Tutorials

How to Measure Your Brand's AI Visibility

Sep 26, 2026

In one line

Test ChatGPT and Perplexity with real customer questions, record mentions and cited sources, then retest monthly to track your brand's AI visibility.

The most direct way to find out whether AI recommends your brand is to ask it yourself. Draft a set of questions your customers would actually ask, put each one to ChatGPT and Perplexity, and record whether your brand appears and which sources get cited. Then retest on a fixed schedule to track how the answers change.

What AI Visibility Means

AI visibility is the degree to which your brand is mentioned or cited in the answers that AI search tools give when users ask a question.

The same practice is sometimes called AEO (answer engine optimization) or GEO (generative engine optimization). Be aware that in English, AEO more commonly refers to Authorized Economic Operator, a customs certification program, and the same three letters also serve as a fashion retailer's stock ticker. To avoid confusion, this article uses the term AI visibility throughout.

Traditional SEO measures rankings and clicks. AI visibility measures something else: whether AI writes your brand into its answers. The two are related, but neither replaces the other.

Step 1: Draft Questions Your Customers Actually Ask

Measurement starts with a question list, not a tool. Your job is to simulate how real customers phrase their questions, not to list the keywords you wish you ranked for.

Cover at least three categories of questions.

Recommendation questions test whether AI names you unprompted. For example: "Which web hosting providers in Hong Kong are worth recommending?"

Comparison questions put your brand and a competitor in the same sentence, such as "Which is better for a small business, A or B?" These reveal how AI describes the differences between you.

Regional questions add location and context, which brings them closer to purchase intent. For example: "A Hong Kong SME wants to open an online store. Which service providers should it consider?"

Write 5 to 10 questions per category. A total of 20 to 30 questions is enough for a first round of testing.

Step 2: Ask Each Question and Record the Results

Work through the list in ChatGPT and Perplexity. Perplexity answers generally come with citations, and ChatGPT lists reference links when its search feature is enabled.

For every question, record four data points: whether your brand was mentioned, where in the answer it appeared, how the AI described you, and which web pages were cited as sources.

Cited sources deserve special attention. If AI cites a third-party review article instead of your own website, your site's content is not yet treated as a reliable source, and that is a gap your content strategy needs to close.

A spreadsheet works well for this. Useful columns include date, tool, question, mentioned or not, a summary of the description, and the cited URLs.

Step 3: Set a Baseline and Retest on a Schedule

Your first round of results is your baseline. Retest with the same question set every month and compare three metrics: mention rate (how many questions produce an answer that includes you), description accuracy, and the composition of cited sources.

Keep the questions identical between rounds, or the comparison breaks. If you want to add new questions, track them as a separate set instead of mixing them into the old baseline.

After you update your website content, publish new articles, or adjust structured data, a retest shows whether those changes are reflected in AI answers.

The Blind Spot: A Single Result Proves Nothing

AI answers are non-deterministic. Ask the same question twice and you may get different answers, and a brand that appears in one run can vanish in the next. This is normal.

That means a single test result is never a conclusion. Repeat each question several times and look at how often your brand appears, not whether it appeared once.

Other blind spots exist. Conversation history on a logged-in account can influence answers, so an incognito window gives cleaner results. Answers can differ by region and language, and a model update can shift results across the board.

Read AI visibility as a trend, not a precise score. A mention rate moving from low to high, or citations shifting from third-party pages to your own website, are the directional changes that actually mean something.

FAQ

What is AI visibility?

AI visibility is the degree to which a brand is mentioned or cited in the answers of AI search tools such as ChatGPT and Perplexity.

How is AI visibility different from SEO?

SEO optimizes for rankings and clicks in search results. AI visibility focuses on whether your brand is written into AI-generated answers. The two complement each other.

How often should I retest AI visibility?

Retest monthly with the same question set, and add an extra round after major content updates. Keeping the question list consistent is what makes results comparable over time.

Why does the same question return different answers?

AI answers are inherently non-deterministic, so a single result proves nothing. Repeat each question and track how often your brand appears, then read the long-term trend.

Is AEO the same as the customs AEO certification?

No. In customs and trade, AEO stands for Authorized Economic Operator, a supply chain security program with no connection to AI search. The two only share an abbreviation.

From Manual Testing to a Technical Audit

Manual testing measures the outcome: how AI answers questions about you today. But visibility has a technical half as well: whether AI crawlers can read your site without obstacles, whether your structured data is complete, and whether llms.txt is deployed. The free AEO Auditor (ai.ud.hk/aeo-auditor) checks that technical half in one pass. Enter your URL and it maps the gaps. Compare the audit results with your manual test log, and you will know whether your next move is better content or better technical groundwork.