Wondering How to Improve Website AI Visibility? It Starts (But Doesn’t End) with SEO

Generative AI is really a misnomer. This technology doesn’t generate new answers so much as it reconfigures existing information. So if you’re trying to improve a website’s AI visibility, the site itself is the first, most important resource that LLMs will turn to.

That means a site has to be accessible, perform well (load quickly and render correctly) and provide an intuitive framework for information delivery reflected in its organization and content. If that sounds pretty much identical to the factors that determined a site’s visibility to search engines, that’s because it is. There are people who say that SEO is AEO/GEO, and they’re not wrong. But it also doesn’t end there.

Internet Innovations and the Laws of the Universe

Imagine a scenario with me: You’re an AI exec pouring trillions into development and billions into data infrastructure. The entirety of the investment in this technology is oriented towards the math that underlies linguistics, because that’s a tremendously complex nut to crack.

Are you going to start from scratch when it comes to the technology that makes up the cogs in the larger machine? Probably not. So even though this innovation is truly revolutionary, it’s still going to ride the tailwinds of existing technology wherever possible.

Innovation almost always piggybacks on what’s come before. Because the name of the game isn’t just innovation, but cost-effective, rapid innovation. That’s especially true for technology that’s a part of the primary innovation, but not all of it.

Why expend valuable computational power (read: money) to effectively reinvent how information is harvested from the web if you don’t have to? My hunch (and that of many other, smarter people than myself) is that LLMs utilize the same well-worn paths that have been driving SEO for many years.

First, Make the Site Accessible

Google is a user experience company. Read that one again: Google is a user experience company. Sure, most people know the company as the first name in search, but user experience is how it got there. Specifically, not its own user interface and experience, but rather, yours.

Early on in the evolution of traditional search, the powers that be at Google figured out that if a site was included in search results, but it offered an awful user experience, people would fault the search result, not the site. That’s why the company made site performance a core part of how it ranked sites in results. The better the experience on the site, the more favorably people regarded Google itself.

Google weighs all of it: how fast your site loads, how much outdated content or dead-end content you have, major jumps between desktop and mobile presentation. It’s only maintained market share this long because it treats your site like an extension of its product. The math is simple: if someone finds a result on Google that fails to meet their needs, they may look elsewhere the next time.

For current AI systems, and even more so in the near agentic future, access is the first qualification for indexation of a site. If your site is outdated, slow, clunky, gated or has a ton of pop-ups, odds are AI is just going to jump right over it. Again, part of the value proposition of this technology is speed, and if you think human attention spans are short, take a look at machines’.

Second, Text = Language in “Large Language Models”

We know that these “large language models” need text, and a lot of it. Where and how to consistently access text across websites was Google’s chief early innovation.

Their search technology piggybacked on visual impairment technology, specifically web browser screen readers. In the early days of the Internet, a standard protocol was adopted to allow visually-impaired users to access what was then, and is still now, a highly visual medium.

But for screen readers to be able to perform consistently, certain text fields had to be available across any page on the web. So four fields became the backbone of the technology:

• Meta title
• Meta description
• H1 (the primary heading on the page)
• The actual copy itself

Knowing that those fields should be present on any page, Google made them the basis of its search engine information retrieval process.

Yes, Google actually “read” each of those fields in order to determine what a page was about. It logged consistencies in terminology across them as a basis for contextual relevance. All of this was recorded on a page-by-page and site-by-site basis and stored in Google’s “index” of the Internet.

We know that web indexes play a huge role in LLMs. Gemini’s training model essentially is Google’s index, and ChatGPT retrieves information from the web using the Bing index (no surprise there either, given how much Microsoft has invested in OpenAI).

Just like that, four key fields on web pages became more important than ever. If you haven’t been able to make it clear to Google what your business is, or what you do, odds are AI is going to struggle with that very same thing.

You might laugh, but this is the essence of SEO. The problem is that Google published the rules in this space decades ago (explaining them poorly to this day), but I still encounter huge brands that have zero search visibility because they use abstract, esoteric language to talk about what they do.

Third, Close Your Own Gaps

With AI, that actually could get worse. If you’re not playing by the rules, it may skip over you altogether. Or, even worse, if you leave gaps in your positioning, it may fill them in with inference, speculation, hearsay or hallucination.

That’s why AI search optimization begins with SEO. A solid traditional search footprint is the first thing to account for. From there, encouraging accurate retrieval is the name of the game, which I’ll cover in my next post.