Losing revenue to competitors? Check what AI tells your buyers

The signs: when the problem is not your ads

AI search does not explain every lost deal. But some patterns point to it. Check if two or more of these are true for your business:

• Fewer inbound leads, same ad performance. Clicks and cost per click look normal, but fewer people contact you on their own.
• New names in lost deals. Buyers mention competitors that you rarely met before, often smaller ones.
• Buyers arrive with a shortlist. They already compared three providers before the first call, and you are the one who must prove you belong on the list.
• "I asked ChatGPT." When you ask new customers how they found you, AI tools appear in the answers. Or, more important, lost buyers say it.
• Revenue is down over the last one or two years and the usual explanations (season, price, market) do not explain all of it.

None of these signs proves that AI is the cause. Together, they tell you that it is worth checking.

Why spending more on ads does not fix it

The old path to a purchase was: the buyer searches Google, opens a few websites, compares, and decides. Your ads, your SEO and your landing pages work on that path.

One path that is becoming more common is shorter: the buyer asks an AI model, reads one answer with a few recommendations, and decides which brands to look at. The AI model makes the first selection. The buyer may never see the search results page where your ad sits.

And many buyers do not check the sources behind the answer. They read the recommendation, take it as trustworthy, and move on. For these buyers, the AI answer is the shortlist. If your brand is not in it, they may never see the pages that prove you are a good fit.

So the question is not "should we spend more on ads?". Ads still work for the buyers who search. The question is: what happens to the buyers who ask?

What our data shows

Between 7 August and 19 September 2026, we collected 4,010 answers from ChatGPT, Gemini, Perplexity and Google AI Overviews. The answers replied to buyer questions in six markets, from CRM software to adventure travel. In each market, we tracked one real brand, which we do not name. For every answer, we checked one thing: did the AI model name that brand?

What the 18.8% measures. It is a brand-mention rate across our research sample. It is not a measure of lost customers or lost revenue. It shows how often AI named the tracked brand in the answers we collected. Your results depend on your category, market, buyer questions and AI models.

On average, the tracked brand was not named in about four of five answers. The result was very different from market to market. In two of the six markets, AI named the tracked brand in more than half of the answers. In the other four markets, it named the tracked brand in only about 2% to 11% of the answers. Same type of question, same AI models, very different results.

The sources differ too. For the claim analysis, we used the 567 answers from ChatGPT, Gemini and Google AI Overviews that contained claims about the brand. These answers cited 952 distinct websites (domains). 83.8% of those websites were cited by only one of the three AI models. A buyer who asks ChatGPT and a buyer who asks Gemini can get a different shortlist, built from different pages.

A note on these numbers. These figures come from six markets and one tracked brand per market. They describe this sample, not all brands. The full method, the dates and the limits are in our analysis of 4,010 AI answers.

What this looks like for your business

Imagine that a buyer asks three AI models questions like these:

Your business might appear in one answer, be missing from another, and be described wrongly in a third. One answer cannot tell you if this is a pattern. You need to see which brands appear across repeated runs, which sources the AI models cite, and how the results differ by model.

An AI visibility report should show you:

• Where you are missing: buyer questions where competitors appear and you do not.
• Why it may happen: the pages and sources behind the recommendations.
• What to check first: gaps in your content, in third-party mentions or in your brand facts.
• Whether anything changes: how the results move over time.

The goal is not to chase every AI mention. It is to find the gaps that matter to your buyers, and to decide what to do about them.

What happens next? Add your brand and buyer questions, then review the results for the AI models you track.

AI visibility is something you can work on

Being in an AI answer is not luck. AI models recommend the brands they can find, understand and trust. That depends on things a company can work on: clear facts on its own pages, comparison pages, reviews, and mentions on the websites that AI models cite.

This work has a name: GEO (generative engine optimization) and AEO (answer engine optimization). Any company in your market can do this work, and some may already do it. If competitors improve their presence in the sources AI relies on while you do nothing, your relative visibility may fall. Measure the gap. Do not assume it.

A 15-minute self-check

You can test this today, without a tool. It is a quick diagnostic, not a statistically sound measurement:

1. 1
   Write five buyer questions

   Use the words your buyers use, not your product names. For example: "Which accounting software is best for a small business in the UK?"

2. 2
   Ask them in three AI models

   For example ChatGPT, Gemini and Perplexity. Use a private window, so your own history does not change the answer.

3. 3
   Ask each question twice

   AI answers change from one run to the next. One answer is a coin toss, not a measurement.

4. 4
   Write down every brand named

   Are you there? In which position? Which competitors appear again and again?

5. 5
   Look at the sources

   Which pages does the answer cite? Those are the pages that shape your buyers' shortlist, even if your buyers never open them.

If your brand is missing from the answers your buyers rely on, or if AI describes your offer incorrectly, you may have found an AI visibility gap that advertising alone will not fix. Track the same questions over time to understand whether the pattern persists.

Two layers, four steps

The Vosapio approach has two parts that work together. First, it splits AI visibility into two layers: Live and Memory. Second, it runs the same four steps on both layers: See, Understand, Act, Measure. Together, they show where you lose, why you lose, and what to fix first.

1Live vs Memory: two ways AI sees your brand

When AI answers a buyer, it relies on two layers: what it finds on the web right now, and what it already knows from training. Your competitors can win in one layer and lose in the other. They change for different reasons, so Vosapio measures them as two separate metrics: Live visibility and Memory visibility.

Why this matters when you lose deals: the fix depends on the layer. If competitors win the Live answers, the cause is usually in the pages that AI cites today, and you can work on those pages now. If competitors win the Memory answers, the model already knows them better than you. That fix is long-term: state the same facts clearly and in many trusted places, so that future models learn them.

A note on the two layers. One answer can use both layers, so not every answer can be attributed definitively to one or the other. Treat the split as a strong signal of where to look first, not as an exact cause.

A note on recommendations. In some cases, AI answers a recommendation question mostly from Memory, rather than by searching the web. So a strong website alone does not guarantee that AI recommends you.

2See, Understand, Act, Measure: on both layers

For each layer, Vosapio uses the same four steps:

1. 1
   See

   Find the buyer questions where AI recommends competitors and not you, in each AI model, in Live answers and in Memory answers.

2. 2
   Understand

   Find out why. For Live answers: which pages does AI cite, and what do they say about your competitors that they do not say about you? For Memory answers: what does the model already believe about you and about your competitors?

3. 3
   Act

   Investigate and address the likely cause. For Live: improve your own pages, and work to get mentioned on the websites that the answers cite. For Memory: state the same facts consistently, in many trusted places, so that future models learn them. No single fix guarantees a result.

4. 4
   Measure

   Check again every week. Live visibility can move when pages and cited sources change, but the timing and the impact differ by AI model and question. Memory visibility changes more slowly, with new model versions.

Being named is only one step. We measure AI visibility as a ladder of five steps: read, cited, mentioned, recommended, and described correctly. A brand that AI names but describes wrong can still lose the deal.

Vosapio runs these steps for you: it asks your buyer questions in each AI model on a schedule, shows which competitors win each answer, measures Live visibility and Memory visibility as two separate metrics, and keeps the sources behind every answer. See how Vosapio tracks AI visibility, or compare the plans for brands and the plans for agencies.

For brands

You know your buyers and your products better than any AI model does. Use that. Run the check on the questions your buyers ask just before they buy: which products AI recommends, which ones it leaves out, and which facts it gets wrong. If you sell ten products but AI recommends only seven, this is where you find out why. Start with the questions closest to a purchase, and fix the pages behind them first.

For agencies

Your clients ask why leads are down when the campaigns perform. AI search is one part of the buyer journey that traditional campaign reports may not capture. Run the same check for each client: show which questions they lose, to which competitors, in which AI model, and which pages to fix first. It turns a difficult client call into a plan.