Minel Gunesoglu is the founder of Is My Brand in AI and Vouch. Is My Brand in AI is a research and consulting practice that measures whether AI answer engines name a company when buyers ask for recommendations, and works with that company on what to change. Vouch scores AI tools and publishes each score next to the evidence it rests on.
Both are built on measurement that is published rather than asserted. The AI Search Evidence Index asked whether pages ranking for AI-visibility queries can support their own numbers. The Claude Retrieval and Citation Study measured how far an AI answer moves between identical runs of the same question. Both datasets are public with permanent DOIs at CERN’s Zenodo, released under CC BY 4.0, alongside a written record of the defects found in the measurement instrument and how they were corrected.
Company: Is My Brand in AI
We are thrilled to have you join us today, welcome to ValiantCEO Magazine’s exclusive interview! Let’s start off with a little introduction. Tell our readers a bit about yourself and your company.
Minel Gunesoglu: I am Minel Gunesoglu. I run two companies that price in opposite directions, and the reason they do is the same body of research.
Is My Brand in AI is the research and consulting side. When a buyer asks ChatGPT, Perplexity, Gemini or Google’s AI Overviews which company to use, we measure whether a brand is named, which of its pages the engine actually retrieved, and what would have to change for that to improve. Vouch is the product side. It scores AI tools and publishes every score next to the evidence it rests on, with the source document saved rather than merely linked, so the number can still be checked after the vendor’s page has changed.
Neither business would exist in its current form without the other, and that is the part I find worth talking about. The research tells us what actually decides whether a page gets named in an AI answer, which is the exact thing a review desk has to get right. Running the desk shows us what buyers ask before they spend money, which becomes the next research question and then the next consulting engagement. It is one ecosystem with two revenue models, not two companies that happen to share a founder.
The work underneath is published rather than asserted. The AI Search Evidence Index asked whether pages that rank for AI-visibility queries can support their own numbers. The Cross-Engine Citation Study put the same buyer questions to four engines in two rounds seven weeks apart. The Claude Retrieval and Citation Study logged, across 150 runs, what a model’s search retrieved separately from what it chose to cite. All three datasets are public under CC BY 4.0 with permanent DOIs at CERN’s Zenodo, and each ships with a written record of the defects we found in our own instrument and how we corrected them.
What was the pivotal pricing insight or experiment that most dramatically improved your company’s growth or profitability?
Minel Gunesoglu: The insight was that trust and outcome have opposite pricing properties, and that one ecosystem can sell both if it prices them in opposite directions. The growth came from acting on it in the checkout page rather than in the about page.
Vouch sells trust. A verdict stops being worth anything the moment somebody can buy it. That is not a slogan, it is the economics of the category: the instant a reader suspects the ranking is purchasable, the ranking has no informational value left, and there is nothing to sell. So the verdict itself is the one thing on that site that never appears on a price list.
Here is what that looks like as an actual price list, because a principle that never reaches a checkout page is only a sentence. Submitting a tool for review costs nothing and always will. The free queue runs in submission order and the backlog currently sits at around three months. Fifty dollars, once, moves a tool to the front and it is measured within seven days. Ninety-nine dollars a month buys a slot labelled Sponsored everywhere it appears, held outside the scored category listings. None of the three touches the score, the position, or the top pick badge, and we say so in one line on the page: you pay for when, not for what.
The counterintuitive part, and the reason this improved profitability rather than costing us revenue, is that refusing to sell the verdict is what created the two things we can sell. Speed is only worth fifty dollars because the free queue is genuinely slow and genuinely honest, so a three month backlog is not a capacity failure, it is the asset that gives the expedited tier its price. A sponsored slot is only worth ninety-nine dollars a month because every reader knows it cannot buy a score, which is exactly why it can be sold at all. Both prices are parasitic on the refusal. Remove the refusal and both products lose the reason anyone would pay for them.
We made the refusal structural rather than rhetorical, because a promise in the terms of service is only as durable as the person who wrote it. In our codebase the sponsors file is never imported by the scoring module. A paid record has no path to the ranking function. Not by accident, not by a developer working quickly on a Friday, not by me in a month when revenue is short.
Is My Brand in AI sells outcome, which has the opposite property. Outcome cannot be priced in advance because it does not exist until the work is done, so the fee is tied to measured movement against a baseline both sides agree before the work begins. The research told us this was right rather than merely clever: the Claude Retrieval and Citation Study showed that two runs of an identical question overlap only 0.555 in the sources they return. If a single reading is that unstable, a monthly snapshot is the wrong unit to bill against, and a firm charging for one is charging for noise.
Same principle, two directions, two revenue lines that do not rise and fall together. Never charge for the thing itself. Charge for access to it, or for its effect.
How do you determine the true value your product or service delivers to customers, and how does that shape your pricing strategy?
Minel Gunesoglu: We measure it, and the measurement is itself the product. That sounds circular, so let me give the number that made it concrete and then show where it lands on the price list.
The AI Search Evidence Index started from a question about our own market. Pages that rank for AI-visibility queries are full of statistics. Can a reader who wants to check one actually reach its source? We sampled 38 such pages and scored the 13 that carried enough numeric claims to rate against a fixed instrument. On the median page, a reader could reach the source of roughly one figure in three.
The part that shaped our pricing was not that finding. It was what the finding cost. Collecting those 38 pages took under two minutes. Checking them took 5,706 model calls across a week, and that week contained seven revisions of the instrument, blind coding where the machine’s verdicts stayed sealed in a separate file until a human had finished reading, an independent re-implementation written without reusing the original code, and hand verification of every sourced block. We also sampled our own instrument and found that 22.4 percent of what it had scored as a factual claim was not a claim at all. That correction is published with the data.
So the value is an asymmetry. Asserting is nearly free. Verifying is expensive. Anyone can publish a number in two minutes, and almost nobody will spend a week establishing whether it holds. That gap is the commercial position of both companies, and once you see it the pricing writes itself.
It tells you what to charge for and, just as usefully, what to refuse to charge for. Because verification is the expensive part, the queue for it is genuinely long, currently around three months. That backlog is not an embarrassment to be hidden, it is the thing that makes fifty dollars for a seven day review a fair and obvious trade. The customer is not buying a better outcome, they are buying an earlier one, and both sides can see exactly which is which.
On the consulting side the same logic says we are not selling hours of attention, we are selling the ability to tell movement from noise. That is what the Cross-Engine Citation Study established. We put the same twelve buyer questions to four engines in two rounds seven weeks apart, and roughly half the sources one engine had cited in June were gone by August. A client paying monthly for a dashboard is paying to watch that churn. A client paying for movement against an agreed baseline is paying for the only part of the data that means anything.
Can you share one pricing decision (such as packaging, tiering, or a bold price increase) that felt risky at the time but ultimately paid off significantly?
Minel Gunesoglu: Tying the consulting fee to measured movement instead of to a calendar.
A retainer quietly moves risk onto the client. They pay in month two whether or not anything changed in month one, and the agency’s incentive is to keep the relationship alive rather than to finish the job. We inverted it. Before any work starts we agree a baseline: which buyer questions we will run, against which engines, how many times each, and what counts as movement. Then we measure against that baseline and the fee follows what actually moved.
It felt risky for the obvious reason. It means carrying a cost the industry has arranged not to carry, and it means some months are lean by design. It also removes the comfortable position where you can be busy without being effective.
The reason it works is not courage. It is that we can measure well enough to know the difference between movement and noise, and that came from the research rather than from confidence. The Cross-Engine Citation Study is what gave us the nerve. We put the same twelve buyer questions to ChatGPT, Perplexity, Gemini and Google’s AI Overviews, then repeated the whole exercise seven weeks later. Roughly half the sources one engine had cited in June were gone by August. If you do not know that, a client’s dashboard moving looks like your work. Once you do know it, you understand that most of what a monthly report shows is the market breathing, and you stop billing for it.
There is a version of this idea that I want to warn other founders away from, because it sounds similar and is not. Promising to charge nothing unless a result appears is a marketing line, not a pricing model, and it hands the client a document that will be read back to you in a bad quarter. What we do is narrower and more defensible: a stated measurement, agreed in advance by both sides, that determines the fee. The client knows exactly what will be measured before a single hour is spent. So do we.
The second thing that makes it survivable is the ecosystem. Vouch generates revenue on a completely different mechanism, on the product side, and the two are not correlated. A consulting model with variable months is only sane if something else in the business is not variable in the same way. Founders who copy the fee structure without having that second leg tend to discover the problem in month four.
A firm that cannot measure its own effect cannot price this way, and honestly should not try. The pricing model is downstream of the measurement. Not the other way round.
How do you continuously test and refine pricing without alienating existing customers or damaging brand perception?
Minel Gunesoglu: By keeping the thing that cannot be bought completely outside the pricing conversation, so that changing a price never touches the product’s credibility.
On Vouch the score is not a variable. It is not a discount lever, it is not a package tier, it is not something a sales conversation can reach. That single decision is what makes the rest of the pricing testable in practice. We can move the fifty dollar expedited fee, we can change what the ninety-nine dollar sponsored slot includes, we can add a tier or retire one, and not a single reader has to wonder whether a verdict moved with it. The perception risk that normally accompanies a price change simply does not attach, because the price was never anywhere near the thing readers trust.
Most brand damage from pricing does not come from the size of an increase. It comes from a customer discovering that the thing they valued was negotiable all along. If your product’s credibility sits on the price list, every experiment you run is an experiment on your reputation, and you will eventually stop running them. That is how companies end up frozen on a price they set in year one.
On the consulting side the baseline does the same work. Both parties write down what will be measured before the engagement starts, so when the fee structure changes later we are changing a stated, shared number rather than reopening the question of what the relationship is worth. It turns a negotiation into an arithmetic conversation, which is a much better conversation to have with someone who already trusts you.
For existing clients specifically, the rule is that nobody learns about a change from an invoice. Anything new applies from a renewal both sides can see coming, and the old terms run to their end. That costs money in the short term and is the cheapest reputation insurance available.
The research feeds this too, and it is the part people miss. The AI Search Evidence Index and the Claude Retrieval and Citation Study are both published under CC BY 4.0 with permanent DOIs at CERN’s Zenodo, including a written record of the defects we found in our own instrument and how we corrected them. Publishing your own errors sounds like a strange way to protect a brand. It is the opposite. Once a market has watched you correct yourself in public with the data attached, a price change reads as a business decision rather than as a sign that something is being quietly rearranged behind the curtain. Credibility built on published evidence is unusually resistant to commercial noise, because the evidence does not move when the price does.
For other CEOs who still treat pricing as an afterthought, what’s the single most important mindset shift or first step you would recommend they take?
Minel Gunesoglu: Ask what you are actually selling, then ask whether that thing survives being priced. Most pricing damage I see comes from a company charging for the one item in its business that should never have carried a price.
Some things survive a price list without losing anything. Access survives it. Speed survives it. Capacity, support, implementation, effect, all of them can sit on a page with a number next to them and remain exactly as valuable as they were. Other things do not. A verdict, an audit, an independent assessment, a recommendation: these lose the property that made them valuable the moment a buyer can influence them. Price those and you have not monetised an asset, you have destroyed one and collected a fee on the way out.
The first step costs nothing and takes an afternoon. Write down every distinct thing you sell. Next to each one, write what happens to its value if a customer could pay to change it. Anything that becomes less valuable under that test belongs permanently outside the price list. Then price everything that touches it, generously, because that is where your revenue actually lives.
The second step is harder and it is the one I would push hardest. Whatever you decide belongs outside the price list, put the decision somewhere it cannot be quietly reversed. A promise in your terms of service is only as durable as the person who wrote it and the quarter they are having. We put ours in the codebase: the sponsors file is never imported by the scoring module, so a paid record has no path to the ranking function. Reversing that is a visible act that leaves a trace. Rewriting a paragraph in the terms is not.
The third thing, and the one founders resist most, is that pricing is downstream of measurement. You cannot price an outcome you cannot measure, and if you try, you will end up charging for time because time is the only thing you can count. Both of our pricing models came out of studies rather than out of a strategy session. The AI Search Evidence Index told us that verifying costs vastly more than asserting, which is why a verdict nobody bought is the scarce asset. The Claude Retrieval and Citation Study told us that two runs of an identical question overlap only 0.555, which is why a monthly snapshot is the wrong unit to bill against. Neither number came from thinking harder about pricing. They came from measuring the market we sell into, and the pricing followed.
So the mindset shift is this. Pricing is not a commercial decision you make after the product exists. It is a statement about what you believe your product is, and the market will eventually test whether you meant it. Decide what cannot be bought, build the structure that makes that permanent, measure what you actually deliver, and let the price follow the measurement. Do it in that order and pricing stops being the thing you dread reopening every year.
