How to check whether AI answers mention your brand

What questions should you test?

Start with three to five questions that could realistically influence a decision to buy from you.

For a desk brand, that might include:

• How much should I spend on a desk?

• What should I look for when buying a desk?

• What are the best desks for working from home?

• Is X worth the price?

• X vs. Y: which should I buy?

Prioritize questions where the answer could change what someone researches, compares, or buys. You are not trying to measure every possible prompt. You are trying to understand whether your brand appears when the decision matters.

How many times should you run each question?

Run each question at least three times in fresh conversations. If the answers vary significantly, run it a few more times before drawing a conclusion.

You are looking for patterns: brands that repeatedly appear, sources that repeatedly get cited, and competitors that consistently show up where you do not.

Do not treat one response as a ranking.

What is the difference between an AI mention and an AI citation?

An AI mention means your brand appears in the answer. An AI citation means your website is used as a source supporting the answer.

They tell you different things.

Mentioned and cited: Your brand is part of the answer and your content is being used as a source.

Mentioned but not cited: The model knows your brand, but other sources are supporting the answer.

Not mentioned but competitors are: You may have a brand, relevance, content, or authority gap.

Nobody is consistently cited: The topic may still be open for a stronger source to become the default reference.

Track mentions and citations separately. Improving one does not necessarily improve the other.

What types of pages do AI answers cite?

Across the citation patterns I’ve reviewed, the pages that get cited repeatedly tend to make useful information easy to identify, extract, and verify.

• Answer the question directly. Put the answer near the top and state it in a complete sentence.

• Make answers easy to isolate. Use descriptive headings, short answer-first paragraphs, lists where useful, and clear page structure.

• Use specific, checkable information. Numbers, dates, comparisons, named methods, original data, and clearly attributed claims give a model something concrete to reference.

• Support claims with evidence. Cite primary sources where appropriate and make it clear where facts and data come from.

• Demonstrate real authority on the topic. Useful first-hand research, expertise, original data, and established topical relevance are harder to replace than generic summaries.

Keyword density and word count are not useful proxies for whether a page is likely to become a cited source.

What if no source consistently gets cited?

If different sources appear every time you run the same question, the answer may not have an established source yet.

That is an opportunity.

Look at what the existing results fail to answer clearly. Then create or improve the page that should answer the question.

Start with the answer itself. Make it specific, complete, supported by evidence, and easy to understand without relying on the surrounding page.

Do not create a new page just because the query exists. First decide whether the answer belongs on a page you already have.

Should you update an existing page or create a new one?

If a competitor is cited and you are not, first check whether an existing page already covers the same topic. If it does, improve that page before creating another one.

Compare your page with the cited source. Look for:

• A direct answer you are missing

• Important facts or evidence you do not provide

• A comparison or decision point you leave unanswered

• Information buried too far down the page

• Claims that are vague or unsupported

• An answer spread across several paragraphs instead of stated clearly

Then edit the smallest thing that solves the problem.

Sometimes the difference between a useful page and a citable one is a single paragraph that clearly answers the question.

What should you do with the results?

Use the result to decide what happens next:

• Mentioned and cited: Protect what is working. Keep the page accurate and current.

• Mentioned but not cited: Identify which sources are supporting the answer and why your page is not.

• Competitor mentioned and cited: Compare the cited source with your closest relevant page and close the meaningful gaps.

• Neither you nor competitors are consistently cited: Look for an opportunity to create the clearest authoritative answer.

• Results change significantly between runs: Collect more data before changing anything.

The goal is not simply to increase mentions. It is to understand where you are absent, why another source is winning, and whether there is an action worth taking.

What should you avoid when optimizing for AI visibility?

There is no special markup that forces an AI model to mention or cite your page.

Avoid:

• Treating one AI response like a search ranking

• Creating new pages when an existing page needs a better answer

• Adding FAQ or schema markup solely because you expect it to trigger citations

• Measuring hundreds of prompts without knowing what decisions the data will inform

• Assuming a citation automatically produces traffic or conversions

• Blocking AI crawlers without understanding the business tradeoff

AI visibility, citations, referral traffic, and business impact are different things. Measure them separately.

How often should you check AI visibility?

For most sites, monthly monitoring is enough to establish direction. Check more frequently after making a meaningful change if you want to see whether the pattern shifts.

Track three things:

• Whether your brand is mentioned

• Whether your site is cited

• Which pages and domains are cited instead

Watch the trend over time rather than individual responses.

The most useful result is often not that your visibility changed. It is discovering which sources keep winning the questions that matter to your business.