30-Day Manual AI Visibility Playbook for One-Person Teams

Question 1

How do I find 20 buyer prompts worth testing in two hours?

Accepted Answer

The article recommends spending one 120-minute block to build a list of 20 questions worth testing, cutting vanity searches and broad educational topics. It directs the reader to pull candidates from sales conversations, support tickets or inbox threads, and question-shaped queries where the site gets impressions but few clicks, then rewrite each as a buyer would ask it and group them into direct alternatives, constraint-based qualification, and commercial evaluation and pricing. If there are 25 candidates, the article says to cut the five with the weakest purchase intent because the list is a work queue, not a vanity dashboard.

Question 2

How do I manually test buyer prompts to get an AI visibility baseline?

Accepted Answer

According to groas, the second 120-minute block should be used to run the 20 prompts across four surfaces in fresh sessions without personalized memory or chat history, marking an answer unavailable rather than trying to coax one out. For each engine and prompt the article says to log four things: whether the business was mentioned as 1 or 0, whether a citation pointed to its domain as 1 or 0, which rival was suggested, and the visible source URL, leaving the URL blank where no source is shown. The article stresses keeping mentions and citations separate because a model can name a business without linking to it or cite a page while recommending someone else. It warns against building a league table from brand order in one answer because generative responses can change when the same prompt is run again.

Question 3

How do I decide which page to fix when competitors appear in AI answers?

Accepted Answer

The article recommends starting Week 3 with the prompts where the business is absent and a competitor appears, beginning with the visible citations and reading the pages supplying facts to the answer. For each losing prompt it says to note what the buyer asked for, what the cited page states clearly, what the corresponding page says or omits, and what can be substantiated without inventing a claim such as an SLA that cannot be met. It cites a local-business audit of 500 companies to warn against assuming existing search presence handles this job. The block ends by ranking fixes by purchase intent and feasibility and leaving with one URL and one buyer question written down, or the audit is not finished.

Question 4

How do I fix one page in two hours to answer a buyer question?

Accepted Answer

The article recommends opening the existing URL chosen in Week 3 and placing a direct answer near the top under a heading that reflects the buyer’s question, followed by supportable constraints, pricing information, turnaround times or who the service fits, using a table for comparisons. It advises cutting a new 2,000-word article, sitewide title-tag tinkering and unbackable claims, and keeping the rest of the page coherent. After publishing the edit and noting the date, the reader should re-run the affected prompts in a later weekly block, recognizing that indexing and answers may not change immediately.

Question 5

When should a small business keep doing manual work versus moving to autonomous execution?

Accepted Answer

According to groas, after day 30 the routine should continue as one page per week: review the prompts, choose the next page, make the fix and re-test earlier edits, without expecting every answer to change on Friday. The article states the constraint stops helping when catalog, service territories or query volume outrun one person’s Tuesday morning, noting that testing 80 prompts across four engines is 320 checks before inspecting a source or improving a page. At that point it positions groas as a fully autonomous growth engine for paid and organic search with specialized models doing ongoing execution and a human strategist setting direction and guardrails.