You can write the best content in your sector, but if an AI system cannot reach, read or understand your website, you appear nowhere. Many brands work on their message and forget the technical layer beneath it: is my site readable for AI at all? That is what an AI readiness score measures. This article explains what AI readiness is, which technical factors really count, and where most sites stay invisible to AI.
What is an AI readiness score?
An AI readiness score measures how well AI systems like ChatGPT, Claude, Gemini and Perplexity can reach, crawl, read and understand your website. It is not a measure of whether you get cited, that is your GEO Score, but of the technical condition behind it. If AI cannot process your site, it cannot cite you either.
The difference matters. Your GEO Score tells you whether you appear in the answers, and how to raise it is covered separately. Your AI readiness score tells you whether it is technically possible to get in there at all. A low AI readiness is often the hidden cause of a low GEO Score: your content is fine, but AI cannot reach it or reads it wrong. Veesie measures the AI readiness of your website for marketers and agencies on several factors and sums them up in a score from 0 to 100, alongside the GEO Score it tracks across ChatGPT, Claude, Gemini and Perplexity.
Can AI reach your site at all? The AI crawlers
The first question is access: does your website let the right AI bots in? Every major AI provider uses separate crawlers with a different role, and that is often where it goes wrong.
According to the official OpenAI documentation, four OpenAI bots are in play, each with its own task: GPTBot crawls content to train models, OAI-SearchBot makes sure your site appears in the ChatGPT search function, ChatGPT-User fetches pages when a user actively asks for them, and OAI-AdsBot checks advertising pages. Anthropic works in a similar way with ClaudeBot (training), Claude-User (real-time questions) and Claude-SearchBot (search). Perplexity has PerplexityBot for search results and Perplexity-User for answers.
Here is the biggest blind spot. OpenAI confirms it directly: sites that block OAI-SearchBot do not appear in the ChatGPT search answers. Many websites block AI bots in one sweep to "stop scraping", and in doing so accidentally shut out the very search crawler that makes them findable. The distinction is crucial:
- If you want to be visible in AI answers, let the search crawlers in: OAI-SearchBot, PerplexityBot and the search crawlers of the other providers.
- If you do not want your content to train models, block the training crawler specifically (GPTBot via
robots.txt), without giving up your search visibility.
A robots.txt for anyone who does want to appear in AI answers but not train models looks like this, for example:
User-agent: GPTBot
Disallow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
This keeps you visible in the search functions of ChatGPT and Perplexity, while you do not hand your content over to model training. The opposite, blocking everything with one blanket rule, is exactly how sites make themselves invisible in AI answers.
For Google there is a separate case: Google-Extended is not a separate crawler but a robots.txt token that controls whether Google uses your content for Gemini training. Google states explicitly that this token does not affect your position in regular Google search results.
Does AI see your content, or only a blank page?
Even when an AI crawler gets in, the question is whether it actually sees your content. Many AI crawlers do not execute JavaScript. According to an analysis by Vercel, the crawlers of ChatGPT and Claude do fetch JavaScript files, but they do not run them. The result: content that is only built up in the browser with JavaScript does not exist for those bots.
This is a silent killer for modern sites. Webshops and apps that load their product info, prices or text client-side often show an AI crawler a near-empty page. The fix is to keep your most important content in the server-rendered HTML, not only after a JavaScript step. For AI readiness there is a simple test: turn JavaScript off in your browser and check whether your core content is still there. What you cannot see then, an AI model usually cannot see either.
A second check is "view page source" in your browser. If your real text, your headings and your key facts are simply present in that HTML, an AI crawler can read them. If you mostly see empty containers and scripts, that is a red flag for your AI readiness, however strong your content is on substance.
Does structured data help for AI?
Structured data helps to make your page machine-readable, but it is no silver bullet for AI citations. That nuance matters, because a lot of nonsense gets sold around it.
Ahrefs studied 1,885 pages that added JSON-LD schema between August 2025 and March 2026, against 4,000 control pages. The outcome: schema produced no meaningful rise in AI citations on any platform. Citations in Google AI Overviews even dropped slightly, while the changes in Google AI Mode and ChatGPT were statistically indistinguishable from zero. In other words: adding schema does not push a page higher in AI answers.
Even so, structured data is not useless. As Searchviu notes, the negative result applies mainly to direct chatbot retrieval. For search engines, Google AI Overviews and entity recognition, schema stays valuable: it helps machines understand who you are and what your page is about. The advice is therefore nuanced: use structured data for your broader visibility and clear machine readability, not as a lever for AI citations on their own. The most common misunderstandings about schema and AI are worked out in 8 GEO myths marketers still believe.
What about llms.txt?
llms.txt is a proposed file to point AI to your most important content, but expect little from it for now. The idea comes from Jeremy Howard (Answer.AI), who published the proposal in September 2024: a markdown file at the root of your site that hands AI models your key pages in a single overview.
In practice, adoption has stalled. No major AI provider publicly confirms that it reads or honors llms.txt. Google's Search Advocate even compared it to the keywords meta tag, a signal that search engines have ignored for years. The conclusion: an llms.txt takes little effort and does no harm, but it is not a part of your AI readiness from which you should expect growth. Put your energy into access and renderability instead.
The four pillars of AI readiness
Veesie measures AI readiness across four categories that together decide whether AI can use your site. For marketers and agencies, Veesie makes concrete where the technical gains are.
- Technical accessibility: do you let the right AI crawlers in? Is your core content in the server HTML, with no JavaScript dependency? Is your site fast and on HTTPS?
- Structured data: are your pages fitted with correct Schema.org markup for clear machine readability and search engine context?
- Content signals: is your content clearly structured with headings, lists and definitions, so a model can pull passages from it?
- Authority: are you recognized as an entity through consistent mentions and authoritative sources?
A high score across all four does not automatically mean you get cited, but a low score almost certainly means you do not. AI readiness is the gateway, your GEO Score measures what makes it through.
How to improve your AI readiness in 5 steps
Improving AI readiness starts with removing technical blockers, not with more content. Five concrete steps:
- Check your robots.txt. Make sure you do not block the search crawlers (OAI-SearchBot, PerplexityBot) by accident. Block only what you genuinely do not want, such as training crawlers.
- Test your renderability. Turn JavaScript off and check whether your core content holds up. If not, make sure it is server-side rendered.
- Structure your content. Use clear H2 headings, lists and direct definitions, so a model can extract relevant passages.
- Add correct structured data. Not as a silver bullet, but for machine readability and search engine context.
- Measure and repeat. An AI readiness audit from Veesie scans these factors automatically and shows you per category where the biggest blind spots are.
If you want to work through this as a checklist you can tick off, use the GEO audit checklist. And if you want to know whether it works, compare your position with the GEO benchmark by sector and track your GEO Score.
Conclusion: readable first, visible next
AI visibility does not start with better text, but with a site that AI can reach, read and understand. Most blind spots are technical and invisible: a wrongly blocked crawler, content that only appears with JavaScript, or the wrong expectations of schema and llms.txt.
AI readiness is therefore the quiet base under your whole GEO strategy. Veesie makes that base measurable for marketers and agencies, scoring your GEO Score across ChatGPT, Claude, Gemini and Perplexity, so you first make sure you are readable, and only then fight for the citation. Because invisible to the crawler is invisible in the answer. See What is GEO? for the full framework, read 8 tips to improve your GEO Score for the path after your AI readiness, or create a free Veesie account to measure your own score.