We started in January 2026 by experimenting with a new AEO tracking tool on a single client account. The tool let us track prompts, mentions, and visibility in a way we never could before. By August, that experiment had become a fully productized service with tiers, a defined workflow, and its own SOP library.
Here are the 7 lessons that got us there.
1. All the data in the world is useless if nobody can read it
From sentiment scores and visibility rankings to citations and source URLs, the data is endless.
It is one thing to learn that you were mentioned a few times alongside two competitors in your market. The useful questions start immediately after that.
- Where did the AI get that information?
- Why did it list you third when you are better known and considerably larger than the two businesses above you?
That second question has an answer, and it is the one that changes what you do next.
The model was not evaluating your business. It was summarizing its sources. If the directories and articles it pulled from happened to present those two competitors more prominently, you are third, regardless of your actual standing in the market. Once you understand that, the work becomes obvious: go fix the sources. Without that reading, all you have is a disappointing number and no idea what to do about it.
This is the difference between knowledge and wisdom.
Knowledge is the reading the stats on the dashboard. Wisdom is knowing what that reading means in context and what it should change. A lab report is not a diagnosis. The numbers are real and necessary, and on their own they will not tell you whether to act, wait, or ignore what you are looking at.
Every AEO tool on the market will hand you dashboards. Visibility percentages, share of voice, competitor ranks, citation counts. None of that is an answer. It is just raw material. The question that matters is why the number moved, whether the movement is real or noise, and what to do on Monday because of it.
This was always true of SEO.
It is more true of AEO, because the data is newer, noisier, and almost nobody has built intuition for it yet. So every report we send gets read by a person first, gets a written interpretation in plain language, and ends in a ranked list of actions. Sometimes we watch the needle move inside the same week.
And here is the point most people miss. If you are already showing up somewhere, and you probably are in some way, you need to know why. That is the most valuable thing in the data. A win you cannot explain is a win you cannot repeat.
2. The channel was wide open, and third parties owned it
Our own client’s story makes the case. In April, we ran a source-level pull across one competitive local market and examined 2,128 URLs that large language models were citing. Directories and third-party publications carried nearly all of them. Almost no business in that market was being cited from its own website, including our client.
Two conclusions came out of that.
- This was not a deficit to close against entrenched competitors. It was an open position, and in most local markets it still is.
- How other people mention you matters as much as what you publish. Your site tells the model what you claim. Everyone else tells it whether the claim holds.
That finding reframed the entire engagement, and you can read how it played out in full here.
3. Mention, citation, perception, and traffic are four different questions
Most reporting collapses these into one number. They are not one thing.
- Visibility is how often you appear in an AI answer.
- Citation is how often your own site is the evidence behind that answer. Harder to earn, and it is what builds authority.
- Perception is what the AI actually says about you, and how it decides a source is trustworthy.
- Referral traffic is the people who arrive on your site from an AI platform. It is a real and separately measurable channel now, and most businesses are not watching it at all.
Perception deserves more room than I can give it here, and it will get its own post soon. The short version is that whether an AI mentions you and what it says about you are two different problems with two different fixes.
A model can bring you up consistently and still describe you in terms that do not sell your work. We regularly see brands described warmly, as established and caring and easy to work with, while trailing the field on credentials, expertise, and sophistication. That second gap almost never traces back to your own website. It traces to the third-party publications, rankings, and directories the model treats as authoritative.
The other half of this is you need to know what AI is saying about your competitors. Your description only means something when its next to theirs.
4. Averages hide the wins. Track specific prompts and specific markets
The useful view is prompt by prompt and market by market: which exact questions you win, which ones a competitor owns, which town is still open.
In our experience, one market nearly doubled inside a single reporting cycle because the work was pointed at it deliberately for a few weeks. That is only visible, and only directable, at the single prompt and market level.
5. Weekly beats monthly and quarterly, and it is not even close
AI answers change week to week. A quarterly report describes a landscape that no longer exists by the time it lands. Our loop is measure Monday, ship that week, verify the following Monday.
Compared to traditional SEO that would take months to see the change or results, AI rankings can change week to week. I am often amazed at how quickly we can course correct and own a particular platform. The results happen fast.
6. The measurement is noisy
Prompt sets change. Platforms surface duplicate URL paths. Competitor rankings sometimes shift retroactively. We disclose all of it on a like-for-like basis rather than letting a headline look better than it is. A business may grow in visibility if we are only comparing them to a few competitors, while others can drop drastically the day we add in a few more. Understanding the context of the data is what directs decisions.
7. Moving the number is not the same as moving the business
Think of a visibility score as a grade on a test. Somebody had to choose the questions.
If the questions (or prompts we are tracking) have your business name in them, you will score well, because your name was already sitting in the question. That is a very different thing from a stranger asking “who is the best family lawyer near me” and hearing yours.
So before you feel good about a number going up, ask what questions it came from. Are those the questions your customers actually ask? If they are not, the score went up and nothing else did.
What actually changes your position is easier to say than to do. AI has to pull from somewhere it can understand: your website, social media, directories, third-party websites, membership or certification. The measurement we care about most is not on the dashboard or report at all. For our client, the best thing that happened all year was the client running out of capacity and hiring to keep up. That is the story, and that is the result that matters.
Where we think this is going
Three shifts we are watching.
Delegation.
People are starting to hand the whole research task to an agent rather than reading answers themselves, which raises the value of third-party corroboration and review depth.
Traffic stops being the scoreboard.
Citation, share of voice, and branded search volume become the honest measures, and referral traffic from AI platforms becomes the conversion signal.
Divergence.
The platforms are describing the same businesses differently from one another, and that spread is widening. Single-platform reporting will not hold up. Every AI model is different, and every AI model account is catered to the user. It does make things more difficult, but it also gives us an opportunity to connect businesses with the right people.
What we do not expect is a shortcut. The businesses doing well here are the ones that were already clear, consistent, and corroborated, and who had someone paying attention to what the data was telling them.
That is an encouraging place to land, because it means the work is available to any business willing to do it carefully.
If you want to know how AI platforms are describing your business right now, that is exactly what our AI Search Results service is built to answer. If you are not sure you are ready for an ongoing program, our Website Audit and Review is a smaller first step that tells you where you stand.
Curious what the AI platforms are saying about you? Let us take a look together.
