The best website in its field, and the firm AI names least.
An established professional-services firm, decades in its field and well known to anyone who follows it, asked a simple question: when people ask AI who to hire, are we in the answer? This is what one first scan showed. The firm is anonymised; every number is as measured.
9% of answers named the firm (7 of 76)
0 times it was the top recommendation
#7 among the firms AI named in its category
38% of answers named its leading competitor
What we asked
Twenty questions a future client would ask: “best firm for…”, comparisons, city searches and problem-first questions (“my application was refused, who should I hire?”). Nineteen of them do not mention the firm. We asked each one on ChatGPT, Gemini, Claude and Perplexity as a searcher in the firm’s main market, and sampled Google’s AI Overviews for the same questions.
Who AI named instead
| Firm | Share of answers that offer it as an option |
|---|---|
| Competitor A | 38% |
| Competitor B | 33% |
| Competitor C | 21% |
| Three other firms | 9% to 12% each |
| The firm | 9% |
It depended on the engine, and on the question
| By engine | Named in | By type of question | Named in |
|---|---|---|---|
| Claude | 21% | Comparisons | 25% |
| Gemini | 11% | “Best firm for…” | 9% |
| ChatGPT | 5% | City searches | 0% |
| Perplexity | 0% | Problem-first questions | 0% |
Google showed an AI Overview for all 19 questions and named the firm in 2. The overviews leaned mostly on forums and social sites.
The surprise: its website was the best of the group
We scanned the firm’s site and three competitors’ sites for how well AI can reach and read them. The firm scored highest of the four. All the main AI search crawlers were allowed, and its pages delivered their text without scripts. An AI agent sent to do nine customer tasks on the site completed four fully and five partly.
So the gap was not the website. It was everywhere else. Three directory and ranking sites that the engines cited again and again (10 to 16 times each) named the competitors and did not list the firm. The answers were being decided on pages the firm had never looked at.
What AI said about the firm
When the engines did describe the firm, several of their statements, about its founding year, its audience and its team, came with no source behind them. Some were right. NUR lists every such statement so the firm can check it and publish the fact plainly where AI can read it.
The plan that came out of it
- Pages for the questions it was losing. Cost, specific services and the two cities where only competitors appeared: one plain page each, saying what is offered, for whom, and how to start.
- Listings where the answers come from. A profile on each of the three directories the engines cite.
- A starting price, published. The cost questions went entirely to competitors that state one.
This is a baseline from one day, not a result. Single answers vary from run to run, which is why NUR keeps asking the same questions on a consistent schedule and reports the rate over many answers. What a first scan gives you is where to look.