For years, measuring search visibility was relatively simple. You chose a keyword, opened Google, and checked where your website appeared. SEO tools later automated the process and turned rankings into dashboards filled with positions, visibility scores, estimated traffic and historical trends.
AI search has made that model much harder to apply.
When somebody asks ChatGPT, Gemini or Perplexity to recommend a software product, an agency, a dentist, a hotel or almost any other business, there is no traditional search results page to monitor. The assistant searches for information, evaluates different sources and generates its own answer.
Your company might appear in that answer. It might be used as a source without being mentioned. Your website might even be consulted before the assistant decides to recommend one of your competitors instead.
This is the problem PerkFuel is trying to make easier to understand with its free AI Visibility Checker.
I spent some time looking at how the tool works, and what I find interesting is that it doesn't try to recreate Google rankings for AI. Instead, it focuses on a much more useful question: if someone who has never heard of your company asks an AI assistant for businesses like yours, will the assistant actually recommend you?
That sounds like a small distinction, but it changes the entire way AI visibility should be measured.
Knowing your brand exists is not the same as recommending it
One of the easiest mistakes to make when testing AI visibility manually is to include your brand in the question.
Imagine you run a dental practice called Northgate Dental and ask ChatGPT whether Northgate Dental is a good dentist. The answer may be detailed and positive, but the test is already biased because you have given the assistant the name of the company you wanted it to investigate.
A potential customer who has never heard of Northgate Dental would probably ask something much broader, such as "best private dentist in Leeds".
That is a far more meaningful question because the assistant now has to decide which businesses deserve to appear without being told which company to investigate first.
PerkFuel's checker follows this logic. You enter a domain and the system creates a non-branded customer-style query related to the website. It then examines the answer generated around that query.
This means the test is closer to genuine discovery.
The difference matters because brand awareness inside an AI model and commercial visibility inside an AI recommendation are not the same thing. A model can know exactly what your company does while still choosing not to mention it when a user asks for recommendations.
For businesses trying to understand GEO, that distinction is far more useful than simply checking whether ChatGPT knows the brand name.
The most interesting result is sometimes when you are not recommended
The checker does more than look for a company name inside an answer. It also examines whether the website was consulted while the response was being generated.
This creates an especially interesting situation: your website can be read even though your company is not ultimately recommended.
From a GEO perspective, this is probably one of the most useful signals the tool provides.
If an AI assistant never discovers your website, your visibility problem may begin with discoverability, relevance or authority. But if the assistant finds your website, reads it and still recommends three competitors, the problem is different.
At that point, the question is no longer simply "Can the AI find me?"
The better question becomes "Why did it find me and still choose someone else?"
That can lead to a much more interesting investigation.
Perhaps competing companies are supported by stronger third-party references. Perhaps they appear more frequently on comparison websites. Maybe their positioning is easier for the assistant to connect with the user's intent. It could also be that external reviews, directories, publications or other sources provide stronger evidence in favor of those businesses.
A single test cannot prove exactly why one company was preferred, but knowing that your site was considered and not selected dramatically narrows the problem.
This is also why I prefer this approach to a generic AI visibility score.
A score of 42 out of 100 might look impressive inside a dashboard, but it doesn't tell me whether an assistant couldn't find the company, found it but ignored it, cited it as a source or actively recommended it.
Those are completely different situations.
Seeing the sources changes the value of the test
The other part of the PerkFuel checker that deserves attention is the source analysis.
The tool doesn't only show whether your company appeared. It also lets you inspect pages that were consulted when the answer was created.
That is where the checker starts becoming more useful as a research tool.
Imagine an AI assistant recommends three competitors in your market. At first, all you know is that those companies won the recommendation.
Once you look at the sources, however, you may discover that the assistant used a competitor's website, an industry directory, a review platform and a comparison article before generating the answer.
That gives you much more information.
You are no longer looking only at the final recommendation. You are looking at part of the information environment behind it.
This is particularly important because GEO is unlikely to be won exclusively on your own website.
Traditional SEO encouraged businesses to think heavily about their own content, their own pages and their own technical optimization. Those things still matter, but generative search also exposes how important external information can be.
If AI assistants repeatedly consult certain industry directories, those directories become strategically interesting. If comparison articles consistently influence recommendations in your niche, being included in those articles becomes more valuable. If your competitors are reinforced by reviews and independent publications while your company exists mostly on its own website, that difference may eventually influence how assistants describe the market.
The checker therefore gives you something that is potentially more useful than a visibility score: clues about where the assistant is getting its information.
For anyone working seriously on GEO, those clues can become the beginning of an actual strategy.
Competitor analysis becomes much more useful
This also changes how I would approach competitors.
If an AI assistant recommends three companies instead of mine, my first reaction would not necessarily be to publish another ten blog posts.
I would first investigate why those companies keep appearing.
I would look at the external websites mentioning them, the language used to describe their products, the comparison pages where they appear and the sources that assistants repeatedly rely on.
If one competitor is mentioned across several respected websites and another has hundreds of detailed reviews, that tells me something about the evidence available to the AI.
The goal is not to copy those competitors. The goal is to understand why the broader web makes their recommendation easy to justify.
This is an important shift in mindset.
In traditional SEO, competitive analysis often starts with keywords and backlinks. In GEO, competitive analysis may increasingly involve understanding how a brand is represented across the collection of sources that AI systems use when constructing answers.
That is why being able to inspect those sources is more interesting than simply seeing that your competitor was mentioned.
One AI answer is not enough
There is an obvious limitation to any tool attempting to measure AI visibility: generative answers are not perfectly stable.
You can ask the same question twice and receive slightly different responses. A small change in wording can affect the companies mentioned. ChatGPT may also produce a different recommendation from Gemini or Perplexity.
This means one successful answer should never be interpreted as proof that a brand dominates AI search.
PerkFuel handles this reasonably well by expanding the test beyond the first result.
Its full report currently compares ChatGPT, Gemini and Perplexity using two different customer-style questions. That creates six separate answers to examine.
Six responses obviously do not represent every possible AI search around a business, but they are much more informative than a single carefully chosen prompt.
If your company appears in five of six answers, that begins to look like a pattern. If it appears only once, your visibility is probably much less stable. If ChatGPT regularly mentions you while Gemini does not, that difference itself becomes worth investigating.
The value is not in pretending those six answers form a permanent ranking. The value is in using multiple observations to avoid drawing conclusions from one lucky result.
That is exactly how I think AI visibility measurement should develop.
Instead of pretending there is a single fixed position, the goal should be to identify patterns across assistants, prompts and time.
Citation and recommendation should not be treated as the same thing
One of the most useful ideas behind the checker is that it separates being used as information from being presented to the user as a recommendation.
These two events are often mixed together when people talk about GEO.
A marketer might say that a company is "visible in ChatGPT" because one of its articles was cited. Technically, that is a form of visibility, but it does not necessarily mean the brand itself was recommended.
Imagine someone asks for the best tools in a category.
ChatGPT reads one of your articles while preparing the answer, but ultimately recommends three competing products and never mentions yours.
Your content played a role in the research process, but your company did not receive the commercial visibility.
That difference is important.
Being cited indicates that your content entered the assistant's information environment. Being recommended means your brand reached the final answer shown to the potential customer.
For businesses thinking about AI as an acquisition channel, the second event is usually far more valuable.
PerkFuel makes that distinction visible instead of merging everything into one metric.
The tool also has useful limitations
I actually think it is important not to oversell what an AI visibility checker can do.
No tool can currently give you a perfectly complete picture of how every user will see your brand across every possible prompt, model and location.
Two customer questions are still only two questions. Six answers are still only six answers. AI responses can change over time, and the sources used today may not be exactly the same tomorrow.
The fact that a page appears among the sources also does not prove that one specific sentence on that page directly caused the assistant to recommend or reject a business.
There is still uncertainty.
But that is exactly why I like the way PerkFuel presents the test.
The useful output isn't a claim that your company has an objective AI visibility score of 78.4%.
The useful output is evidence.
You see the question that was tested, you see which businesses appeared, you see whether your website was consulted and you see some of the sources involved.
You can then investigate further.
For a field as new and unstable as GEO, I think that is much more valuable than false precision.
Where this fits into a real GEO workflow
The way I would use the tool is not as a replacement for analytics, SEO platforms or manual research.
I would use it near the beginning of a GEO audit.
First, I would run the website through the PerkFuel AI Visibility Checker and look at the customer question the system generated.
Then I would study the businesses that appeared and the sources behind the answer.
If my company was read but not recommended, I would investigate what external evidence supports the companies that were chosen.
If the company was not discovered at all, I would look more closely at its content, entity clarity, relevance and external presence.
After that, I would repeat the process across more questions and more assistants before deciding that a particular pattern is real.
This is where the tool becomes valuable.
It doesn't tell you exactly how to fix GEO.
It tells you where the investigation should begin.
Why I think this is a genuinely useful tool
AI search is creating a measurement problem that SEO tools were never designed to solve.
There isn't always a stable ranking position. The same query can produce different answers. Assistants can use your content without mentioning your business. They can discover your company and still decide that competitors are better recommendations.
Trying to compress all of that into a single score misses most of what makes generative search different.
PerkFuel takes a simpler approach.
It tests a realistic customer question, examines the answer, looks at whether your business was actually selected and exposes some of the sources that contributed to the result.
That doesn't solve every GEO measurement problem, but it turns something extremely abstract into something you can actually inspect.
And that is why I think the checker is genuinely useful.
If you work in SEO, GEO, SaaS, local search or digital marketing, I would recommend testing your own website and then testing a competitor immediately afterwards.
The free checker is available here:
https://perkfuel.io/ai-visibility-check
Don't just look at whether your company appears.
Look at who appears instead, which sources the AI consulted and whether your website was discovered but ultimately passed over.
That is where the useful information begins.
Top comments (0)