Start with the questions, not the tools
Write down 10 to 20 questions a real customer would ask before choosing a business like yours. Mix three kinds: questions that name your brand, questions that name only the category and the place, and questions that compare options or ask how to choose. Keep the wording plain, the way a person types or speaks.
Save this list with a date. The same list is reused at every remeasurement so the results can be compared. Changing the questions between rounds makes any improvement impossible to attribute.
Run each question and keep the raw answer
Run every question in the AI tools your customers actually use. For each run, record the date, the tool or model, the exact question, the full answer text, the URLs it cited, and whether your business was mentioned and how it was described. A screenshot alone is not enough, because it cannot be searched or compared later.
Answers change between runs and between accounts. Run the list on at least two different days before drawing a conclusion, and report mentions as a fraction of the tested responses, not as a share of all AI searches.
Look at what gets cited instead of you
For each answer where a competitor appears, note the sources the tool cited: the competitor's own pages, a directory, a review site, a news article, or a social profile. This tells you where the gap is. If competitors are cited from their own service pages, your website content is the priority. If they are cited from directories and reviews, your third-party footprint is.
Check whether your site answers the decision questions
Take each question and find the single page on your site that answers it. Mark it as answered directly near the top, buried in a longer page, or missing. Typical decision questions cover what you offer and for whom, where you serve, what affects the price, how you compare with alternatives, what proof exists, and what happens next when someone contacts you.
The missing and buried items are your content gap list. They usually matter more than adding new blog posts.
Review the technical basics Google describes
Google states that its generative AI features are rooted in its core Search ranking and quality systems, that a page must be indexed and eligible to show with a snippet, and that no special AI markup or files are required. So the audit checks the ordinary things: index status in Search Console, canonical tags and redirect chains, robots and noindex rules, mobile rendering, and page speed.
Structured data can describe supported details such as a business name, address, hours, and phone number, but Google does not guarantee that features using it will appear. Google also announced in 2026 that FAQ rich results are no longer shown, so FAQ markup should not be the reason you write FAQs. Write them because customers ask those questions.
Audit the business profile footprint
Check the Google Business Profile against Google's own rules: the name should match the real-world name with no added keywords, taglines, or location descriptors, categories should describe what the business is rather than what it has, and each business should have one profile. Then compare the name, address, phone, hours, and services across the website, Google, Apple, Yelp, industry directories, and social profiles. Every mismatch is a signal that can confuse both people and AI systems.
Prioritize by revenue distance and effort
Score each gap on two questions: how close is the underlying customer question to a purchase, and can you fix it without outside help? A practical order is to repair indexing and redirect problems first, correct profile facts second, write the missing decision pages third, and add proof such as reviews, case details, and credentials after that. Publishing a large volume of new content before the basics are fixed rarely helps.
What the audit cannot tell you
An audit cannot predict whether an AI tool will recommend you. A sampled answer is not market share, a before-and-after difference does not by itself prove that one change caused it, and a missing mention is not a lost customer unless you can trace it. Keep the raw records so the next round is comparable, and treat any number that lacks a date, a denominator, and a saved run as an estimate.
Related pages
Sources
The technical statements above were checked against these Google documents on September 18, 2026. They describe Google's systems and should not be generalized to every AI provider.