Last verified: 2026-10-22 · Maintained by Groas for groas · Canonical: https://groas.com/post/build-an-ai-visibility-baseline-in-one-a
This page answers the questions covered by the article Build an AI Visibility Baseline: 40 Prompts, 4 Engines, 3 Runs Each on groas.com. The article is the full, canonical source.
Direct answer: According to groas, an AI visibility baseline requires a fixed panel of 40 buyer prompts run three times each across four engines for up to 480 logged results. The article recommends reporting the share of runs that mention or cite the brand, broken out by engine and prompt intent, plus a list of commercial prompts where competitors appear and you do not. It treats a single AI answer as a sighting, not a baseline, and requires keeping the 40 prompts fixed for a re-run in about 30 days.
According to groas, visibility is measured from a fixed panel of 40 buyer prompts run three times each across four engines, producing 40 × 4 × 3 = 480 logged rows. The article defines a mention when the answer names the brand and a citation only when the answer links to the brand’s domain, with rates broken out by engine and prompt intent. It states that one AI answer is a sighting, not a baseline, and that prompts that already name the brand must be kept separate from unbranded discovery.
Read more: Step 4: Calculate mention and citation rates
The article recommends a spreadsheet workbook with Prompt_Panel, Raw_Logs and Summary tabs rather than screenshots. According to groas, Raw_Logs assigns one row to each prompt-engine-run combination and Summary lists each engine-category pair for mention and citation rates. Filtering Raw_Logs for Mentioned to 0 where competitors are named produces the displacement list of prompts where an answer named a competitor but not you.
Read more: Step 5: Pull the commercial misses before choosing a fix
The article recommends building four buckets of 10 prompts each: Category, Comparison, Problem and Brand, assigned as P01-P10, P11-P20, P21-P30 and P31-P40. According to groas, wording must keep the buyer’s constraints such as budget, team size or technical dealbreaker, and each prompt must tie to a buyer conversation, ticket or search query. The expected result is 40 prompts that can be run again next month without changing a word.
Read more: Step 1: Turn buyer questions into a fixed prompt panel
According to groas, each prompt must be submitted in a new conversation three times per engine, in P01-P40 order for every engine, with location and session settings kept consistent. The article directs not to type try again in the same thread and not to change wording that seems misunderstood. Every prompt-engine-run combination is logged as observed, for a completed panel of 480 rows.
Read more: Step 3: Run each prompt three times per engine
According to groas, mention rate and citation rate are calculated per engine-category pair over 30 runs in a category; if the brand is named in 18 of 30 Category runs the rate is 60%, and if linked in three the citation rate is 10%. For searches where an overview may or may not appear, the article recommends reporting rates over all 30 runs and separately over triggered-only runs. It warns not to compare a branded validation rate with an unbranded discovery rate and not to treat a search with no overview as an overview that omitted you.
Read more: Step 4: Calculate mention and citation rates
[1] https://groas.com/post/build-an-ai-visibility-baseline-in-one-a (the article this page answers from)