What AI reads
on your website
AI does not read your website the way a human does. It does not scan pages visually. It reads HTML code, structured data and semantic elements. It prioritises headings, meta tags and JSON-LD schema. It ignores JavaScript-rendered content, images without alt text and hidden elements. Sites with complete structured data receive 3.1 times more citations in AI answers. This article explains exactly which elements AI reads, what it ignores and how to optimise your website for maximum AI readability.
3.1x
more citations with complete structured data
90%
of AI crawlers process schema markup
0%
JavaScript rendering by AI crawlers
H1-H3
headings determine how AI structures content
The elements AI reads first
AI crawlers process your website through a hierarchy of elements. They do not start at the top and read everything. They prioritise the elements that contain the most information in the fewest tokens. This priority order determines what AI takes from your website and what it skips.
Priority 1: Structured data (JSON-LD). This is the most direct and unambiguous source of information for AI. Schema markup tells AI in a structured format what the page contains: business type, location, services, author, FAQs. 90% of all AI crawlers process schema markup and it is the first layer that gets read. For UK businesses, this means marking up your LocalBusiness details including your Companies House number, VAT registration, service areas and opening hours. 68% of AI-cited websites have strong structured data.
Headings as information structure
Priority 2: Headings (H1 to H3). The H1 tells AI what the main topic of the page is. H2 headings define the sections. H3 headings indicate subsections. This hierarchy tells AI how the information is organised and what the relationship is between different topics on the page. A page without heading structure is, for AI, an unsorted pile of text. Think of your headings as the table of contents that AI uses to navigate your page.
Priority 3: Meta tags. The title tag and meta description give AI a summary of the page before it reads the content. A title tag of maximum 60 characters and a meta description of 155 characters that accurately describe what the page contains help AI to categorise the page correctly. Open Graph tags are also read and used. For a UK solicitor's page, a title tag like "Employment Law Solicitor Manchester | Free Initial Consultation | Smith & Partners" is far more useful to AI than "Our Services | Smith & Partners."
Priority 4: Paragraph text. The actual content of your page. AI reads the text in your paragraphs, lists and tables. Short paragraphs are processed better than long blocks of text. The first two sentences under a heading receive the most attention. That is the "quotable passage" that AI potentially incorporates into an answer. For maximum AI readability, keep paragraphs between 40 and 60 words and lead with the key point.
Links as context
Priority 5: Links and navigation. AI reads the anchor texts of links to understand which other pages are related. Internal links help AI understand the structure of your website. A link with text "read more about our bookkeeping for sole traders" tells AI more than a link with text "click here." Descriptive anchor text is a small change that significantly improves AI's ability to understand your site's content structure.
What AI ignores on your website
JavaScript-generated content. If your content only appears after JavaScript has executed, AI sees an empty page. This is the biggest technical risk for AI visibility. Websites built with React, Angular or Vue without server-side rendering are completely invisible to AI crawlers. Many UK businesses use website builders that rely heavily on JavaScript. If your site is built on a JavaScript framework, ensure you have server-side rendering (SSR) or static site generation (SSG) enabled. Test by viewing your page with JavaScript disabled in your browser.
Images without alt text. AI cannot analyse the visual content of images. It only reads the alt text as an HTML attribute. A product photo without alt text is, for AI, as though that photo does not exist. A photo with alt="DeLonghi Magnifica espresso machine, silver, in use on a kitchen counter" gives AI concrete information. For an estate agent, an alt text like "Three-bedroom semi-detached house in Didsbury, Manchester, front view with garden" is far more useful than "property-image-1.jpg."
Hidden and dynamic content
Hidden content. Elements with display:none, visibility:hidden or content in accordions that are collapsed by default are skipped by some AI crawlers. If your FAQ is in an accordion that only opens after a click, AI might not be able to read the answers. Consider making FAQ content visible by default in the HTML, even if you use JavaScript to collapse it for users. The HTML source code should contain the full text.
Styling and visual formatting. AI does not see colours, fonts, spacing or animations. A button that stands out visually because of its colour and size is, for AI, simply a link element. Your CSS framework does not matter for AI readability, except that complex frameworks generate more HTML tokens. The more bloated your HTML, the less actual content AI can process per crawl. Clean, semantic HTML gives AI more content per token.
Video and audio content. AI crawlers cannot watch videos or listen to audio. A YouTube video embedded on your page is, for AI, an iframe element and nothing more. If your video contains valuable information, put that information on the page as text as well. A transcript or summary below the video makes that content available for AI. Many UK businesses have useful video content that is completely invisible to AI because it is not transcribed.
Cookie banners and pop-ups. The content in cookie consent notices, pop-up windows and slide-in panels is treated as noise by AI. It reads the text but assigns it no relevance to the page's actual topic. Do not put important information in overlays or modals. Keep your core content in the main body of the page.
Rule of thumb: if it is not in your HTML source code, it does not exist for AI. Test by viewing your page with JavaScript off and CSS off. What you see then is what AI sees.
Is AI reading your website correctly?
VestVale automatically monitors whether ChatGPT, Gemini, Claude and Google AI cite your business. Find out whether your content is being picked up.
Optimising your website for AI readability
Use semantic HTML
Use article, section, nav, header and footer instead of only divs. Semantic elements tell AI the function of each section. A div without context is, for AI, an anonymous block. An article element tells AI that it contains content. A nav element tells it those are navigation links. This distinction matters for how AI interprets and prioritises your content.
Structure headings hierarchically
One H1 per page, H2s for sections, H3s for subsections. Do not skip levels (not from H1 to H3). The heading structure is the skeleton AI uses to understand your content. Phrase H2 headings as questions where possible: "What does conveyancing cost in London?" is better than "Our conveyancing fees."
Put important content in HTML
All text you want cited must be in the HTML source code. Not in JavaScript, not in images, not in videos. Duplicate visual content as text. A pricing table as an image is unreadable for AI. A pricing table as an HTML table is perfectly readable and quotable. This is a common issue on UK business websites.
Write descriptive alt texts
Not "photo-1" but "Interior of our accounting office in Manchester city centre with meeting room for 8." Alt texts are the only way AI understands visual content. Make them descriptive and informative. For product images, include the product name, key features and use case in the alt text.
Implement llms.txt
An llms.txt file in your root directory helps AI models navigate your website. It contains a summary of your site, your most important pages and contact information. Only 7.2% of websites have this. It is an early advantage that takes minutes to implement. Include your business name, what you do, where you operate and links to your key pages.
Minimise HTML bloat
AI crawlers have token limits. The cleaner your HTML, the more content AI can process per crawl. Remove unnecessary div nesting, inline styles and unused classes. More content per token equals more visibility. If your site uses a page builder that generates excessive HTML, consider whether the bloat is worth the convenience. Learn more in how AI reads websites.
Structured data: the most important element for AI readability
If you do only one thing to improve your AI readability, implement structured data. JSON-LD schema markup is the single most impactful change you can make. It gives AI a machine-readable summary of your page that requires no interpretation. Sites with complete structured data receive 3.1 times more AI citations than those without.
For a UK local business, the minimum structured data should include: LocalBusiness schema with your business name, address, phone number, opening hours, service area and description. FAQPage schema for any FAQ sections on your pages. Service schema describing what you offer with pricing information where applicable. Review and AggregateRating schema if you display reviews on your site.
Which schema types matter most
LocalBusiness schema is the foundation for any business serving a local area. It tells AI exactly what you are, where you are and what you do. FAQPage schema is the most directly quotable: it presents questions and answers in a format AI can immediately use. Service schema describes your offerings in a structured way that AI can match to user queries. For regulated UK professions, consider adding hasCredential or knowsAbout to demonstrate professional qualifications.
Testing your structured data is straightforward. Use Google's Rich Results Test to validate your markup. Check that all required fields are populated and that there are no errors. Pay attention to warnings too, as incomplete schema is less useful to AI than complete schema.
A common mistake UK businesses make is implementing schema but leaving fields empty or using generic values. A LocalBusiness schema with "Business Name: Company" and "Service Area: UK" is nearly useless. A schema with "Business Name: Smith & Partners Solicitors" and "Service Area: Greater Manchester, Cheshire, Lancashire" gives AI the specificity it needs. The more detailed your schema, the better AI can match you to relevant queries.
For e-commerce businesses, Product schema with offers, pricing, availability and reviews is essential. When ChatGPT answers "What is the best cordless drill under £100?", it looks for product pages with structured pricing data, specifications and ratings. The retailer whose product page has complete Product schema with an AggregateRating gets cited ahead of the one with just a product image and a buy button. Read more about why structured data matters in what is GEO?
Frequently asked questions
Does AI read my JavaScript content?
No. AI crawlers read raw HTML by default. JavaScript content is not rendered and is therefore invisible. If your website loads content dynamically via JavaScript, implement server-side rendering (SSR) to make that content available in the HTML source. Test by loading your page with JavaScript disabled in your browser. What you see then is what AI sees.
Which meta tags does AI read?
Title tag (max 60 characters), meta description (max 155 characters) and Open Graph tags are read by AI. But structured data (JSON-LD schema markup) is now more important than meta tags for AI visibility. Invest first in schema markup, then in meta tag optimisation. Good meta tags help AI categorise your page. Good schema markup helps AI quote you.
Can AI read images on my website?
Not the visual content. AI only reads the alt text as an HTML attribute. Vision models can technically analyse images, but AI crawlers do not use this when indexing websites. Put all important information that appears in images as HTML text on your page as well. Pricing tables, infographics and diagrams should always have a text equivalent.
How important is semantic HTML for AI?
Very important. AI uses the accessibility tree (the semantic structure of your HTML) to understand your page. Semantic elements like article, section and nav tell AI the function of each part. A page built entirely with divs gives AI no structural clues. A page with proper semantic markup is significantly easier for AI to parse and cite.
Does my website builder affect AI readability?
Yes. Website builders like WordPress with well-coded themes generally produce clean HTML that AI can read. Builders that rely heavily on JavaScript (some no-code platforms) may produce content invisible to AI. Page builders that generate excessive div nesting and inline styles create HTML bloat that reduces how much content AI can process per crawl. Check your source code.
What is llms.txt and should I have one?
An llms.txt file sits in your website's root directory and provides AI models with a structured overview of your site. It includes your business description, key pages and contact information. Only 7.2% of websites have one, making it an easy early advantage. It takes minutes to create and deploy. Think of it as robots.txt for AI models.
Is your website being read by AI?
VestVale automatically monitors whether ChatGPT, Gemini, Claude and Google AI cite your business. Find out whether your content is being picked up.
From £19.95/mo ex. VAT. Cancel monthly.