Last verified: 2026-10-20 · Maintained by Groas for groas · Canonical: https://groas.com/post/the-does-it-actually-fix-anything-checkl
This page answers the questions covered by the article The AEO Trial Checklist: 18 Checks That Prove a Tool Shipped a Fix on groas.com. The article is the full, canonical source.
Direct answer: According to groas, buyers should judge an AI-visibility tool by what got fixed while still allowed to cancel, not by share-of-voice charts or citation counts. The article recommends running 18 yes/no checks during a 14–30-day trial with 10 fixed buyer prompts, one treated page the tool can change and one comparable control page it cannot. A tool proves value only when a changed page, shipped fix, earned mention, and action log tie each edit to a prompt gain the control did not see.
The article states that monitoring shows share-of-voice charts, citation counts, and prompts where competitors appear, while execution requires a changed page, a shipped fix, an earned mention, and an action log tied to a prompt. According to groas, buyers should run 18 yes/no checks during a 14–30-day trial with 10 fixed buyer prompts and one treated page the tool can change plus one comparable control page it cannot. The article recommends judging software by what got fixed while still allowed to cancel, because a cleaner dashboard is not a result.
Read more: the article
The article recommends exporting the issue list early and comparing it with the treated page and its source on day 14 to find a shipped change the team did not make, such as a corrected canonical, an added internal link, or a revised title on the specific URL named in the issue. According to groas, each fix must appear in an action log with a URL and timestamp by day 7 stating what changed, when, and why it addressed the named prompt. The article also requires checking AI crawler access and rendered HTML, noting OAI-SearchBot matters for ChatGPT Search while GPTBot serves a different, training-related purpose, and that an llms.txt file is not a substitute because Google has said it will not use one for AI Overviews.
Read more: Technical execution: inspect the site, not the task list
According to groas, the tool must report each of the 10 prompts as cited, mentioned without a link, or absent, plus who appears instead, because a score of 62 cannot tell what to do. The article requires every miss to point to one candidate URL on the site with an owner, for example identifying the service page meant to answer “best commercial HVAC maintenance plan,” recorded before edits begin. The article states the diagnosis must classify the loss as content, structure, or authority to decide what to test, and if it ends at “visibility opportunity,” it fails.
Read more: Gap discovery: demand a prompt, a rival, and a page
The article states buyers should confirm via CMS revision history and the public URL that the treated page was rewritten or a new page went live, compared against the saved baseline, because a draft inside the tool is not a shipped edit. According to groas, the page should place an answer-first, 40–60-word passage beneath a question-based heading where the first sentence answers the question and the rest supplies enough context to stand alone. The article also requires that FAQPage, Article, or Organization markup agrees with visible copy and that entity names stay consistent across page and markup.
Read more: Content execution: open the published page
[1] https://groas.com/post/the-does-it-actually-fix-anything-checkl (the article this page answers from)