Free AI visibility audit: can ChatGPT read your website?
Some of your buyers now ask ChatGPT, Perplexity or Gemini who to hire before they open Google. If an assistant cannot fetch and read your website, it has nothing of yours to quote. The free AI Audit checks whether ChatGPT, Claude, Perplexity, Gemini and Copilot can technically reach and understand your site, and gives you a score from 0 to 100 with a list of what to fix.
What the audit checks
You type in one address, usually your home page. The tool scores that page across five areas. Every check is marked as a pass, a warning or a fail, with a one-line explanation and a note on how to fix it.
- AI bot access (30 points) - whether your robots.txt lets in the crawlers that feed AI answers: OAI-SearchBot (ChatGPT search), PerplexityBot, Claude-SearchBot, Googlebot (Gemini and AI Overviews) and Bingbot (Copilot). It also reads the rules for training crawlers such as GPTBot, ClaudeBot and Google-Extended, and looks for a noindex tag or header. 15 bot tokens in total.
- Content readable without JavaScript (25 points) - how many words of real content sit in the raw HTML, whether the page is an empty JavaScript shell, one H1, H2 sections, the ratio of text to code and semantic main or article tags.
- Structured data and metadata (20 points) - valid JSON-LD (schema.org) and the types it declares, such as Organization or LocalBusiness, Service, FAQPage and Article, plus the title, meta description, Open Graph tags, canonical URL and page language.
- Index and fundamentals (15 points) - a working sitemap, HTTPS, server response time, the size of the HTML, and whether an llms.txt file exists. llms.txt is worth one point only, because it is an experimental format.
- Content AI wants to cite (10 points) - an FAQ or question-style headings, lists or tables, a phone number or other contact details in the text, and publication or update dates.
On top of the score, you get a verdict for each of the five assistants: sees your site, sees it partially, won't read the content, or access blocked.
What the tool fetches, and what it does not
Everything runs on a server. You do not install anything and nothing touches the admin panel of your site. For the address you enter, the audit requests:
- Your page, as a browser - the HTML (up to roughly 800 KB), with a 9-second time limit.
- Your page again, as an AI bot - a second request that identifies itself as GPTBot, to see whether your server or CDN answers it differently. This is a warning signal only and is worth no points, because the request does not come from OpenAI's own IP addresses. Your server logs give the final answer.
- /robots.txt - parsed by the rules of the robots.txt standard (RFC 9309) for each of the 15 bot tokens.
- /llms.txt - checked for presence.
- Your sitemap - the one declared in robots.txt, or the usual locations such as /sitemap.xml and /wp-sitemap.xml.
What it does not do: it does not run JavaScript, it does not crawl your whole site, it does not log in anywhere and it does not ask ChatGPT about you. It reads the same public files a crawler would and tells you what it found. If your firewall blocks even this plain request, you get a message saying so instead of a score, and that is a finding in itself.
Why AI assistants may not see a site
The causes are usually mundane, and most of them are invisible to anyone who only looks at the site in a browser:
- Content that only exists after JavaScript runs - Google renders JavaScript before indexing; most AI crawlers do not. A December 2024 analysis of crawler traffic on Vercel's network found that none of the major AI crawlers it measured, including GPTBot and ClaudeBot, rendered JavaScript, even though they downloaded JavaScript files. If your text, services or contact details appear only after a script runs, those crawlers receive an almost empty page.
- robots.txt rules that block AI bots - a Disallow line for GPTBot or "all bots" is sometimes added by a plugin or copied from a template. Note the difference: OpenAI uses GPTBot for training and OAI-SearchBot for ChatGPT search, and each is controlled separately. Anthropic separates ClaudeBot (training) from Claude-SearchBot (search), and Perplexity builds its index with PerplexityBot. Blocking a training bot is a legitimate choice. Blocking a search bot keeps you out of that assistant's live answers.
- A CDN or firewall that stops bots - since July 2025, Cloudflare has blocked known AI crawlers by default on newly added domains unless the owner allows them. Security plugins and hosting firewalls can do the same without anyone noticing.
- No structured data - without schema.org markup, an assistant has to guess what the page is: a company, a product or an article. JSON-LD states it plainly: name, services, address, phone, opening hours and FAQ.
- A forgotten noindex - a staging setting left on after launch tells search engines to skip the page. ChatGPT search, Copilot and Perplexity rely on search indexes, so the page drops out there too.
How to read the score
The score is the sum of the five areas, from 0 to 100. Hard caps keep it honest: an empty JavaScript shell or a noindex tag caps the score at 25, and a page with fewer than 40 words of text in its HTML is capped at 40. No amount of metadata helps if a crawler finds nothing to read.
- 85-100: Excellent, AI sees you. The technical side is in place. Your next gains come from content and from your reputation on other sites.
- 65-84: Decent, but you are leaving room for competitors. Typically missing structured data, an FAQ or a sitemap. Quick fixes.
- 40-64: Weak, AI skips your business. Something important is blocked or missing, often thin HTML content or a blocked search bot.
- 0-39: Nearly invisible to AI. Usually a JavaScript-only site, a noindex tag or a bot wall. This needs a developer, not a plugin.
The verdicts for each assistant follow the same logic. "Sees your site" means its search bot is allowed and the raw HTML holds at least 300 words. "Sees it partially" means access is fine but the content is thin. Gemini is treated differently: Google renders JavaScript, so a JavaScript-only site counts as partial for Gemini and as unreadable for the others. The checks in the report are sorted with failures first, so the top of the list is your to-do list.
What to fix first
Work from access to content, not the other way round. A perfect FAQ does nothing if the crawler is stopped at the door.
- 1. Unblock the search crawlers - remove robots.txt and firewall rules that stop OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot. Decide separately, and deliberately, about the training bots.
- 2. Remove any stray noindex - check both the meta robots tag and the X-Robots-Tag header.
- 3. Put the content in the HTML - server-side rendering or static generation, so your services, location and contact details are in the page source. On a JavaScript-heavy site this is the biggest job, and the one that matters most.
- 4. Add structured data - Organization or LocalBusiness, Service and FAQPage in JSON-LD, matching what the page actually says.
- 5. Fix the fundamentals - a sitemap referenced in robots.txt, HTTPS, a fast server response and a sensible page weight.
- 6. Write content worth quoting - an FAQ with the questions customers really ask, clear lists, visible contact details and dates.
A high score is the entry ticket, not the prize. For what happens after the technical fix, read my guide Does ChatGPT recommend your business?
How the full result is unlocked
The scan itself is free and needs no account. When it finishes, you see how many serious issues and warnings it found. To see the exact score and every check:
- 1. Enter your email and phone number and tick the consent box. I use them to send you the report and, if you want, to talk about the fixes. Your details are not sold or published.
- 2. Type in the 6-digit code sent to your email. It is valid for 10 minutes.
- 3. Read the result on screen - a copy with a link to it also goes to your inbox, and you can copy the link for your developer. The link stays active for 30 days.
The phone field accepts international numbers - enter yours with the country code, for example +44, +971 or +1. If anything in the form does not work for you, email me your website address and I will run the audit and send you the full result myself.
Who builds the fixes
Me. REELM is a one-person studio in Wrocław, Poland. I am Sebastian: I wrote this audit, and I design and code the websites that pass it. You talk to the person who does the work, not to an account manager.
- Remote, in CET - full overlap with the UK and the rest of Europe, two to three hours behind the UAE, and my afternoons cover the mornings on the US East Coast. Calls on Google Meet or Zoom.
- Fixes on your current site - robots.txt, firewall rules, structured data, sitemap, FAQ and rendering. If your platform cannot output proper HTML, I will tell you so.
- Or a new site built for this - hand-coded business websites and immersive, cinematic websites with content in plain HTML, schema.org markup, an open robots.txt and an llms.txt as standard, and Lighthouse 90+.
- Quoted individually - after a 15-minute call, in EUR, GBP or USD. No packages and no price list.
Want the Google side as well? The free SEO Audit checks indexing, speed and Core Web Vitals through Google PageSpeed, structured data and rich results, and the consistency of your brand and local details.
Frequently asked questions
Is the AI visibility audit really free?
Yes. The scan costs nothing and needs no account. To see the full result, you leave your email and phone number and confirm the email with a 6-digit code. The report then appears on screen and arrives in your inbox. There is no paid version behind it.
Does a high score mean ChatGPT will recommend my business?
No. The score measures technical access: whether AI crawlers may fetch your page and whether they find readable content and structured data when they do. That is a necessary condition, not a guarantee. Whether an assistant recommends you also depends on your content, reviews, mentions on other sites and your reputation.
Which AI assistants does the audit cover?
ChatGPT, Claude, Perplexity, Gemini together with Google AI Overviews, and Microsoft Copilot together with Bing. The tool reads your robots.txt rules for 15 bot tokens from OpenAI, Anthropic, Perplexity, Google, Microsoft, Apple, Meta, Amazon and Common Crawl, and gives a separate verdict for each of the five assistants.
My site ranks well on Google. Why would AI not see it?
Because Google renders JavaScript and most AI crawlers do not. A site built as a JavaScript application can rank on Google while the crawlers of ChatGPT and Perplexity receive an almost empty page. robots.txt and firewall rules can also treat AI bots differently from Googlebot.
Does the audit check my whole website?
No. It scans the one address you enter, usually the home page, together with your robots.txt, sitemap and llms.txt. Blocked bots or a firewall usually affect the whole site, so one page already tells you a lot. If you want every page reviewed, I do that by hand as part of a project.
Find out what AI sees. Then fix it.
Run the free audit. If the result shows problems, send me the link and I will tell you which ones are worth fixing first and what it would take. Every project is quoted individually.