Your website is probably invisible to ChatGPT, and the reason is technical, not editorial. Most AI assistants read only the raw HTML a server sends. They do not run the JavaScript that builds modern sites in the browser. If your content, headings or structured data only appear after that JavaScript runs, the AI sees an empty page and cites a competitor instead.
That gap matters more every month. AI-referred website sessions rose 527% between January and May 2025, and ChatGPT alone now serves around 800 million weekly users. The buying journey increasingly starts inside an AI answer rather than a list of ten blue links, so being absent from those answers is a commercial problem, not a curiosity.
This post explains why it happens, how to check whether it is happening to you, and where the line sits between a quick fix and a proper engineering job.
Why can’t ChatGPT see my website?
AI search crawlers fetch your page and read what the server returns. Unlike Google’s main crawler, most of them do not execute JavaScript. GPTBot, OAI-SearchBot, PerplexityBot and ClaudeBot see only the initial HTML response. If your site renders its content client-side, which is common with many React, Vue and single-page-app builds, that initial response can be close to empty.
The result is blunt. A page that looks perfect to a human in a browser can be a blank shell to an AI. Everything your marketing team wrote is there on screen, but it arrived too late for the crawler to read.
This is also why a beautifully designed site can quietly underperform in AI search while an older, plainer competitor gets cited. The competitor’s content is in the HTML. Yours is assembled after the crawler has already left.
How do AI engines decide who to cite?
Each major engine sources its answers differently, which means there is no single trick that wins all of them. Where Google returns ten links, large language models typically cite just two to seven sources per answer, a far narrower window, so position matters enormously.
ChatGPT Search leans heavily on Bing’s index (the two correlate by roughly 73% to 87%) and strongly favours encyclopedic, authoritative content, with Wikipedia accounting for close to 48% of its citations. Perplexity is different again: around 47% of its citations come from Reddit, and it heavily favours content published or updated recently. Google’s AI Overviews stay closest to traditional search, with roughly 92% of their citations coming from pages already ranking in the top ten organic results.
The practical lesson is that your own website is necessary but not sufficient. Getting cited also depends on third-party mentions, community presence and freshness, surfaces most businesses ignore entirely.
How can I tell if my site has this problem?
You can check the most important signal yourself in under a minute. Open your homepage, right-click and choose View Page Source, not “Inspect”, which shows the assembled page. View Source shows the raw HTML a crawler receives. Then search that source for a sentence from your hero text.
If your headline and key copy are in the source, you clear the biggest hurdle. If the source is mostly empty containers and script tags and your text is not there, your content is being rendered client-side and AI crawlers cannot read it. While you are there, search the source for application/ld+json, which is your structured data. If it only appears via your tag manager or after JavaScript runs, AI crawlers miss it too.
For a definitive view, run your page through Google’s Rich Results Test, which renders the page like a crawler and lists exactly which structured data it can detect. And ask the engines directly: prompt ChatGPT or Perplexity with “What does [your company] do?” The answer tells you what they currently believe, which is often an outdated version of your business.
What actually fixes it?
Three things do most of the work, and they are all technical.
First, server-side rendering. Your critical content and structured data must be present in the initial HTML response, before any JavaScript runs. For some stacks this is a configuration change. For a client-rendered app it can mean re-architecting how pages are delivered.
Second, server-rendered structured data. Schema markup is one of the strongest AI-visibility signals. Around 81% of pages cited by AI systems include structured data, and comprehensive schema makes a page roughly 36% to 40% more likely to appear in AI answers. But it only counts if the crawler can see it in the raw HTML, not if it is injected client-side.
Third, answer-first content structure. Analysis of thousands of AI citations found that a concise, self-contained answer of 40 to 60 words placed immediately under a clear question-style heading is the single most consistent predictor of being cited. Add specific statistics with sources roughly every 150 to 200 words, since one study found this alone improved citation rates by over 40%, and keep your most important pages genuinely fresh, since content updated within 30 days earns around 3.2 times more citations.
None of this is mysterious. But doing it correctly across an entire site, auditing what crawlers actually receive, fixing the rendering, structuring the schema properly, and maintaining it as engines change their behaviour, is engineering and editorial work combined, and it is easy to get subtly wrong in ways that silently cost you visibility.
Is this just SEO with a new name?
No, though it builds on the same foundations. Traditional SEO optimises for ranking in a list of links. AI visibility, often called GEO, for Generative Engine Optimisation, optimises for being cited inside a synthesised answer. The technical groundwork overlaps, covering crawlability, structure and authority, but the goals and success metrics differ. You measure SEO in rankings and clicks. You measure GEO in citations and mentions.
Crucially, GEO does not replace SEO. Because Google’s AI Overviews draw so heavily from top-ranking pages, strong traditional SEO remains the foundation that AI visibility is built on. The two reinforce each other.
Frequently asked questions
Does my website need to be rebuilt to appear in ChatGPT?
Not always. If your content is already server-rendered, the fixes are mostly additive, such as structured data, better headings and fresher content. If your site renders entirely client-side, the rendering itself needs addressing, which is a larger job. A short technical audit tells you which situation you are in.
Will blocking AI crawlers protect my content?
It will also make you invisible in that engine’s answers. Blocking OAI-SearchBot removes you from ChatGPT Search, and blocking PerplexityBot removes you from Perplexity. For most businesses seeking visibility, allowing the search crawlers is the right call, even while you decide separately about training crawlers.
How long until I appear in AI answers after fixing this?
It varies by engine. Once crawlers re-read your improved pages and any caches refresh, changes can surface within days to a few weeks. Resubmitting your sitemap in Google Search Console and Bing Webmaster Tools speeds up re-crawling.
Do I need an llms.txt file?
Probably not. Despite some early enthusiasm, no major AI provider has confirmed it reads llms.txt files, and recent guidance suggests they have little practical effect. Your effort is far better spent on server-rendered content and schema.
At Flux Dynamics we build sites that are visible to both search engines and AI assistants: server-rendered, properly structured, and engineered to be cited. If you are not sure whether your site can be read by ChatGPT, tell us what you are working with and we will take a look.