility tracker?
Trackers ask language models what they say about your brand and chart the answers. Crawlable measures the layer underneath: whether the crawler can complete a request and whether your content is in the response. If that fails, prompt tracking is measuring a symptom.
Is the scan really free?
Yes. Unlimited one-off scans, full findings with evidence, no account and no card. Paid plans add continuous monitoring, regression alerts, generated fixes and white-label client reports.
Check your site now
Forty seconds, no account. You will get a shareable report either way.
No signup. No credit card. Results are a shareable link.