A Dummy's Guide to AI Visibility

I’ve spent the last couple of months exploring the new, enchanting land of AI Visibility. This is what I learned - consider this a Dummy’s Guide to AEO.

So we’ve collectively entered the Intent Economy. What’s that, you may ask?  Well, it’s about LLMs figuring out your intent really, really well and then matching it with the most relevant answer, solution, or product.

Things were not like this before.

Say you wanted to buy trainers. You’d go to Google, type “women’s trainers,” and get a page of blue links. You’d click around, compare options, and eventually make a decision. Brands were competing to appear as high as possible in that list. They did it through technical site optimisation, a mountain of content, and backlinks that confirmed their credibility.

That was (and still is) SEO. Then AI came along and shook this fairly stable system.

Because now you’re not the one clicking through links and comparing things; the model does that for you. It checks sources, compares options, and gives you the answer it believes is the best match. The searching, filtering, and comparing happens behind the scenes without your participation. How we search has also changed. In the past, we’d just type “women’s trainers” and see sports brands competing for higher rankings in search results. Now, we share our full life story with an LLM: “I want cute trainers for spring, light colour but they won’t get dirty easily, mainly for streetwear and occasional run on a track outside”. This is good stuff. The more context, the better for an LLM. Now, what does it mean for brands? 

1. What’s in the name?

First things first. AEO (Answer Engine Optimisation), GEO (Generative Engine Optimisation), AI Search, AI Discoverability, AI Visibility - they all basically refer to the same thing.

It’s a set of tactics aimed at improving your brand’s visibility inside LLMs.

2. Search is shifting to LLMs.

The numbers vary, but it’s safe to say that around 30% of users already start their search with an LLM.

Call me a doomer, but I can see a scenario where the entire search journey collapses into a single window: query, comparison, clarification, discussion, and decision, all happening in one place.

We’re already seeing website traffic gradually declining year by year. 

3. Site-centrism is out, Omnipresence is in. 

This is probably the most important one (but still, please stay till the end).

The website used to be the major source of truth about the product. No more. LLMs cross-reference information from many places: your website, your socials, third-party listings, reviews, articles, and comments.

So it’s no longer just about what you say about yourself,  it’s also about what others say about you. And these should ideally match.

As an example, a friend of mine who runs a dance school was looking for a CRM and asked her LLM for recommendations. Here’s roughly what the model will do: it will check all those sources.

Why “roughly”? Because language models are:

a) probabilistic b) weight those sources slightly differently

So you can never be 100% sure how the answer will come out.

4. What others say about you suddenly matters a lot.

Brands need to pay much closer attention to how they appear outside their own controlled channels.

LLMs care about consistency. If what your website claims doesn’t match what people say in reviews, forums, or articles, the chances of being recommended go down.

5. Structured content on your website is even more important (hi, SEO!)

LLMs love it when a site has a clear structure and specialisation. They like when everything is organised, connected, and clearly positioned in a niche (who doesn’t, right?) Ideally your brand becomes an expert in a specific domain. The old type of content like “What is project management?” isn’t particularly useful anymore. Models have already consumed, digested, and internalised that.

What they need now is current, specialised, living content. That’s what feeds RAG - retrieval-augmented generation (also a fun term to memorise for a quiz night at a local pub).

6. From keywords to entity associations.

In traditional SEO, we picked keywords we wanted to rank for and built everything around them. Now we’ll be playing a different game, called “entity association”.

First, what’s an entity? Pretty much anything: a company, a person, a product, a concept… or your dog. LLMs build knowledge graphs that map how entities connect to one another. Which means it suddenly becomes very important what company you keep. My grandmother used to say that too; turns out she was right.

You need to be clear about who you are and what entities you want to be associated with.

In practice, those associations are built through things like: product documentation, blog posts and articles, media mentions, GitHub repositories, and community discussions.

7. Multimedia rules.

Google literally published a blog a couple of days ago about where search is heading, and it’s heading in that direction.

People are increasingly lazy (no judgment, I’m among them) and don’t want to read walls of text. They want videos, visuals, and images. Search will prioritise content that includes visual elements. Of course, it depends on the brand and the industry, but the general direction is pretty clear.

8. If you have a physical business, keep your Google Business profile in good shape.

All models look and they strongly prefer businesses with complete, well-filled profiles. A properly filled-out profile increases the likelihood of being cited in LLM answers by up to 7 times.

9. How do you actually see how your brand appears inside LLMs?

There are a few tools. Semrush and Ahrefs are great places to start: they can tell you how often your brand is mentioned and cited, and where it appears.

If you want deeper insights, like how people talk about your brand and with what sentiment, there’re specialised tools for that. I’ll be honest, I don’t think sentiment analysis is faultless at the moment, so I’ll refrain from recommending a particular tool; go and test them out.

And if you’re a true connoisseur, you can always run your own experiment: ask each LLM 100 slightly different questions and see how often your brand appears. You can also do this via APIs, but you’ll need a budget.

10. So what are we actually measuring?

Well, I’m glad you asked. This field is still evolving, but most AI visibility reports track something like this:

Mentions. Are we mentioned at all? Out of 100 prompts, how often?

Citations. Are we referenced with a source link? This is your gold. Citations include a link so the user can actually visit your website, not just see your name mentioned.

Sentiment. How is the brand described? Popular? Niche? Emerging?

Rank. Where do we appear in the answer? Are we the model’s favourite, or just one of several options? (sigh)

Traffic. How much traffic comes from LLMs? This one is tricky because not every conversation leads to a click. But there are still ways to detect whether those conversations are influencing your pipeline.

Most brands are just setting off on their AEO journey. If you need someone to sail this boat with, let’s chat.