AI Competitor Analysis: Benchmark

An AI competitor analysis measures how often AI systems such as ChatGPT, Perplexity and Google AI Overviews mention your brand compared to your competitors, expressed as share of voice. Across a fixed set of buying questions you count how often each brand is mentioned, then divide your mentions by the total in your category. This article explains how to set that measurement up systematically, what a healthy benchmark is, and how to turn the result into a plan that produces leads instead of a nice-looking number.

What is an AI competitor analysis?

An AI competitor analysis is a structured measurement of how often and how prominently AI models mention your brand versus your competitors, within a defined set of questions your buyers actually ask. The central metric is share of voice: the percentage of all relevant brand mentions in your category that go to you. If a model mentions your brand in 25 of the 100 answers that feature a brand from your category, your share of voice is 25 percent.

That differs from your standalone mention rate, the percentage of answers in which you show up at all. The mention rate tells you whether you are visible. Share of voice tells you whether you are winning. It is the AI counterpart of a top position in Google: it is not about presence alone, but about your share at the moment a buyer asks for a recommendation. To see how this logic fits the bigger picture, read our guide to Generative Engine Optimization.

Why is your own mention rate meaningless without competitors?

A mention rate in isolation gives you no decision framework at all, because you have no reference point to set it against. Say ChatGPT mentions your brand in 20 percent of the relevant answers. Is that good? That depends entirely on the rest of your category. If you are mentioned in 20 percent while your three biggest competitors together sit at 70 percent, you are losing the fight for the recommendation. If the model mentions you in 20 percent while nobody else gets above 10 percent, you are effectively the default choice.

That is why the competitor is the real yardstick, not an abstract target number. A benchmark without competitors measures movement without direction. You can see whether your number is rising, but not whether you are rising faster than the market. And in AI search results that distinction is sharp, because models often work with a short list of a handful of recommendations. Whoever falls outside that list is invisible to the buyer, however high the absolute mention rate may look.

How do you measure how often competitors are mentioned versus you?

You measure share of voice by running a fixed prompt set repeatedly through multiple AI models and counting the mentions per brand. The method consists of four steps that you repeat identically every measurement round, so your numbers stay comparable.

Build a prompt set around buying questions. Collect the questions your buyers genuinely ask in the consideration phase, not abstract keywords. Think of phrasings like “best fleet management software for transport companies in Belgium” or “alternative to supplier X”. These are the moments where a recommendation counts.
Fix your competitor set. Decide up front which brands you count. Usually these are your direct competitors plus the generic players the model often names, even if you do not see them as rivals yourself. It is precisely those unexpected names that reveal where AI slices your category differently than you do.
Run the prompts across multiple models. Ask the same questions in ChatGPT, Perplexity, Gemini and Google AI Overviews. The same brand can differ sharply in visibility from model to model, so an average across all engines hides where you win and lose. Repeat per model as well, because answers vary from run to run.
Count, normalise and calculate share of voice. Count the mentions per brand, optionally weighted by position (being named first weighs more than a passing mention). Divide your total by the total of all brands combined. That percentage is your share of voice per model and across the whole set.

You do not have to do this by hand. There is a growing supply of specialised measurement tools, which we cover in our overview of AI visibility tools. Whichever one you choose, the value sits in a consistent prompt set and a fixed measurement cadence, not in the tool itself.

There is no universally healthy percentage, because a realistic benchmark depends on your category, the AI model and your position in the market. There are, however, three anchors that help you set your own norm.

The first anchor is the market leader. In many categories the best-known player pulls a disproportionate share of the mentions towards itself, because models lean on repeated, broadly confirmed associations. As a newer or smaller player, your realistic goal is therefore rarely first place in the short term, but a solid second or third position within your real competitor set.

The second anchor is the model. Share of voice differs sharply per engine: the same brand can be far more visible in one AI environment than in another, and the sources each model cites often overlap only partially. So set a target per model instead of an average, and start with the engine where your buyers spend the most time.

The third anchor is time. AI answers are volatile and can shift within weeks. A healthy benchmark is therefore not a snapshot but a trend: measure on a fixed cadence, for example monthly for the full set and weekly for your most important buying questions. The question is not only “what is my number”, but “am I moving up faster than my competitors”. Whether your mentions are also carried by independent sources determines whether that position stays stable, as we explain in brand mentions over backlinks.

What mistakes do companies make with an AI competitor analysis?

The biggest mistake is measuring vanity: celebrating a high mention rate while the mentions have nothing to do with a buying decision. If you are frequently mentioned in informational questions but never in the comparative questions where a choice is made, your number looks good while you miss the leads. So always start with the prompts that sit close to the purchase.

A second mistake is changing the prompt set or the competitor set between measurement rounds. If you adjust questions or brands, you are measuring noise instead of progress. Lock your set down and change it deliberately and with documentation, not ad hoc. A third common mistake is throwing everything onto an average across models, which stops you from seeing in which engine you are falling behind. And finally: treating a one-off measurement as a conclusion. Given the volatility of AI answers, a single measurement round says little; only a trend across several rounds is reliable.

How do you translate a benchmark into action?

A benchmark is only valuable when it points to where improvement yields the most return, not as a scoreboard to show off with. The practical translation starts with the prompts where you lag behind your competitors and that at the same time sit close to a buying decision. That is your priority list.

For each such prompt, work out why the competitor gets mentioned and you do not. Often the cause lies in the underlying indicators of AI visibility: are you correctly understood within your category, and do enough independent sources confirm your story? We describe that mechanism in the 5 core indicators of AI visibility. You then work in a targeted way on content, positioning and external mentions around those specific buying questions. To see how to organise that as an ongoing process, read our complete GEO audit.

At Customer Impact we deliberately steer on leads and revenue, not on a high percentage in itself. A share of voice that rises on questions without buying intent looks nice in a report, but changes nothing about your pipeline. We are also honest about the limits: you rarely overtake an entrenched market leader in the short term, and nobody can guarantee a fixed position in AI answers. What is possible is winning ground structurally on the questions your buyers ask. That is exactly the work we deliver within our GEO optimisation.

The short summary

An AI competitor analysis measures your share of voice: how often AI models mention you versus your competitors on the questions your buyers ask. Your own mention rate only gains meaning next to that of your competitors. A healthy benchmark does not exist as a fixed number, it follows from your category, the specific model and your trend over time. And the measurement is only worth something once you translate it into targeted action on the buying questions where you lag behind.

Want to know how often AI mentions your brand versus your competitors, and where you can win ground? Schedule your free intake and we will map out your AI share of voice together.