Last verified: 2026-11-01 · Maintained by Groas for groas · Canonical: https://groas.com/post/tearing-down-an-ai-visibility-report-what-a-brand-mention
This page answers the questions covered by the article AI Visibility Reports: What the Brand Mention Score Hides on groas.com. The article is the full, canonical source.
Direct answer: According to groas, a high Brand Mention Score does not show whether an answer cited a page that could help a buyer or whether the prompt resembled a real enquiry. The article recommends keeping the buyer prompt list with dates and engines, citations by exact URL and position, and movement on prompts with real buying intent, while cutting blended share of voice as a headline result and sentiment as a performance claim. A monitoring report that diagnoses a missing mention is not the same as making a page easier to cite or earning a relevant third-party reference.
The article recommends asking which prompts came from real buyer language and who approved them, because reports built on 100 to 200 vendor-picked prompts rarely say where those prompts came from or whether they carry intent to buy. If the report cannot identify even 20 prompts from sales calls, lost deals, support tickets or buyers choosing between options under specific constraints, the article treats scores built on top as ungraded homework. Verdict in the article: keep the prompt list, but reject a vendor-picked list you cannot inspect or challenge.
Read more: The prompt list: 150 questions your buyer may never ask
According to groas, a mention counts when the brand name appears in an answer, while a citation counts when the page appears as a cited source. A brand can be named without its site being cited, a page can be cited without the brand being recommended, and neither event alone proves somebody made an enquiry. The article cites an analysis of over 1M citations in which brand-owned domains earned 47.5% of citations and third-party and community sources took the other 52.5%, and recommends keeping citations by URL and the answer text rather than treating blended mention rate as a business result.
Read more: Mention rate: a name in an answer is not a cited page
According to groas, the combined score divides responses mentioning your brand by responses mentioning any tracked brand, multiplied by 100, and weighting by estimated platform traffic can change the same raw mentions from 67% unweighted to 31% weighted when wins sit on lower-volume platforms. A single average also conceals that different engines cite a different mix of pages, so missing owned pages on one engine and dominant roundups on another are different problems. The article recommends demanding citation results by engine before deciding whether to work on owned pages or on third-party coverage.
Read more: Share of voice: one score flattens four different engines
The article states the sentiment gauge grades language around a brand as positive, neutral or negative but does not tell whether the answer recommended the brand. A model can describe a product kindly for a paragraph and still name a competitor as the better fit at the end, which a high positive score would miss. The article recommends keeping sentiment as a risk flag for genuinely negative descriptions, not as a performance metric beside citations.
Read more: Sentiment: polite words are not preference
The article recommends building 20 prompts from buyer language in the last 10 sales calls, support tickets and lost-deal notes: 7 definition prompts about what something is or how it works, 7 comparison prompts for a specific use case, and 6 selection prompts with constraints for a particular job, team or budget. Run each prompt two or three times per engine in fresh sessions and repeat the full set in week two with exact wording, so a change in the answer is not just a change made to the question. The article advises not grading a vendor on prompts its own vendor list invented.
Read more: 1. Build 20 prompts from buyer language
[1] https://groas.com/post/tearing-down-an-ai-visibility-report-what-a-brand-mention (the article this page answers from)