Does website design affect AI visibility?

Ask three agencies whether website design affects AI visibility and you’ll hear three different pitches: a redesign here, a schema package there, a new file to upload at the root of your site. The confusion is built into the question, because the word design means two separate things. One is how your site looks. The other is how it is built and organised under the surface. Those are not the same lever, and treating them as one is how marketing budgets end up in the wrong place.

AI search behaves differently from the ten blue links. When someone asks ChatGPT or Google’s AI Overviews a question, the system pulls passages from across the web and writes an answer, often splitting one question into several searches before it does. The prize is being chosen as a source, not holding position one. If you want the mechanics in full, we have covered how AI engines retrieve and cite content separately.

This guide walks through which kind of design actually moves AI visibility, the technical fault that can hide your content completely, how to structure pages for citation, the truth about schema, the tactics worth ignoring, and how to tell whether any of it is working.

Which kind of website design affects AI visibility?

Two kinds of design affect AI visibility, and they don’t carry equal weight. Visual design, meaning your brand styling and layout, has almost no direct effect on whether an AI engine cites you. Structural and technical design, meaning how the page is built, delivered and organised, is what decides whether your content can be read at all.

The reason is simple. Most AI systems read the text and code of a page, not the rendered result a person sees. Your colour palette, your hero image, the polish of your layout: none of it reaches the model in the way it reaches a visitor. A page can look dated and still be cited constantly. A page can look beautiful and be completely absent from AI answers.

There is one shift worth watching. Google has noted that browser agents, the AI tools that complete tasks on a site, may read a page by taking screenshots, inspecting its underlying structure, and reading its accessibility tree, according to Google’s guide to optimising for generative AI search. As those agents grow, the visual and structural layers may start to converge. For now, though, the line holds: structure is what counts, and the rest of this guide stays there.

Can AI engines actually read your pages?

If your content only appears after JavaScript runs in the browser, most AI engines can’t see it. As of 2026, the major crawlers behind ChatGPT, Claude and Perplexity fetch the raw HTML of a page and don’t execute JavaScript, so client-side rendered content is effectively invisible to them, however good it is.

This is the fault that sits underneath everything else. Two terms make it clear. Server-side rendering means the full content is delivered as finished HTML before it reaches the browser. Client-side rendering means the browser receives a near-empty shell and builds the content with JavaScript afterwards. Google’s own crawler can handle the second case, because it runs the scripts. The newer AI crawlers, for the most part, do not.

How common is the problem? More common than most teams assume. A scan of the most-linked 1,000 sites on the web found that, of those that could be read at all, 57% showed their content only after JavaScript ran, which works out at around 40% of the full top 1,000 serving a near-empty page to any agent that does not render (RenderPeek, measured 28 May 2026). These are not small or neglected sites. Many are names everyone knows.

The behaviour is now well documented. None of the major AI crawlers render JavaScript, including GPTBot, ClaudeBot and PerplexityBot, with Google’s Gemini the main exception because it uses Google’s existing rendering infrastructure (Lantern, March 2026). An AI crawler is simply an automated agent that fetches pages for an AI system. Here is who renders and who doesn’t:

• GPTBot, OAI-SearchBot, ChatGPT-User (OpenAI): do not render JavaScript
• ClaudeBot, Claude-SearchBot (Anthropic): do not render JavaScript
• PerplexityBot (Perplexity): does not render JavaScript
• Gemini (Google): renders, via Googlebot’s infrastructure

The commercial sting is the quiet bit. You can rank well on Google and still be missing from AI answers, and your traffic reports won’t tell you, because the visit never happens.

Are AI bots even reaching your pages?

Rendering is one access problem. Permission is the other. Some sites block AI crawlers without meaning to, often because a security setting or content delivery network defaults to blocking AI bots. There is also a distinction marketers miss: blocking the search bots that feed live citations is not the same as blocking the training bots that collect data for future models. Cut off the search bots and you disappear from the source links in AI answers, even if your only intention was to keep your content out of model training. It is worth asking your developer to confirm which bots your site currently allows.

How to structure content so AI will cite it

AI engines retrieve and cite specific passages, not whole pages, so content that’s easy to lift gets cited more often. The patterns that help are practical: a direct answer in the opening line, self-contained sections under clear headings, and specific figures backed by named sources. This is the part the property and enterprise teams we work with tend to underinvest in, and it is where the gains are largest.

Extractability is the idea underneath all of it. Extractability is how easily an AI system can pull a self-contained, accurate passage out of your page and drop it into an answer. Three habits drive it.

Lead with the answer

Open every section by answering the question it raises, in the first sentence, in full. An AI engine scanning your page is looking for a clean, quotable response it can trust. Bury the answer three paragraphs down and you make the model work for it, which it often won’t. The senior growth team at HubSpot put it plainly in EMARKETER’s analysis of GEO and AEO: the first sentence of a page should resolve the primary question, because answer engines look for that quick confirmation (EMARKETER, April 2026).

Write sections that stand alone

Because engines pull individual chunks, each section needs to make sense on its own, without leaning on the paragraph before it. Clear headings help, ideally phrased the way a real person would ask the question. This is not a new skill. It is search engine optimisation applied with retrieval in mind, and Google’s guidance agrees that organising content under clear headings serves both human readers and machines, while adding that you don’t need perfect code or content broken into tiny fragments.

Back claims with sourced figures

Specific, sourced numbers are one of the strongest signals you can send. The original academic study on this, by researchers at Princeton, Georgia Tech, the Allen Institute and IIT Delhi, found that adding statistics, citations and quotations lifted source visibility by up to 40% across a wide range of queries, and by up to 37% on Perplexity (Aggarwal et al., presented at SIGKDD 2024). The pages that gained the most were not the ones already sitting at the very top of search. Mid-ranked, well-evidenced content saw the biggest lift, which is good news for any brand that does not yet dominate its category.

Does schema markup help your AI visibility?

Schema markup is worth keeping, but it isn’t the AI visibility lever many vendors claim. Google says no special structured data is required for AI search, and a controlled study found that adding schema produced no citation uplift. Schema clarifies your content for search systems; it doesn’t unlock AI citations on its own.

Schema markup, also called structured data, is code (usually JSON-LD) that labels what the content on a page means, so machines can read it without guessing. Plenty of agencies sell it as the route into AI answers, with claims of large citation lifts. The evidence does not support treating it that way.

Three positions sit on the table, and they are worth holding together:

• The vendor pitch: widespread claims of big gains, often citing a 2.5x higher chance of appearing in AI answers for pages with full schema.
• The controlled test: Ahrefs tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026 against 4,000 control pages, and found no citation uplift on any AI platform, with a small 4.6% decline in AI Overviews (via Search Engine Roundtable, May 2026).
• Google’s line: structured data isn’t required for generative AI search and there’s no special schema to add, though it remains useful for rich results in traditional search.

The honest read is that schema still earns its place for the search benefits it has always had, and it helps some Bing-powered surfaces. It does not work as a citation strategy, and it is not a reason to fund a rebuild. If you want the basics laid out plainly, our GEO and AI SEO FAQs cover the common questions.

The AI design tactics you can safely ignore

Several widely sold AI tactics do little or nothing for visibility, and a senior marketer planning a budget should know which to skip. Google has named most of them directly in its official guidance, which makes them straightforward to drop without second-guessing.

Four come up again and again:

• llms.txt files. llms.txt is a proposed text file that summarises your site for AI tools. Google Search ignores it. Creating one won’t help or harm your Google visibility, though it may have niche uses for other systems.
• Chunking content into tiny blocks. There is no requirement to break content into small fragments for AI to understand it. Google’s systems read longer pages fine, and mechanical chunking earns nothing extra.
• Rewriting content just for machines. You don’t need a separate, robotic version of your copy. AI systems understand synonyms and meaning, so writing naturally for people is enough.
• Chasing inauthentic mentions. Buying or manufacturing brand mentions across the web is not the shortcut it appears to be. Quality content and genuine coverage are what the systems reward.

None of this means technical and content work is wasted. It means the work should go into access, structure and evidence, not into files and hacks that Google has already told you it ignores.

How do you know if your AI visibility is working?

You can’t manage AI visibility you can’t see, and standard analytics won’t show it. Traffic reports weren’t built to tell you whether an AI engine cited you, so measuring AI visibility needs dedicated tracking rather than clicks and sessions alone.

The gap is wider than it looks. Major publishers including Reuters and The Guardian receive less than 1% of their referral traffic from AI platforms despite being cited frequently, according to Similarweb’s 2026 GenAI Brand Visibility Index (via EMARKETER, April 2026). You can be part of the answer thousands of times and barely register a click, which means clicks are the wrong scoreboard. Share of voice in AI answers, meaning how often your brand is cited or mentioned relative to competitors across AI platforms, is closer to the real measure.

What you actually want to track is citation and mention frequency across the platforms your buyers use, how that compares to your competitors, and how it moves over time rather than in a single snapshot. We built a visibility tracking tool to measure exactly this. It started in property and now runs across all industries, recording where brands appear across organic, paid, map and AI results, which feeds directly into our benchmarking work.

This is the same discipline behind our enterprise SEO results. Our long-running SEO partnership with Hamptons International is built on the structural foundations this guide describes: clean delivery, well-organised content, and measurement that ties activity back to commercial outcomes rather than vanity figures. The principles scale from a single landing page to an estate of hundreds.

The bottom line

Website design does affect AI visibility, but the version that matters is structural, not decorative. For a senior marketer deciding where to spend, the order of priority is clear:

• Make sure AI can reach and read the page, which usually means serving content server-side.
• Structure content so it can be lifted and cited: answer first, self-contained sections, sourced figures.
• Keep schema clean for its search value, without overpaying for it as an AI tactic.
• Skip the hacks Google has already told you it ignores.
• Measure visibility properly, because traffic alone hides it.

Almost all of this is the SEO and GEO discipline your team may already understand, applied with AI retrieval in mind, rather than a costly rebuild. If you want a clear read on where your site stands today, our generative engine optimisation work is the place to start.

Frequently asked questions

Does AI search read JavaScript websites? Mostly no. The major AI crawlers behind ChatGPT, Claude and Perplexity fetch raw HTML and don’t run JavaScript, so content that only appears after scripts load is invisible to them. Google’s Gemini is the main exception, because it uses Google’s rendering infrastructure. If your main content depends on client-side rendering, it likely needs to be served server-side instead.

Is GEO different from SEO? They overlap heavily. Google’s position is that optimising for AI search is still SEO. Generative engine optimisation is best understood as applying SEO and content fundamentals with AI retrieval and citation in mind, rather than a wholly separate discipline. A site with solid SEO foundations is far more likely to be cited by AI engines.

Does schema markup improve AI visibility? There is no clear evidence that it does. A controlled study of 1,885 pages found no citation uplift from adding schema, and Google says no special structured data is needed for AI search. Keep clean schema for its traditional search benefits, not as an AI citation strategy.

Does page speed affect AI visibility? It can. AI crawlers work to tight time limits and may abandon slow, heavy pages before reading them, so fast delivery and lean pages help your content get fetched. Treat speed as a prerequisite for visibility rather than a direct ranking lever.

Do I need an llms.txt file to appear in AI search? No. Google Search ignores llms.txt and similar files, so creating one won’t help or harm your Google visibility. It may have niche uses for other systems, but it is not a route into AI Overviews or AI Mode.

How do I check whether AI can read my website? View the page source, or load the page with JavaScript disabled, and confirm your main content is still there. If the page is nearly empty without scripts, AI crawlers likely see the same blank shell, and your content needs to be delivered server-side.