How to Build a Competitor Benchmark in AI Search

• An AI Visibility score alone isn’t enough. What counts is the context and the brand’s position against competitors in its niche.
• Your competitors in AI Search may differ from the ones you know from industry reports and SEO work.
• An AI Search benchmark should cover the entire purchase journey. An averaged score can hide problems at a specific decision stage.
• Don’t base the benchmark on a single measurement. An analysis of more than 45,000 citations found that only 30% of brands stayed visible when the same prompt was run again.
• A benchmark should lead to specific actions. The KPIs and findings you report should show where competitors have pulled ahead, why and what you can do to narrow the gap.

A company gets a report from its agency saying its AI Visibility in its niche is 40%. Which raises the question: is that a good result or not? The answer depends on what its direct competitors are scoring.

Let’s look at a concrete example:

Our brand’s AI Visibility sits at the 40% mentioned above. That puts it second in the category, which is a solid result. It also has a comfortable lead over the other competitors, but at the same time the niche is clearly dominated by a single leader, and the brand trails it by as much as 28 percentage points.

AI Visibility on its own shows the brand’s score, but only a competitor benchmark adds business value and tells you what to do next.

A benchmark should come in two forms: open and controlled.

Open Benchmark

Here you monitor prompts without suggesting any brands. These are general queries, such as:

• Which e-commerce SEO agencies are worth considering in Poland?
• Which Polish suit brands are worth a look?
• Which running shoe brands are the best?

This shows whom AI recommends in its answers and lets you identify the brand’s real competitors in AI Search for its niche.

Controlled Benchmark

The next stage is monitoring visibility with a focus on specific brands:

• your own brand
• your 3 to 5 most important competitors
• possibly 1 or 2 brands discovered during the first stage of the benchmark

You then track this same group of brands across specific prompts over a longer period.

A benchmark should rest on a broader view of AI visibility than a single prompt. It works the same way as keywords on Google: ranking well for one phrase doesn’t mean the brand is also visible for other industry queries.

A prompt like “best SEO agency” is too general to judge a whole category by. For a benchmark, you need a topic cluster of prompts that reflects a potential customer’s real purchase or decision journey.

You can group the monitored prompts by stage:

E-commerce businesses can also add constraints tied to product features to the monitored prompts, for example:

• a product under $150
• a product for beginners
• a product in a specific color
• a product delivered by Friday

Why Is Splitting by Intent So Important?

An averaged AI Visibility score can hide a real visibility problem at one of the stages of the customer’s purchase journey.

For example:

The benchmark should also be split by the LLMs you analyze. Strong visibility in ChatGPT answers doesn’t automatically mean the brand will score similarly in Gemini. Even if the benchmark reports an averaged AI Visibility metric against competitors, the detailed data should include a breakdown by platform.

A one-off visibility check for a handful of prompts isn’t enough to draw conclusions in a benchmark. An AirOps analysis of more than 45,000 brand citations in AI answers found that only 30% of brands stayed visible when the same prompt was run again later. So a single result may be a fluke rather than a stable position for the brand.

That’s why a benchmark should be based on:

• a fixed set of prompts, measured repeatedly
• the same language
• the same market
• the same methodology

To keep the benchmark dashboard readable, you don’t need 30 metrics. A few KPIs that answer specific questions are enough:

The whole benchmark should roll up into one main KPI: the AI Competitor Gap, which shows the difference between the category leader’s score and your brand’s score. Tracking it over time gives you a starting point and lets you check how well your AISO strategy is working.

Once you’ve compared the overall AI Search visibility of your brand and its competitors, it’s time to single out the areas where the brand is falling behind.

Prompt Gap

The first step is the Prompt Gap: checking which queries trigger answers that include your competitors but not you.

Example:

A comparison like this shows right away which customer needs your competitors are winning visibility for in AI Search.

Source Gap: Why Does AI Trust Your Competitors?

You now know the areas where a competitor wins in AI. The next step is to find out why. To do that, look at the online sources AI bases its answers on. It may turn out that the LLM gets its information about a competitor from:

• the brand’s YouTube videos
• industry articles
• rankings
• review sites

Entity Gap: What Does AI Associate Your Competitor with More Strongly?

The third angle is the Entity Gap: checking what AI associates your brand with and what it associates your competitors with.

Taking SEO agencies as an example, when asked about a specific agency, an LLM might say that:

• Company A specializes in e-commerce SEO
• Company B is an expert in international SEO
• Company C handles SEO in general

That’s why, when someone asks about an SEO agency for international e-commerce, companies A and B show up in the recommendations and company C doesn’t.

The Entity Gap shows which needs, categories and use cases AI links more strongly to your competitors than to your brand. That tells you directly which topics to build content and case studies around, where you lack demonstrated expertise and what your digital PR and AI Authority Link Building should focus on.

Keep the layout simple and move from the big picture to the details. We recommend including the following elements:

1. An executive view: a table with the key KPIs mentioned above.

2. AI Visibility charts for individual platforms and intents.

3. A chart of the AI Competitor Gap over a longer period.

4. Key written takeaways summarizing the Prompt Gap, Source Gap and Entity Gap.

That set of metrics and findings is enough to give a Marketing Manager a complete picture of the brand’s visibility in AI Search compared with its competitors.

This kind of analysis is demanding and time-consuming, so it’s best to refresh it periodically, in stages:

• prompt monitoring: ongoing, using tools
• refreshing the core KPIs and the AI Competitor Gap: once a month
• checking for new competitors in AI Search and reviewing the Prompt Gap, Source Gap and Entity Gap: once a quarter

Before you sign off on the final form of the benchmark, check whether a Marketing Manager who reads the report will be able to answer these questions:

1. Who? Who is the real leader in AI Search in our industry?
2. Where? What AI Visibility do we have on each platform and for each intent?
3. Why? Where does our competitors’ advantage in specific areas come from? What are they doing better?
4. What do we do? What action plan are we putting in place based on the results?

In AISO projects, a competitor benchmark is one of the first stages of our work with a client. The analysis covers the brand’s visibility in AI answers and identifies the gaps where competitors have the advantage. What does that look like in practice?

For the Giacomo Conti brand, we ran a benchmark in which we:

• monitored 38 priority prompts
• analyzed the brand against its competitors in AI Search
• identified the gaps and the reasons competitors had the edge

Based on the results, we put a strategy in place that:

• increased brand mentions in AI tools by 300% in 6 months
• gave the brand 100% visibility in the monitored prompts about local stores

Read the case study: How a Strong Local Menswear Brand Increased AI Mentions by 300% in 6 Months

In this project, the benchmark gave us a reliable baseline, and the monitoring and KPI updates that followed showed whether the chosen strategy was working and delivering the expected results.

Sources:

https://www.airops.com/report/how-citations-mentions-impact-visibility-in-ai-search

https://www.muratulusoy.de/en/research/state-of-ai-search-2026.html