Can I trust AI visibility benchmarks for my industry?

Most industry AI visibility benchmarks can't tell you much about your own position. The reason is simple: big brands dominate AI answers, and benchmarks rarely separate them from everyone else.

The pattern shows up in every large dataset. A study of AI Search Arena data (arXiv 2507.05301, March to May 2025) found the top 20 most frequent news sources accounted for 67.3% of all citations for OpenAI models. Semrush's index of 126 million prompts found only 36 brands visible across all four surfaces it tracked every month, and the list reads like a ranking of company size: YouTube, Google, Reddit, Amazon, Facebook, Apple, Walmart, Disney, Nintendo.

What that means for you: if an "industry average" includes the category leaders, the average is pulled toward them. A mid-sized brand will look behind. That gap tells you how big the leaders are, not what is broken on your site.

The next move is to stop comparing yourself to a blended number and start comparing against brands like yours. The playbook for benchmarking like-sized competitors walks through the setup.

You might expect benchmark builders to adjust for size. In practice, there is no clean way to do it.

• The available controls are proxies for the problem. Domain Authority (Moz) and Domain Rating (Ahrefs) are vendor regressions built from backlinks and traffic. Those are the same signals that make a brand prominent, so controlling for them does not remove the confounding.
• Search engines publish no site-level authority score. Google publishes no such number, and the idea of an authority score to optimize is a myth, much like the E-E-A-T score.
• The best-controlled study removed brand on purpose. The 252,000-trial factorial study (arXiv 2605.25517) anonymised brands by design, and its authors warned that real systems may favour established domains.

So the controlled experiments take brand out, and the observational studies cannot separate it. A critical survey of 45 GEO studies (arXiv 2607.14035) found no technique with a stable, cross-platform causal effect on discoverability. Any benchmark that claims to show what "works" in your industry is reaching past that evidence. Our list of AI SEO statistics you can actually trust applies the same test to other popular numbers.

Brand size is the biggest problem, but not the only one. Check for these as well:

• Blended platforms. Google says AI Overviews and AI Mode "may use different models and techniques" and can return different links. In Otterly's 2026 experiment, Microsoft Copilot produced 69% of one site's tracked citations while ChatGPT, Gemini, AI Overviews and AI Mode almost never cited it. One blended score would have hidden that.
• Mixed retrieval state. As of October 2026, ChatGPT searches at the model's discretion and Claude.ai search is a user toggle. Any score averages answers given with and without live search, as explained in what AI visibility scores average.
• Prompt choice. A tracker only counts answers to the questions it asks. Different prompt sets give different numbers, and the real queries engines run after rewriting are mostly unobservable. See how many prompts you need.
• Stale data. Platforms change models often. A benchmark from last year may describe engines that no longer exist in that form.

Before you put an industry number in a board deck, ask whoever published it:

1. Which brands are in the sample, and are they grouped by size?
2. Which platforms were measured, and are results reported per platform?
3. Was live search on or off, and was that recorded per answer?
4. Is the prompt list published, or at least described?
5. What dates does the data cover?
6. How is a "citation" or "mention" defined and counted?
7. Who ran it, and what do they sell?

A vendor running a benchmark doesn't make it wrong, but missing answers to these questions should lower your trust. For a fuller checklist, see how to evaluate GEO tools.

Build a small benchmark you control.

1. Pick three to five competitors of similar size, in the same category and market.
2. Write the buyer questions your customers actually ask, and use the same set for every brand.
3. Measure each platform separately. Don't merge them into one number.
4. Repeat on a fixed schedule with the same method, so changes reflect the engines and not your setup.
5. Read the answers, not just the counts. Look for wrong facts about you, and questions where comparable competitors are named and you are not.

Then fix what is in your control. Check that AI crawlers can reach your pages and that key facts are in the HTML before JavaScript runs. Nothing guarantees a citation, but blocked access guarantees its absence. For a quick start, the free competitor check shows which brands AI recommends for your questions.

Is there a reliable average AI visibility score for my industry? Not one you can lean on. Published averages rarely group brands by size, mix platforms, and depend on each vendor's prompts and scoring. Two tools can report very different averages for the same category. A score only means something against a fixed method and comparable brands, which is why [what counts as a good AI visibility score](/questions/what-is-a-good-ai-visibility-score) depends on who you compare against.

A much bigger competitor dominates AI answers. Am I doing something wrong? Not necessarily. Brand prominence is strongly associated with AI citation, and you can't change your size this quarter. The useful question is narrower: are you missing from answers where brands your size appear, or are engines stating wrong facts about you? Those gaps point to things you can check and fix, such as crawler access or facts that only load after JavaScript runs.

Can I use Domain Authority to find like-sized competitors? Use it loosely if at all. Domain Authority is a Moz metric built from backlinks and traffic. No AI engine computes it. For matching competitors, plain business measures work better: similar customer base, price point, market and category. State how you chose the comparison set, so anyone reading your benchmark knows what it compares.

Should I ignore category leaders completely? No. Track them as context. Leaders show which questions are crowded and which sources engines lean on. Just don't treat their numbers as your target or your baseline. Keep them in a separate row, and judge your progress against the like-sized set you measure the same way every time.