How to Tell If AI Is Citing Your Business

Why AI visibility is now its own metric

For twenty years the question was simple: where does your site rank in Google. That question still matters, but it no longer captures where buyers actually make decisions. A growing share of people ask an assistant a question, read the summarized answer, and act on whatever names and sources that answer contains. If your business is not in the answer, you never entered the consideration set, and your Google ranking did not save you.

The problem is that AI citations are harder to see than a ranking. There is no single dashboard that lists every time an engine quoted you. So measuring AI visibility means combining a few imperfect signals into a picture you can trust and repeat. It is closer to running a small research routine than reading a report, and once you set it up it takes fifteen minutes a week.

Query the assistants directly for your target questions

The most direct check is also the most overlooked: ask the engines the questions your customers ask, and read what comes back. Write down ten to fifteen prompts a real buyer would type, phrased naturally rather than as keywords. Then run each one through ChatGPT, Perplexity, Google AI Overviews, and Microsoft Copilot, and note whether your business appears, whether it is linked, and what the engine says about you.

• Best [your service] providers, phrased for a national or specific-region search
• Comparison prompts that pit your category against alternatives
• How-to and definitional questions your content should be the authority on
• Direct brand prompts asking what your company does and who it serves

Read the one report that actually shows citations

Bing Webmaster Tools is the closest thing to a first-party window into AI citations, because Copilot is built on Bing's index and retrieval. Inside the tool you can see how your pages perform in Copilot answers, which queries surface them, and how that traffic behaves. It is free, it is underused, and for any business serious about answer-engine optimization it is the first account to set up and verify.

Treat the Copilot data as directional rather than complete. It covers the Microsoft side of the ecosystem, not ChatGPT or Perplexity, but the patterns it reveals about which pages get retrieved usually hold across engines, because they all reward the same underlying clarity and authority. If you want the deeper mechanics of why engines pick one source over another, our guide on getting cited by AI walks through it.

Watch for referral traffic from assistants

When an assistant links you and a reader clicks through, that visit shows up in your analytics with a referrer. In 2026 those referrers are legible: traffic arrives from chatgpt.com, from perplexity.ai, from copilot.microsoft.com, and from the Google surfaces that carry AI Overviews. Build a segment or a simple filter that isolates these sources so you can watch the trend over time rather than hunting for it.

The volume will look small next to classic organic search, and that is normal. Assistant referral traffic is lower in raw numbers but tends to arrive later in the decision, closer to action. A handful of visits from an engine that just recommended you can be worth more than a large pool of top-of-funnel clicks, so judge the channel by what those visitors do, not by the count alone.

Track sentiment and accuracy, not just presence

Being named is only half the story. The other half is what the engine says about you and whether it is correct. As you run your prompt list, record the tone of each mention and flag anything wrong: an outdated service, a location you left years ago, a confusion with a similarly named competitor. A citation that carries a factual error can do more harm than no citation at all.

Over time you want two things trending in the right direction. First, more of your target prompts returning your business at all. Second, the mentions reading accurately and favorably. In our client work, positive sentiment across the major engines tends to settle into the high seventies to mid eighties percent once the underlying data and content are clean, and the remaining gap is usually a specific fixable error rather than general vagueness.

Build a simple monitoring routine

None of these signals is decisive alone, so the discipline is combining them on a schedule. A spreadsheet is enough. One tab for your prompt list with a column per engine and a date, one tab pulling Copilot data from Bing Webmaster Tools, and one tab for the referral trend. Run it on the same day each week and the picture assembles itself.

• Re-run the same prompt set weekly so you are comparing like with like
• Check Bing Webmaster Tools for shifts in Copilot impressions and queries
• Review the assistant referral segment in your analytics
• Log any accuracy errors and route them to whoever owns the fix

The routine matters more than any single tool, because the engines change their behavior often and a one-time snapshot goes stale fast. Consistent measurement is what lets you connect a content or structured-data change to a real movement in citations, which is the whole point. If the monitoring reveals that your business is essentially invisible to the engines, the next step is the strategic work behind our SEO and answer-engine services.