There are two kinds of AI visibility report, and your agency needs both

If you sell digital marketing to local businesses, you have probably spent the last year watching various new AI tools spring up - all promising to make your clients more visible.

I get it. The space is noisy, everything is new, and the labels have not settled down yet. Is it AEO? GEO? AIO? Who knows.

But I think a lot of confusion comes from a much simpler missing distinction, and once you have it, the whole thing gets a lot simpler. So I want to borrow something we already understood perfectly well in the old world of SEO.

We used to know exactly what each tool was for

Cast your mind back to how search used to work (you know, those bygone times of... about 6 to 12 months ago). There were two very different jobs in SEO, and nobody in the industry confused them.

On one side you had the technical audit. You crawled a site, found the broken things, the missing title tags, the slow pages, the pages accidentally blocked from indexing, and you fixed them. It was concrete. You could point at a problem, do the work, and point at the same thing again to show it was gone.

On the other side you had rank tracking. That measured performance. Where does this site sit for these keywords, and is it climbing or sliding over time. You did not control that directly. You influenced it, then you watched.

A fixer and a measurer. Two separate jobs. And nobody sane ever promised a client a guaranteed number one on Google. The people who did were charlatans, and everyone knew it.

Winning at AI visibility has exactly the same two jobs. We just have not named them yet.

This is the bit the market has skipped over. Many of the new tools on the market have failed to separate the two: they promise the earth but in reality contain a confusing blend of actionable fixes and abstract, indirect performance metrics.

As an agency it's pretty hard to understand what you need to fix or how to explain it in simple terms to your SMB clients.

In our view there are two separate jobs do be done with two separate solutions, just like in the old SEO days.

There is the technical AEO audit, which looks at the factors you control and lets you fix them. And there is the AI visibility monitoring, which measures how visible a business actually is across the AI platforms. They answer different questions, they are used differently, and if you are reporting to a client, you want both for different use-cases. Let me take them one at a time.

The audit: the things you can fix, and prove you fixed

This is the AEO side, answer engine optimisation. It is the on-page and listing factors that you directly control.

An AEO audit looks at things like:

• How clearly the content is written, and how complex it is to read. AI has to be able to work out what a business does, where, and for whom.
• Whether the pages actually say what the business does, in real depth, and whether that content is fresh rather than years old.
• Whether there are clear calls to action.
• The basic technical hygiene, like structured data and an llms.txt file.
• The quality and accuracy of the listings and reviews across the places AI looks.

One honest word of warning here. Not all of these carry equal weight. When I dug into the data on 10,000 US businesses earlier this year, the technical hygiene like schema and llms.txt barely moved the needle. The things that did were FAQs, clear content, accurate listings and detailed reviews. So fix the housekeeping items because they are housekeeping, but do not sell them as the main event, and spend the real effort where it actually counts.

The reason this side matters so much to an agency is simple: every one of these is something you directly control. If a business's phone number is wrong in a directory, that is not a mystery. You can sell a solution to fix it, deliver that solution and then prove it is fixed. If the content is thin, you make it better. This is your safe ground.

With the right kind of tools, you can:

1. Show a prospect the problem;
2. Do the work; and then
3. Show them the same report again with the problem cleared.

The tracking: the thing you can influence, but never guarantee

This is the measurement side. Think of it as rank tracking for the AI world.

Instead of crawling the site, you ask LLMs directly. You put questions to ChatGPT, Perplexity, Gemini and Grok, and look at the answers they give.

• Does the AI know this business exists at all?
• If it does, has it got the details right (the address, the hours, the phone number, the services)?
• When someone asks for a business like this one, does it get recommended, and how is it described?

The clever part is in the questions. You build a fan-out of the topics a business should be known for, then ask both branded and unbranded versions of each.

If it is a café known for its coffee, the branded question is "is this place known for a good latte?". The unbranded one is "where is the best place to get a latte in this town?". The first tells you whether the AI agrees with the business's own story. The second tells you whether the business even shows up when a real customer is choosing, with no name to go on.

That is the closest thing we have to rank tracking in the AI world, and it is genuinely useful.

But here is the part nobody wants to say out loud: you cannot guarantee it.

Just as you could never promise a number one on Google (and doing so would land you in hot water pretty quickly) you cannot promise that ChatGPT will recommend a business. You can improve the odds tremendously, and most of what improves them is the same content and reputation work the technical audit points at. But it is not deterministic.

Sometimes you do everything right and the needle barely moves. Sometimes it moves for reasons you cannot fully explain. And if you chase the wrong tactic, you can burn a month optimising for something that was never going to matter.

So if anyone offers a client a guaranteed spot in AI results, treat them exactly as you would have treated someone promising a guaranteed number one back in 2010. Same promise, same problem, new label.

Same two jobs as before, then, just pointed at a new set of engines:

Why the split matters most when you are the one reporting to a client

This is where the difference stops being academic and starts shaping how you win and keep clients.

Lead with the audit. It is the work you can definitely deliver and definitely prove. You can walk into a pitch, show a prospect exactly what is wrong with how AI sees their business, and it lands because it is specific and it is fixable. Then you do the work and show the same report again with the issues gone. That is how you win a client, and it is how you earn the trust to keep them.

Use the tracking to tell the longer story. Visibility across the AI platforms is your evidence that the work is paying off over time, and it is a brilliant reason for a client to stay engaged month after month. But present it honestly. Show it as a journey, not an overnight fix. Some months will be up, some flat, the odd one will dip. If you set that expectation on day one, a wobble is just part of the story. If you promise a straight line up, the first dip makes you look like you have failed, when in truth nothing has gone wrong at all.

The teams that get this right treat the two reports as a pair. The audit proves competence and hands the client something concrete. The tracking proves progress and keeps them around. One without the other is only half the conversation.

What even is a "share of voice", anyway?

There is one more thing I would add, and it applies to both reports equally. Keep them simple. SEO taught us how badly this can go wrong. The dashboards got more and more technical over the years, stuffed with metrics that made perfect sense to an analyst and washed straight over the client, and quite often over the sales rep trying to explain them too. A report loses most of its value the moment the person reading it stops following it.

"Share of voice" is the one that always gets me. It is an indirect, woolly metric that sounds meaningful and falls apart the second a client asks what it actually means. Share of what? Measured how? Against whom? If the person reading the report cannot tell you what a number means, that number is not doing any work for you.

So whatever you put in front of a client, and whatever brand is on it, boil it down. Simple, clear scores they actually grasp will beat twenty charts they nod along to and forget. That goes double for the AI visibility tracking, where the numbers are less intuitive to begin with. If a client can read the report on their own, without you sitting next to them translating it, you have done the job properly.

The short version

In SEO we had two jobs, fixing and measuring, and we never muddled them up. AI visibility is no different. There is the audit, which finds the controllable factors so you can fix them and prove it.

And there is the tracking, which measures how visible a business really is across the AI platforms, and which you present as a journey rather than a promise. Keep the two separate in your head, keep whatever you show the client simple, and you have a far clearer story to tell than "we bought an AI tool".

This two-sided problem happens to be exactly what we work on at Insites, both parts and fully white-label, if you ever want to take a look or book a demo.

Andrew

CEO, Insites

You can connect with me on LinkedIn.