I Stopped Asking “Does AI Know My Brand?” and Started Measuring Something Harder

Devanand Kushwaha — Thu, 08 Oct 2026 07:03:34 +0000

I thought AI search visibility would be a simple number.

Something like:

Your brand appears in 38% of AI answers.

Easy.

Put it on a dashboard. Track it every week. Done.

Then I started building Aurvikon.

And I realized the percentage isn't the interesting part.

The interesting question is:

Why did the AI choose them instead of you?

That question changed how I started looking at AI search.

I stopped testing branded searches

The easiest test is to ask:

“Tell me about [company name].”

Of course the company is likely to appear.

That's not very useful.

A real buyer might never mention your company.

They might ask:

“What's the best CRM for a 10-person sales team?”

Or:

“What are good alternatives to Salesforce for a small company?”

Or:

“Which tools are best for automating outbound sales?”

That's where things get interesting.

So I started testing category questions instead of brand-name questions.

And I started seeing a very different picture.

A company can be visible without really being competitive

Imagine an AI system answers a category question and mentions your company.

At first glance:

Great. We're visible.

But then you look closer.

Your competitor was mentioned first.

Another competitor was recommended for the exact use case you target.

Your company was mentioned briefly but wasn't supported by any strong source.

And in another AI system, you weren't mentioned at all.

So what does “38% visibility” actually tell you?

Not much on its own.

That's when I started breaking visibility into smaller pieces.

Are you mentioned?

The most basic question.

Did the AI include your company in the answer?

Are you relevant?

Being mentioned isn't enough.

Was your company actually recommended for the problem you solve?

Who gets mentioned instead?

This is probably one of the most useful signals.

Because the real business question isn't:

“Does AI know us?”

It's:

“Who is getting the buyer that we wanted?”

What sources are influencing the answer?

This is where things get even more interesting.

The answer may be influenced by:

comparison pages
customer reviews
industry articles
directories
documentation
community discussions
other third-party sources

So your website isn't necessarily the entire story.

This created a weird problem

I started noticing situations where a company could have:

a decent website

good traditional SEO

lots of content

and still lose the AI recommendation.

Not because the product was necessarily worse.

But because the AI had a different picture of the category.

Maybe competitors had more third-party coverage.

Maybe there were more comparison pages mentioning them.

Maybe reviews were stronger.

Maybe the company's positioning was inconsistent across the web.

Maybe the AI simply had more evidence connecting the competitor with the use case.

That's a much harder problem than “rank higher.”

That's why I'm building Aurvikon

The idea behind Aurvikon became much simpler:

Ask the questions buyers actually ask.

Then measure what happens.

Which brands appear?

Which competitors appear?

What sources are cited?

Where does your company disappear?

Which questions are the biggest opportunities?

Instead of giving a company another generic SEO checklist, I want to show the actual discovery problem.

Something like:

Question: What are the best tools for automating outbound sales?

Then:

AI mentions: Apollo, Outreach, Hunter

Your brand: Not mentioned

Top cited sources: X, Y, Z

Biggest opportunity: Specific use case where competitors dominate

That's much more actionable than:

“Your AI visibility score is 42.”

I'm still figuring out the hard part

This is probably the part I find most interesting.

I don't think the industry has completely figured out what a good AI visibility measurement should look like yet.

I'm still testing questions like:

How many prompts are enough to make the measurement meaningful?

Should every AI engine have the same weight?

How much should citations matter?

Should a mention in position #1 count very differently from position #5?

How do you separate a random mention from an actual recommendation?

How stable are these answers over time?

And probably the biggest one:

Can an AI visibility score explain what to do next, or is it just another dashboard number?

I don't want Aurvikon to become another tool that gives you a colorful score and leaves you wondering what it means.

The useful part should be the explanation.

The metric isn't the product

That's the lesson I'm currently learning.

A number is interesting.

A reason is useful.

If an AI doesn't mention your company, I want to know why.

If it mentions your competitor, I want to know what evidence is behind that recommendation.

And if there's an opportunity to change the situation, I want the product to tell you where to start.

That's the problem I'm working on now.

I'm still early.

The methodology is still evolving.

And honestly, that's the fun part.

We're watching a new layer of search appear while we're still figuring out how to measure it.

If you're building SaaS, try one experiment today:

Ask an AI:

“What are the best [your category] tools for [your target customer]?”

Don't mention your company.

Then look at who gets recommended.

You might learn more from that answer than from another week of staring at your Google rankings.

How to check if ChatGPT recommends your brand (the method we use)

Devanand Kushwaha — Wed, 07 Oct 2026 16:28:32 +0000

Ask ChatGPT "What is [your product]?" and it will probably find your website and describe you nicely. That proves one thing: it can find you.

Your buyers don't ask that. They ask "what's the best tool for…" or "which one should I use for…". The real question is whether you're named in that answer.

Here's the manual method we use to get a real number, plus the mistakes that quietly inflate it.

The method

Write 20–30 real buyer questions. Mix category questions ("best accounting software for freelancers"), problem questions ("how do I stop late client payments") and comparisons ("X vs Y"). Leave out anything that contains your own brand name.
Use a clean session. Logged out, or a temporary chat with memory turned off. Personalisation and chat history change the answer.
Make sure web search is on. Recommendations for current products come from live search results, which is what most buyers see, not from the model's training memory.
Ask each question at least 3 times, ideally on different days. Answers vary from one run to the next, so one answer proves very little.
If you sell in several countries, ask from each one (or put the country in the question). Answers and sources differ by market.
Record four things for every answer: were you named (yes/no), your position in any list, which brands were named instead, and which websites were cited.

Turning the answers into numbers

Visibility = answers that name you ÷ all answers. Named in 6 of 60 answers = 10%.
Share of voice = answers naming you ÷ answers naming any brand you track.
Position = where you appear when you are named (first, third, last).
Sources = the websites cited most often. These are the pages the AI trusts for your category, and the places you most need to be mentioned.

Keep the question list fixed and re-run it every week. A change only means something if you measured the same questions the same way.

Common mistakes

Checking once and treating it as the truth. Run it several times.
Counting questions that contain your brand name. You will always "win" those, which inflates the number.
Checking only ChatGPT. Claude, Gemini, Perplexity and Google's AI Overviews often name different brands and cite different sources.
Reading the answer but ignoring the sources. The cited pages tell you what to fix.

FAQ

Does ChatGPT give everyone the same answer?
No. Answers vary between runs, users, countries, and whether web search is used. That's why you ask each question several times and keep the method fixed.

How often should I check?
Weekly is enough to see a trend by hand. Daily checks help once you're changing things and want to see the effect sooner.

Why does ChatGPT name my competitor instead of me?
Usually because the pages it reads for that question (lists, reviews, comparisons, forums) mention your competitor and not you. The cited sources show you which pages those are.

Doing it automatically

30 questions × 3 runs × 2 countries is 180 answers a week by hand. That's why we built Aurvikon: it asks your buyer questions every day with live web search, records who is named and which pages are cited, and turns the cited pages into a to-do list.

*Originally published at aurvikon.com