An AI visibility audit should answer three business questions: where your company appears today, why relevant answers may overlook it, and which improvements are worth making first. It is not a single score from an unknown algorithm, and it should not begin with a promise to rank first in ChatGPT.
Below is a practical audit framework for service businesses, including the evidence to collect and how to prioritize findings.
1. Define what visibility is worth measuring
Choose the services, customers and markets that matter commercially. A Dutch consultancy selling to finance teams should not measure itself only against generic questions such as "What is AI?" It should focus on decisions potential buyers actually make.
Create three groups of prompts:
- Education: "What does automated invoice reconciliation involve?"
- Comparison: "When should we use reconciliation software instead of spreadsheets?"
- Selection: "Which providers support invoice reconciliation for mid-sized firms?"
Keep language, country and search mode consistent where possible. The goal is to understand a defined audience, not generate a flattering headline number.
2. Establish a reproducible baseline
Test a fixed set of relevant prompts across the AI platforms your target customers use. Record the exact question, platform, date, language, company mentions, cited URLs and other brands appearing. Repeat important questions because AI responses vary.
Separate outcomes:
| Signal | Useful interpretation | Limitation |
|---|---|---|
| Brand mention | Your name appeared in the answer | Does not prove a citation |
| Website citation | Your URL was linked as a source | Does not prove recommendation |
| Google AI impression | Google displayed your URL in a generative feature | Does not prove a visit |
| AI referral visit | Identifiable traffic reached your website | Does not capture all AI influence |
Google introduced dedicated generative AI performance reporting in Search Console in 2026, with worldwide rollout reported by August 31. Check the report where your property has sufficient data; it covers Google's own features, not ChatGPT or Perplexity.
3. Audit whether important pages are eligible to appear
Take your five most commercially important URLs and inspect them in Google Search Console. Check indexing, canonical status, crawl access and whether the page has a readable snippet. Google's guidance says supporting links in AI Overviews and AI Mode must be eligible for normal Search indexing and snippets.
For ChatGPT search, check OpenAI's documented OAI-SearchBot access separately. Note if a firewall or content delivery network may block otherwise permitted crawlers.
Do not mix technical eligibility with actual visibility. A perfectly indexable page can still be absent from an answer.
4. Review the content gap, not just keyword use
Take a prompt where a competitor appears and examine what useful information that competitor provides. Is their service described more specifically? Do they address cost drivers, implementation stages, locations, constraints or buyer risks that your website ignores?
Create a gap table:
| Question | Our best page | Missing detail | Proposed action |
|---|---|---|---|
| How do we start? | Services overview | Process and responsibilities | Add a process section |
| Who is it for? | Homepage | Clear target audience | Specify ICP on service page |
| What can go wrong? | No suitable page | Limitations and risks | Publish a practical guide |
Avoid creating a new blog every time. Sometimes a better paragraph on the existing service page is more valuable.
5. Verify authority and business information
Check whether the website clearly identifies the organization, services, contact details and any accurately stated credentials. Are sources provided for technical assertions? Does the About page help readers understand who is behind the advice?
Then look at real external context: relevant publications, trade organizations, partners and legitimate mentions. Do not treat manufactured reviews or purchased links as an authority strategy.
6. Prioritize actions by business impact
Give each action a simple score from 1 to 3 for customer relevance, confidence and effort. Address issues that block discovery first. Next, improve important service pages and high-intent content. Consider original research only when you can execute and document it properly.
A useful audit deliverable contains:
- The exact prompt set and test methodology.
- Screenshots or links documenting observations.
- Search Console and crawler findings.
- A content gap table with specific URLs.
- A prioritized 30-day action plan.
- A date for repeating the baseline.
An audit should lead to decisions, not another dashboard. Read our measurement guide or explore AEO services to turn the findings into improvements.



