AI systems read a page in HTML order, before any visual design
The first thing an AI system reads on a website is the raw HTML in the order it appears in the source code, starting with the title tag, meta description and first heading, not whatever a person sees first on screen. Many AI crawlers never execute JavaScript, so anything that only appears after the page renders may not exist for them at all.
Short answer
An AI crawler reads a page as source code, top to bottom, before any styling or script changes what a person eventually sees. That means the title tag, the meta description and the first heading and paragraph in the actual HTML carry outsized weight, because for a large share of AI traffic they are the only content that reliably exists.
Source order is not screen order
A browser takes an HTML document and a stylesheet and produces a visual layout: a hero image might sit above a heading in the code but render below it on screen, or a sidebar written last in the markup might display first. People only ever see the rendered result.
A crawler that does not run a full browser sees none of that rearranging. It reads the document as written: whatever tag comes first in the source is what it encounters first, regardless of where a stylesheet later places it visually.
Why many AI crawlers see less than a person does
Vercel instrumented AI crawler traffic across its network and found that the crawlers behind ChatGPT and Claude do request JavaScript files, ChatGPT on 11.50% of requests and Claude on 23.84%, but never execute them. Anything that only exists once a page finishes building itself in the browser is invisible to a crawler working this way.
Google is the exception rather than the rule, and even its own documentation advises against relying on that exception, recommending server-side or pre-rendering because not every crawler can run JavaScript. In practice, a page's true first content, for most AI systems, is whatever exists in the HTML before any script has had a chance to run.
11.50% / 23.84%
of requests where ChatGPT and Claude crawlers respectively fetch a page's JavaScript files without ever executing them.
Vercel, The rise of the AI crawler
A crawler meets a page the way it was written, not the way it was designed to look.
What to put first, in practice
The practical checklist is short, because the position that matters most is the one furthest from any script: the page's static head and the opening of its body.
- Write a title tag that names the actual page, not a generic tagline.
- Write a meta description that states the answer, not a teaser for it.
- Put the real claim in the first heading and the sentence beneath it, in the markup, not after an animation or a client-side data fetch.
- Keep structured data in the document head, where it renders regardless of script execution.
None of this changes how a page looks. It changes what still exists when the styling and scripting are set aside, which is the version of the page a large share of AI systems actually read.
Common questions
Does the order of sections in my HTML matter for AI visibility?
Yes. Many AI crawlers read a page as a stream of source code and never execute the JavaScript that could rearrange it visually. Whatever appears earliest in the actual HTML, not on screen, is what such a crawler encounters first.
Do the title tag and meta description matter if AI does not see a page visually?
They matter more, not less. They are usually the first two pieces of text in the document and often the only ones a system reads before deciding whether the page is relevant at all.
If my homepage is a single-page app, does that hurt what gets read first?
It can. If the meaningful content only appears after JavaScript runs and builds the page in the browser, a crawler that does not execute that script sees an empty or near-empty document first, whatever the finished page looks like to a person.
Should my first heading restate my page title?
No. The title tag should name the page; the first heading should state the actual claim or answer. Repeating the title as the heading wastes the most valuable position on the page.
Is this the same thing as a classic Google ranking factor?
Related but distinct. Classic SEO cared about keyword placement near the top of a page. This is about whether the content exists at all, in the raw document, before any script runs.
AI-readable websites
The full practice this page is one principle of.
What is semantic HTML
The markup discipline that gives a document its structure.
Do AI crawlers read JavaScript sites
The deeper answer behind why source order matters this much.
Why structure beats keywords for AI
What to do with the position once you have earned it.
Find out what your own homepage looks like with the styling and scripts set aside.