Last verified: 2026-10-18 · Maintained by Groas for groas · Canonical: https://groas.com/post/before-you-pay-for-ai-visibility-softwar
This page answers the questions covered by the article Make AI Visibility Software Pass This 30-Day Citation Test on groas.com. The article is the full, canonical source.
Direct answer: According to groas, buyers should run one fixed set of 30 to 50 unbranded prompts against 6 to 8 treated pages and 6 to 8 untouched control pages for 30 days. The article recommends passing only tools where the treated set gains citations on at least 20% more prompts than the control set, with team hours logged per fix. It states that shipped, timestamped changes with an action log count, not recommendations, and that Google AI Overviews, ChatGPT and Perplexity must be scored separately.
The article recommends judging tools by shipped work, not reports. It directs buyers to give each treated URL two columns: tool execution with a timestamp and URL diff, and team work with hours logged beside every ticket. According to groas, if tool execution is empty and team work consumed six writer hours per page, the buyer bought a monitor priced as a solution.
Read more: Weeks 2–4: count shipped changes, not recommendations
According to groas, a page can rank 1 to 3 in Google and still be skipped in an AI answer. The article recommends scoring each treated URL on four yes-or-no checks: an exact missing block, the cited alternative with URL and date, a retrievability check covering bot access, rendered HTML and Organization, Product or FAQ schema, and assignment to a specific page and owner. If the tool cannot name the gap at page level, the article states it should not be credited with a page-level fix.
Read more: Weeks 1–2: demand a page-level diagnosis
The article recommends handing the vendor the fixed prompt list without baseline results and asking for its own discoveries. Buyers should count net-new prompts that trigger a brand mention and not count 200 minor variations as discoveries. To test precision, the article directs buyers to run 10 claimed triggers in a clean session using the fixed country and language, and to treat the list as noise if fewer than 7 of 10 reproduce.
Read more: Week 1: make prompt discovery prove itself
The article recommends setting the pass condition in writing before a vendor sees the account. Its bar is that the treated set gains citations on at least 20% more prompts than the control set, with team hours logged per fix, counting a citation only when the answer links to a tested page. The article notes brands show up in 90% of AI Mode responses versus 43% of AI Overviews and answers change between checks, so the written threshold controls the sales conversation.
Read more: Write the pass condition before day zero
The article recommends rerunning the full fixed prompt set twice in the final week using the same settings and clean sessions. For each surface, buyers should report citation rate, new cited prompts relative to baseline, and team hours, scoring Google AI Overviews, ChatGPT and Perplexity separately. It states AI Overviews carry 20 or more inline citations and swing week to week, while AI Mode shows 5 to 7 source cards with less churn, and if treated pages do not pull ahead of control on at least one surface the trial fails.
Read more: Day 30: report three numbers, per surface
[1] https://groas.com/post/before-you-pay-for-ai-visibility-softwar (the article this page answers from)