Enter any URL and see, in seconds, whether ChatGPT, Perplexity, and Google AI Overviews can actually crawl, parse, and cite that page.
Run GEO Audit
Live robots.txt, llms.txt, and HTML scan
What Gets Checked
GEO Score
Out of 100
No result yet
Enter a URL on the left and hit "Run GEO Audit."
Paste in a URL and the tool runs four live checks against the real page, the same way an AI crawler would encounter it.
We fetch your robots.txt and check it against 13 named AI crawler user agents, from GPTBot to Meta-ExternalAgent, to see who is allowed in and who is blocked.
We parse your JSON-LD structured data, heading hierarchy, FAQ formatting, and raw visible text to see how easily an AI system can extract facts.
We check title length, meta description, canonical tag, and Open Graph tags, the baseline metadata every citation-worthy page still needs.
Each of these has a documented, named user agent. Blocking one does not just hide you from one company, it removes you from a specific downstream AI product.
| Bot | Company | What Blocking It Costs You |
|---|---|---|
| GPTBot | OpenAI | Your content is excluded from the data OpenAI uses to train future models, and pages ChatGPT tries to browse directly may fail to load. |
| OAI-SearchBot | OpenAI | You lose eligibility to appear as a live citation inside ChatGPT's search-powered answers, even if your content is otherwise a perfect match. |
| ChatGPT-User | OpenAI | When a real user asks ChatGPT to open or summarize your page directly, the fetch fails and the user sees an error instead of your content. |
| ClaudeBot | Anthropic | Claude cannot crawl or index your page, which removes you from Claude's training data and any web-browsing citations Claude generates. |
| anthropic-ai | Anthropic | Blocks a secondary Anthropic training crawler, reducing how thoroughly your domain is represented in Claude's underlying knowledge. |
| PerplexityBot | Perplexity | Perplexity's answer engine cannot index your page, so it cannot surface or cite you even when your page is the best available source. |
| Perplexity-User | Perplexity | Live, on-demand fetches triggered by a user pasting your URL into Perplexity will fail. |
| Google-Extended | Your content is excluded from the data that feeds Gemini and Google AI Overviews, separate from your normal Googlebot search indexing. | |
| CCBot | Common Crawl | You are removed from the public Common Crawl dataset that many smaller AI labs and open-source models train on. |
| Bytespider | ByteDance | ByteDance's AI models and TikTok-related AI features cannot ingest your page content. |
| Amazonbot | Amazon | Amazon's AI features, including Alexa's answer generation, cannot use your page as a source. |
| Applebot-Extended | Apple | Apple Intelligence's on-device and cloud AI features are blocked from training on or referencing your content. |
| Meta-ExternalAgent | Meta | Meta AI and Llama model training pipelines cannot crawl your page. |
Several major CDN and hosting platforms have shifted how AI crawlers are treated by default over the last two years. Cloudflare, for example, shipped a one-click option to block all known AI bots in 2024, then in 2025 changed the default for newly onboarded domains to block AI crawlers unless the site owner explicitly opts them in. If you migrated hosting or CDN providers recently, it is worth re-checking your robots.txt, since a platform-level default may have silently blocked bots you never intended to block.
The 100 points are split across four weighted categories. Access comes first because nothing else matters if the crawler cannot get in.
30 points for how many of the 13 AI bots robots.txt allows, weighted toward GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. 10 points for a live llms.txt file. 5 points for llms-full.txt.
10 points for having any valid JSON-LD at all. Up to 10 more for meaningful schema types like Article, Product, FAQPage, or Organization.
Single H1, at least two H2 sections, an FAQ or Q&A block, and enough visible text in the raw HTML response, not just after JavaScript runs.
Title length, meta description length, a canonical tag, and the three core Open Graph tags: og:title, og:description, og:image.
Work through these in order. Access first, structure second. A perfectly structured page still scores zero on Crawlability if it is blocked at the door.
Go to yoursite.com/robots.txt and search for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. If you see Disallow: / under any of them, or under a wildcard User-agent: * block with no explicit Allow for that bot, remove or scope down that rule.
Create a plain Markdown file at yoursite.com/llms.txt listing your site name, a one-line summary, and links to your most important pages grouped under headings. This is the format proposed by Jeremy Howard in 2024 and is now checked by an increasing number of GEO tools.
At minimum, add Organization schema site-wide and Article or Product schema on content pages. If the page has an FAQ, wrap it in FAQPage schema so AI systems can lift each answer as a discrete, quotable unit.
Use exactly one H1 per page stating the topic plainly. Break the body into H2 sections that each answer one specific sub-question. Avoid using headings purely for visual styling.
Write 4 to 8 questions in the exact phrasing a user might type into ChatGPT or Perplexity, followed by a direct one to two sentence answer before any elaboration. This format is disproportionately easy for AI systems to extract.
View the page source, not the rendered DOM. If the visible text you expect is missing from the raw HTML, your content likely depends on client-side JavaScript that most AI crawlers do not execute. Server-render or statically generate that content instead.
Add a title between 10 and 60 characters, a meta description between 50 and 160 characters, a self-referencing canonical tag, and all three core Open Graph tags. These are quick wins that round out the score.
JSON-LD is not just for Google's rich results anymore. It is a structured, unambiguous fact sheet that AI systems can parse without guessing at meaning from prose.
Declares your headline, author, publish date, and publisher explicitly. AI systems use this to attribute claims correctly and to judge freshness, which matters for time-sensitive queries.
Marks each question and answer pair as a discrete, machine-readable unit. This is consistently one of the easiest formats for an AI system to lift into a generated answer with attribution intact.
Gives AI shopping and comparison features exact pricing, availability, and identity data instead of relying on parsing marketing copy, which is where most misattribution happens.
A quick honesty check on schema
Schema helps AI systems parse your page correctly once they have already decided to fetch it. It does not, by itself, make a blocked page reachable, and it does not make thin or generic content suddenly quote-worthy. Treat structured data as a translation layer for content that is already good, not a substitute for writing something worth citing.
llms.txt is a proposed convention, introduced by fast.ai co-founder Jeremy Howard in September 2024, for giving AI systems a curated, Markdown-formatted map of a site's most useful content.
Lists your site's key pages under short, human-readable headings so an AI system can quickly find your documentation, pricing, or core content without crawling and parsing your entire HTML sitemap.
It is not an official standard adopted by OpenAI, Anthropic, Google, or Perplexity, and there is no public confirmation that any major AI crawler actually fetches or prioritizes it today. Treat it as a low-cost, forward-looking addition, not a guaranteed lever.
llms-full.txt
A companion file some sites publish alongside llms.txt with fuller page content inlined, rather than just links. It costs little to add once llms.txt exists and can only help if an AI crawler chooses to read it.
AI systems typically break a page into chunks before deciding what to retrieve. Clean, one-question-per-section headings make chunking accurate. Messy headings make it a guess.
One H1, stated as the core topic. A page about 'GEO audits' with an H1 like 'Welcome' gives an AI system no topical anchor. State the subject plainly in the H1 itself.
H2s framed as sub-questions, not vague labels. 'How is the GEO Score calculated' extracts cleanly into an answer. 'Details' or 'More info' does not carry enough signal to be useful on its own.
Short paragraphs under each heading. A 2 to 4 sentence answer directly under a heading, before any elaboration, is far easier to lift verbatim than a heading followed by 400 words of buildup.
Tables and lists for anything comparative. Comparisons, pricing, and step sequences parse more reliably from an actual <table> or <ol> element than from a paragraph describing the same thing in prose.
Real visible text in the initial HTML. If your framework renders the body only after JavaScript executes, most AI crawlers, which generally do not run a JS engine, will see an empty or near-empty shell regardless of how good the eventual content is.
They overlap more than most guides admit, but the failure modes are different enough that a pure SEO checklist will miss real GEO problems.
| Dimension | Traditional SEO | GEO |
|---|---|---|
| Primary goal | Rank a URL as high as possible on a search results page. | Get your facts, product, or brand quoted directly inside a generated AI answer. |
| Gatekeeper | Googlebot and Bingbot, generally permissive by default. | 13-plus distinct AI crawlers, several of which are now blocked by default on major hosting and CDN platforms. |
| Core lever | Backlinks, keyword targeting, Core Web Vitals, and click-through rate. | Crawler access, structured data, and content that answers one question per section in plain language. |
| Success signal | Ranking position and organic click traffic in Search Console. | Being named or quoted inside an AI answer, often with zero click-through at all. |
| Failure mode | Ranking on page two or three, still technically visible. | A single Disallow rule in robots.txt can make a page invisible to an AI system entirely, regardless of content quality. |
The stakes are not hypothetical. Google's own research has found that around 75 percent of mobile searches with an answer box get resolved without the user ever leaving the results page. As AI-generated answers absorb more of that behavior, whether AI crawlers can even reach your page becomes a bigger swing factor in visibility than another five backlinks.
Most of these are accidental. A CDN default, a copy-pasted robots.txt, or a rebuild that dropped structured data.
Site owners often copy a robots.txt template from a tutorial or another project without checking what user agents it blocks. A single 'User-agent: GPTBot / Disallow: /' line silently removes you from ChatGPT's training and citation pipeline.
Some hosting and CDN platforms now block AI crawlers by default for newly onboarded domains. If you switched providers recently and never explicitly reviewed the AI bot settings, you may be blocked without a single line changed in your own robots.txt file.
A trailing comma or an unescaped quote inside a JSON-LD block makes the entire schema fail silently. The tag renders fine visually, so nobody notices until an audit tool actually tries to JSON.parse it.
Client-rendered single page apps frequently ship an HTML shell with a near-empty body and inject the real content after hydration. Most AI crawlers request the raw HTML once and move on, never seeing that injected content.
We would rather you trust this tool for what it actually measures than oversell it. Here is the honest boundary.
The 13 major AI crawlers checked here are not explicitly blocked by your robots.txt.
Your page has parseable JSON-LD with recognized, meaningful schema types.
Your heading structure and FAQ formatting follow patterns AI systems reliably extract.
Your core meta tags and Open Graph data are present and within recommended length.
Your content exists as real, visible text in the raw HTML response, not just after client-side rendering.
That any specific AI system will actually choose to cite or quote your page for a given query.
That your content is factually accurate, original, or better than a competitor's on the same topic.
That an AI crawler already respects robots.txt at all. Some AI bots have been documented ignoring it entirely.
That your ranking in traditional Google search will improve, since this tool does not measure backlinks or Core Web Vitals.
That llms.txt is read or prioritized by any major AI system today, since it remains an unofficial, voluntary convention.
That is a legitimate business decision and this tool will not tell you which choice is right for your site. What it will tell you is the tradeoff clearly: blocking training crawlers like GPTBot or anthropic-ai also tends to block the same company's citation and browsing crawlers, since many sites apply one blanket rule. If you want to allow citations while blocking training, you generally need separate, more specific robots.txt rules per user agent, which is worth doing deliberately rather than by accident.
Results are cached for 5 minutes per URL, so back-to-back runs on the exact same address return the same numbers. After that window, a changed score usually means the page itself changed, a CDN or edge rule updated robots.txt, or a deploy altered your JSON-LD or meta tags. It is rarely random.
No. It audits the single URL you enter, plus that domain's robots.txt, llms.txt, and llms-full.txt. For a full-site rollout, run the audit on your homepage, your highest-traffic content pages, and any page you specifically want AI systems to cite, since GEO issues can vary page by page even on the same site.
There are very few free GEO auditors that check AI crawler access this specifically, most existing tools focus only on schema or on a generic SEO checklist relabeled for AI. Here is where this one is strong and where it has real limits.
A perfect GEO Score makes your page reachable and parseable. It does not, by itself, get you mentioned. AI systems still need to encounter your brand somewhere trustworthy first, and Reddit threads are one of the most consistently cited sources in AI answers because they read as candid, unpaid opinion. Tools like MediaFast help you show up in those threads with genuine, useful answers instead of trying to force citations through technical settings alone.
Related GEO Resources
A clean GEO Score gets AI crawlers through the door. MediaFast helps you get quoted where AI models pull real answers from: candid Reddit threads, not just your own metadata.
Everything you need to know about GEO audits, AI crawler access, and how the GEO Score is calculated.
A GEO (Generative Engine Optimization) audit checks whether a web page is technically readable and citable by AI systems like ChatGPT, Perplexity, Claude, and Google AI Overviews. It looks at whether AI crawlers are allowed to fetch the page, whether the page has structured data and clean headings AI systems can parse, and whether meta tags and llms.txt files exist to help AI tools understand and summarize the content correctly.
Yes, completely free with no login required. Enter any public URL, click Run GEO Audit, and you get a GEO Score out of 100 along with a categorized checklist and prioritized fixes. We fetch the page's robots.txt, llms.txt, and HTML live, so results reflect the current state of the page.
The GEO Score is a 0 to 100 weighted score built from four categories: AI Crawlability (45 points, covering the 13 major AI crawler user agents plus llms.txt and llms-full.txt), Structured Data (20 points, covering JSON-LD presence and meaningful schema types), Content Structure (20 points, covering heading structure, FAQ format, and visible text volume), and Meta and Social (15 points, covering title, meta description, canonical tag, and Open Graph tags). AI crawlability is weighted heaviest because if the bots cannot fetch the page, nothing else on the page matters.
No. A high score removes technical barriers so AI systems can access, parse, and understand your page. It does not guarantee a citation, because AI systems also weigh topical relevance, the specificity of your answer, how often other trusted sources reference you, and how the underlying model chose to retrieve or was trained on content for that particular query. Think of the GEO Score as clearing the entrance requirements, not winning the competition.
robots.txt is the first gate. If GPTBot, ClaudeBot, PerplexityBot, or Google-Extended are disallowed there, the AI crawler typically never requests the page at all, so none of your schema, headings, or content quality matter for that particular bot. Checking robots.txt first mirrors how AI crawlers actually behave: check the rules, then decide whether to fetch.
A regular SEO audit focuses on ranking signals for traditional search: backlinks, Core Web Vitals, keyword density, and page speed. A GEO audit focuses on whether AI systems can access and extract your content at all, and whether that content is structured in a way that is easy to quote directly in a generated answer. Some checks overlap, like meta descriptions and headings, but AI crawler access and llms.txt are unique to GEO and are not part of most legacy SEO checklists.
