What Marketing Leaders Get Wrong About AI Visibility (It Starts with UX)

The Request that Lands on My Desk

The brief usually arrives the same way. A marketing leader has invested in content, the blog is publishing on schedule, and yet the brand is nowhere to be found when someone asks ChatGPT or Gemini a question it should clearly answer. The request is "fix our AI visibility," and the assumed fix is more content, better keywords, or a new AEO checklist.

That assumption is where most teams go wrong.

AI visibility is a user-experience problem before it is a content problem. Once content quality is met, what decides whether AI tools find and reuse your answer is whether your site is fast and well-structured. Language models behave like impatient visitors: if extracting an answer costs effort, they deprioritize the page.

Content quality is the floor. If you have no content, or thin content, or content written for neither humans nor language models, nothing downstream matters. Content is still king. But most teams stop at the floor. The next constraint is not another blog post; it is whether your website is structured and fast enough for a machine to bother reading it.

That is a user-experience question. And it is the part marketing leaders most often skip.

What the Evidence Actually Shows

Two shifts make this urgent in 2026, and neither is hypothetical.

The first is reach. AI-generated answers now sit on top of a large share of search. BrightEdge's tracker found AI Overviews appearing on roughly 48% of tracked queries by February 2026, up from about 31% a year earlier. Other trackers measure lower depending on the method. Conductor put it at 25.11% across 21.9 million queries in Q1 2026, but every credible dataset points in the same direction: the answer layer is now a regular feature of the results page, not an edge case. Alongside it, a majority of searches now end without a click to any website. If you are not inside the synthesized answer, you are increasingly invisible in the moment that matters.

The second shift is conversational. At Google I/O 2026 in May, Google merged AI Overviews and AI Mode into a single flow: a user can ask a follow-up directly from an Overview and slide into a back-and-forth without losing the original context. Google's head of Search framed it as the biggest change to the search box in more than 25 years. AI Mode itself passed a billion monthly users within a year of launch, with Gemini now the default engine behind AI Overviews. The interaction is no longer one question and one answer. It is a conversation.

Underneath both shifts is a measurement Google has used for years. Core Web Vitals, the LCP, INP, and CLS metrics that score loading, responsiveness, and visual stability, are a confirmed ranking signal within page experience. They are not the primary signal; Google is clear that strong content beats fast-but-empty pages. They behave as a tiebreaker between comparable pages. But that is exactly the point I want marketing leaders to sit with: the same metric that measures human experience is the one that gates machine access to your content.

Why This is a UX Problem, Not a Content Problem

When I audit a site for AI visibility, the first thing I look at is Core Web Vitals, because they give me a fast read on how the site performs for a real person. If pages take too long to load, or a visitor cannot find what they came for above the fold, that is a poor experience. And a language model behaves like your most impatient visitor.

Gemini in particular, since it draws on Google's own performance signals, prefers to extract content from sites that are designed and built to perform. Crawling and parsing cost computation. A model has no incentive to spend that budget on a slow, poorly structured page when a faster, cleaner source answers the same question. People are not patient, and Google spent a decade digitizing that impatience into measurable thresholds. The machines inherited it.

Structure is the other half. Clean heading hierarchy and schema markup help a crawler understand what a section is without working for it. When a model can tell at a glance that a block is an FAQ, a product page, or an "about" section, it spends less effort identifying the content and more confidently reuses it. The caveat is that the schema is an aid to extraction, not a guarantee of citation. It helps a machine read you; it does not make a thin page worth quoting.

Our View: The UX Umbrella Swallowed Everything, Including Content

Here is where we land, and it is a deliberately broad claim. There is no longer a meaningful line between SEO, content, performance, and UX. The UX umbrella expanded until it covered all of them. The content itself is UX. The page speed is UX. The structure is UX. When a marketing leader treats AI visibility as a content initiative that sits next to a separate technical backlog, they have already split a problem that is no longer divisible.

This reframes who owns the work. AI visibility is not a campaign the content team runs while engineering handles "the website." It is a property of the website that the content lives inside. The reason this matters for the budget conversation is simple: you can hire more writers and publish more often and still lose, if every new page lands on a slow, fluffy, badly structured site. The model meets the same wall that the human does.

That is the strategic mistake. Not under-investing in content, but assuming content is the variable when the constraint is the experience around it.

A Page-Level Framework for AI Visibility

If I am revamping a site, I do not start with AI tactics. I lock each page first. The discipline is to refuse the temptation to cram everything onto one page that is supposed to serve a single purpose. For each page:

1. Start from a clear sitemap. Map the logical flow of information in and out of every page before designing anything. AI visibility fails first as an information-architecture problem.
2. Define one purpose per page. Know the single objective of the page and the audience it serves. If you cannot say it in a sentence, the page is not ready.
3. Put the answer above the fold. Make the page's purpose obvious to its target audience immediately. This is the same instinct that helps a human and a model: lead with the answer, then provide context.
4. Design around that purpose, with nothing extra. No fluff, no bloat, no loading more than the page needs. Restraint is a performance strategy, not just an editorial one.
5. Align to the brand and to Core Web Vitals together. The best design serves the brand and the user and stays inside the performance thresholds. These are not competing goals.

A note on what this framework is not for. Heatmaps and analytics are essential for understanding human behavior, lowering bounce rate, and finding the choke points on an existing site. Lean on them when you revamp. But they tell you how people move through a page; they do not tell you how a model reads it. They inform UX decisions, not AI visibility directly. Keep the two honest in your reporting so you do not credit a heatmap fix with an AI visibility gain it did not produce.

Where Drupal Fits

I will be specific about the trade-off rather than sell a platform. For a simple brochure site or a small store, an enterprise CMS is overkill. Where it earns its place is the heavy end: content-heavy, media-rich, multilingual, multi-domain properties that have to perform under load. That is the kind of site, the United Nations tier of organization, where the defaults matter most.

Drupal was built for that tier, and a large share of the performance and structure best practices are addressed before a marketer touches anything. That is the practical benefit: the heavy lifting on performance and structured content is largely done, which frees the marketing team to spend its time on visibility rather than on fixing fundamentals. We have built on that foundation for organizations like UNHCR and UNICEF, where load, languages, and structure are not optional extras. The deeper argument for why structured content underneath the editor is what makes a site citable by AI is one I made in a companion piece on the AI-native CMS.

Beyond AI Visibility: Staying in The Conversation

The I/O 2026 changes move the goalposts, and marketing leaders should move with them. Appearing once, as a link or a cited source in a single answer, is now table stakes. The real question is what happens next when the user keeps probing inside the same conversation.

Think of it like walking through a shopping mall. You pass sixteen shops. They are all visible. But which one do you actually walk into, and which one do you keep coming back to? Visibility gets you the storefront. It does not get you the conversation.

This means treating ChatGPT, Gemini, and the rest as representatives of your brand. You need to understand how you show up in the deeper turns of a conversation, not just the first answer, and what influences it. Part of that is your own site as the source of truth. Part of it is the mirrored profile of your brand on third-party platforms like Clutch, where a summarized version of who you are and what you have delivered feeds the same models. If your site is pristine but your off-site presence is thin, you have optimized for the storefront and ignored the conversation inside.

The mistake is not that marketing leaders care too little about AI visibility. It is that they treat it as a content problem bolted onto a website, when it is a property of the website itself. Get the experience right, for the impatient human and the impatient machine alike, and visibility follows. Then start thinking about the conversation that comes after the first answer.

Wondering where your site sits?