About this sample
The sample is for Quillberry Roofing Co., a fictional roofing contractor in Austin, TX, prepared under a placeholder agency brand, "Your Agency". The answers in it are real: on November 7, 2026, ten buyer-intent questions about roofing contractors in Austin were put to ChatGPT, Claude, Gemini and Perplexity with web search on, and the 40 answers were saved. The sample was produced from those saved answers; no assistant was asked again.
Two things are different from a client report. Because the business doesn't exist, no assistant could name it. And because this page is public, the names of the real local businesses the assistants recommended are withheld ("Local roofer A, B, …"). The counts are unchanged. A client report shows every name.
Open the sample (PDF) Open the sample (HTML)
Headline results: 0 of 40
The top of the report answers the client's first question: was the business named? Here it's 0 of 40 answers — the expected result for a fictional business, and what a business with no AI visibility sees. A real client would see its mention rate, the number of assistants and questions, and its average position when listed.
Mention rate per assistant and the question matrix
Each assistant gets its own row, because they behave differently. In this run, ChatGPT named on average 1.4 businesses per answer, and 4 of its 10 answers named no specific business at all. Gemini named 5.4 per answer and always gave names. Claude and Perplexity sat in between, at around 4. The matrix underneath shows every question against every assistant, so a client can see exactly where they appear — and where they don't.
Competitors named instead
This is usually the section clients read twice. Across the 40 answers, the assistants named about 60 different businesses. The most-named one appeared in 17 of 40 answers (12% of all business mentions); the top ten together took about half of all mentions. The assistants disagreed a lot: only 2 businesses were named by all four assistants, 7 by three or more, and over 40 by just one. For an agency that's a useful message: visibility in one assistant says little about the others.
Cited sources
Assistants that search the web cite their sources, and those sources show where to work. ChatGPT and Perplexity cited sources in all 10 of their answers, Claude in 8. Gemini returned no separate citation list on this route, so only one of its answers contributed links. Across all answers, more than 120 different domains were cited. The most common ones included a roofing manufacturer's website (9 answers), the Better Business Bureau and a contractor-review directory (6 answers each), press-release sites, local news sites and individual contractors' own websites.
Recommendations
The recommendations are derived from the results with fixed rules, so they're consistent from report to report. For this sample they say: answer the high-intent questions directly on the website (all 10 questions were gaps), study the most-named competitors' profiles and reviews, make sure the business is listed on the sources the assistants cite, complete the Google Business Profile, and grow recent, detailed reviews. They're suggestions, not promises: no specific placement in AI answers can be guaranteed.
Methodology and limitations
The last section states how the data was collected and what it can't tell you: the models and search setup used, that API answers can differ from the consumer apps, that answers vary between runs, and that 40 answers is a small sample, so percentages are indicative. Business names are extracted automatically and a few can be missed or mis-parsed. In this sample the section also explains the fictional business, the withheld names, and that the run was originally scored against another placeholder name and re-scored for Quillberry.