If You Remember One Thing About AI Visibility, Make It This

Key Takeaways 

• No tool can show you what people are actually asking AI platforms, so every visibility number you see is modeled from testing, not measured from real behavior. 
• The fundamentals of optimization haven’t changed much, since good content on a crawlable site backed by real authority still does the heavy lifting for both AI and search. 
• Earned media and an active social presence matter more than ever, now that AI leans on third-party sources to validate what it says about a brand. 

The way people use AI is increasingly changing, but one pattern I anticipate will continue growing is people swapping the search bar for LLMs to do research. Instead of typing a few keywords into Google, they’re asking ChatGPT, Claude, Perplexity or Gemini a full question and taking the answer at face value. Naturally, brands want to know how often they’re showing up in those answers and how they compare to the competition. 

But this is where a lot of well-meaning strategy goes sideways. AI visibility doesn’t work like the search visibility we’ve all spent years getting comfortable with, and treating it like SEO is the fastest way to get it wrong. Before chasing AI visibility, it’s worth understanding how it actually works and where it’s different. 

The terminology is a moving target, but a few terms come up constantly. 

SEO is search engine optimization, or improving how a brand shows up in traditional search engines like Google. And while we’re here, please retire the phrase “SEO optimization.” The “O” already covers it. 

AIO is AI optimization. It’s the umbrella term for making sure AI models understand a brand and surface it across platforms, and it also covers using AI to measure and improve that visibility, which is part of why it’s such a useful catch-all. 

GEO, or generative engine optimization, points specifically at generative engines like ChatGPT and Perplexity. The one catch is that GEO is also short for geolocation, so things get confusing fast. 

There are a handful of other variations floating around, and most of them mean the same thing. The label matters far less than what’s underneath it. 

The most important thing to understand is this: AI visibility and search visibility differ in one big place, and that’s the data. 

Search engines share a fair amount of information about what people are searching for. Open Google Search Console and you can see the top queries bringing people to your site, how many clicks each one earned and how many impressions came with them. That’s the backbone of traditional SEO. It’s how a brand figures out where it ranks, where its content is falling short and whether the work is paying off over time. 

On the other hand, large language models share none of it. ChatGPT, Claude, Perplexity and Gemini keep their prompt data private. They don’t share what people are asking or how often, and they’re not about to start. 

Despite what some tools claim, there’s no platform on the market that can tell you what real people are actually typing into AI. Plenty of them imply otherwise with slick dashboards that make it look like you’re seeing real prompts, but you’re not. Any score they give you is modeled, not measured, an estimate built from controlled testing instead of actual user behavior. We dug into why that distinction matters, and how to read those scores without fooling yourself, in a previous post on what really matters in AI visibility. 

This is the expectation that trips most brands up. After years of seeing exactly what people searched, where they ranked and how it moved week to week, it’s natural to expect AI to work the same way. It doesn’t, and getting that straight early saves a lot of headaches down the road. 

Here’s the good news. Not being able to see real prompts isn’t the same as not being able to learn anything. It just takes a different approach. 

Since the real data is off the table, AI visibility tools work off a set of likely prompts, tested over and over across time. You won’t get the exact wording people use, but you can start to see where a brand shows up consistently and where it’s nowhere to be found. 

The trick is grounding those prompts in something real, and traditional search data is the best place to start. The queries in Search Console come from actual people, and someone searching for a topic on Google is usually asking the same thing in an LLM, just in longer, more conversational language. A query is the short search-bar version, a couple of keywords typed in a hurry. A prompt is the fuller, more specific question someone would actually ask a model, usually longer and more conversational. Pull your prompts from the real queries you already know people use, and your testing goes from guesswork to grounded. 

One catch is that LLMs rarely answer the same way twice. Ask the same question a few minutes apart and you can get two different responses. Once you add personalization into the mix, no two people will get the same response. That’s why volume matters. You need the same prompts sent across multiple models many times over, until the noise of any single answer averages into a pattern you can trust. 

It’s also worth knowing that Google is the only one sharing even a sliver here, since its AI answers get folded into Search Console data. Even then, it’s blended in with everything else, and it won’t tell you which results were AI or show you where you landed in them. So even with Google, the real questions stay out of reach.

Once you’re past the measurement problem, optimizing for AI looks a lot like optimizing for search. Underneath it all, a crawler is still a crawler. 

A crawler is that creepy little robot that visits a site, reads what’s there and files it all away. Both Google and the AI platforms use them, and since a crawler only understands code, it can only work with what it can actually read. That’s why a well-built site with clean architecture and proper schema matters just as much for AI as it ever has for search. 

The whole thing has even followed the same arc. SEO started as keyword stuffing and grew into something simpler, which is to write genuinely good content. AI rewards the exact same thing. So, when a company leader asks what they need to do differently, the honest answer is usually “less than you’d think.” It comes down to three familiar pillars. 

• Content, which means high-value writing in the words real people actually use, not stuffed with keywords or buried in marketing speak. 
• Technical, which means a crawlable, well-structured site with the basics in place so both the crawlers and real people can find you. 
• Authority, which means the outside validation that tells every engine your brand is credible and worth citing. 

That third pillar, authority, is where it gets really interesting for communicators. 

A brand’s own website is hardly ever the only thing AI pulls from. These models cross-check their answers against all kinds of sources, which is exactly why authority, built through earned media and an active social presence, has become so valuable. 

Pull the list of sources feeding AI answers and you’ll often find a surprise or two. It’s not unusual to see Reddit, YouTube and even Facebook show up high on the list. Facebook especially has turned into a real source as Meta keeps pushing its content into the AI ecosystem, and honestly, it makes sense. It’s where people go to complain, leave reviews and tag brands in public, and that kind of high-volume, public chatter is exactly what these models reach for. Shaping that is a real opportunity for communicators. 

It also opens up a different kind of competitive insight. By watching how a brand shows up, which competitors get cited more often and which of their pages keep surfacing, you can see exactly how that brand looks to someone asking an AI tool from the outside. And when a competitor’s third-party article or review keeps getting cited while your own content never shows up, that’s telling you something specific about how these models judge authority in your space. 

The thing to remember is that AI visibility is never a single-channel fix. One press release or one earned placement chips away at the authority pillar, but it won’t move the needle on its own. Real visibility comes from content, technical and authority all pulling together, not from any one of them by itself.

No platform knows what people are searching for on AI, and no tool can hand you the real prompts. Anyone who says otherwise is flat out wrong. 

But that doesn’t leave us in the dark. You can still model the likely behavior, watch the trends, dig into the sources and learn enough to start moving the lever in the right direction. 

The measurement really is new, and you can’t treat an AI visibility score like a search ranking without getting burned. But the work that earns that visibility – good content on a clean, crawlable site backed by real authority – is the same work communicators have always been good at. 

Want to understand how your brand is showing up across AI and traditional search? Get in touch.

Lexi Trimpe is Director of Digital + AI at Franco. Connect with her on LinkedIn.