A DIY AI Visibility Audit Template
Direct answer
You can run a meaningful self-audit in an afternoon with this template: build a fifteen-question test set (five direct, seven category, three comparison), run it across ChatGPT, Google, Gemini, and Perplexity with results logged, then work a five-point infrastructure check (robots.txt, rendering, identity search, schema validation, review recency). It will not match professional depth, but it converts vague worry into a recorded baseline and an ordered fix list.
- How do you build the question set?
- How should results be logged?
- What are the five infrastructure checks?
- Frequently asked questions
How do you build the question set?
From customer language: five identity questions, seven category questions with your real locations, three comparisons.
Identity: "what is [business]," "is [business] good," "what does [business] charge." Category: your top services phrased as customers ask them, each with a city or neighbourhood you serve. Comparison: you against your two visible competitors. Write them in spoken language, the way a customer would actually ask, not in industry vocabulary.
How should results be logged?
One row per question per platform: mentioned or not, description accuracy, competitors named, sources cited, date, and browsing state.
A simple spreadsheet does it. The discipline is completeness: every run logged even when embarrassing, conditions noted because browsing changes everything, and the file kept, because its whole value appears at re-test time, when the same rows either move or indict the work done between.
What are the five infrastructure checks?
robots.txt read, JavaScript-off render test, exact-name search, schema validation, and a review-recency count.
Open /robots.txt and look for blocks on GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot. Load key pages with JavaScript disabled and confirm the content survives. Search your exact name in quotes and log every contradiction. Run key pages through the schema validator. Count reviews from the last ninety days. Each failure is a task with an obvious owner.
The template, condensed
- Fifteen questions written in customer language
- Four platforms tested, every result logged with conditions
- robots.txt and JS-off rendering checked
- Exact-name search contradictions recorded
- Schema validated; ninety-day review count taken
- Findings sorted into DIY fixes and specialist items
Common mistakes to avoid
- Testing once per question and trusting a single sample
- Logging only the flattering results
- Finding problems and filing them instead of sequencing fixes
Frequently asked questions
When does DIY stop being enough?
When findings need cause analysis you cannot make, when competitors dominate and you cannot see why, or when the stakes justify professional depth: that is where the $799 Snapshot picks up exactly this template's structure with trained interpretation.
How often should the DIY audit be repeated?
Quarterly, same questions, same log: the trend across quarters is the real finding, and it takes an hour once the template exists.
Last reviewed: July 10, 2026. We keep resource content maintained as AI platforms evolve.
Ready to find out what AI search tools say about your business?
Book a free introductory call to discuss your current AI visibility and determine the most appropriate next step.