How to Check If ChatGPT Is Recommending Your Business (Step-by-Step)

To check if ChatGPT is recommending your business, run 20 to 30 real customer questions through ChatGPT in a fresh session, record whether your business is named, where it ranks in the answer, and which competitors appear, then repeat the test monthly. That process gives you a baseline for AI visibility, and everything else in this guide builds on it.

Local businesses have spent two decades tracking Google rankings, call volume, and review counts. AI answers add a new layer that none of those tools measure. A plumber can rank third in the Google map pack and never be mentioned when a homeowner asks ChatGPT, "Who should I call for a burst pipe at night in Tampa?" This post walks through the exact method to find out where you stand, the metrics worth tracking, and how to turn the results into action.

Why Checking ChatGPT Is Different From Checking Google

Google returns a ranked list of links. ChatGPT, Google AI Overviews, Perplexity, and Gemini return a composed answer, often naming two to five businesses and explaining why. There is no page two. Either your business is in the answer or it is invisible for that question.

Three structural differences make manual tracking necessary:

• Answers vary by wording. "Best dentist near me" and "dentist for anxious patients in Austin" can produce entirely different recommendations.
• Answers vary by run. The same prompt asked twice can return different businesses, so one test proves nothing. You need repeated samples.
• Answers vary by platform. ChatGPT, Google AI Overviews, Gemini, Perplexity, and Microsoft Copilot draw on different sources and show different results for the same question.

Because of this variance, tracking AI visibility is a sampling exercise. You are measuring the probability that your business appears, not a fixed position.

Step 1: Build Your Prompt List

The prompt list is the foundation of every metric that follows. Build 20 to 30 prompts that mirror how real customers ask questions. Pull them from three sources: the questions customers ask on the phone, the search terms in Google Search Console, and the text of your own reviews.

Sort prompts into four intent categories:

• Discovery: "What's the best HVAC company in Phoenix?" or "Top-rated chiropractor near me."
• Problem-based: "My AC is blowing warm air, who can fix it today in Scottsdale?"
• Comparison: "Is a med spa or a dermatologist better for Botox in Denver?"
• Trust and price: "How much does a roof replacement cost in Charlotte and who is reputable?"

Include your city, neighborhood, and service names in each prompt. Aim for roughly 8 discovery prompts, 10 problem-based prompts, 5 comparison prompts, and 5 trust or price prompts. Save the list in a spreadsheet. You will reuse it every month, and consistency is what makes trends meaningful.

Step 2: Run the Prompts Cleanly

Contamination ruins manual tests. ChatGPT can use memory and past conversation context, which skews results toward businesses you have already mentioned. Follow these rules:

1. Use a fresh session. Open a new chat for each prompt, or use a temporary chat that does not save history.
2. Log out of memory features or disable memory in settings so prior conversations do not influence the answer.
3. Turn on web search. Many local recommendations come from live retrieval. Test with search enabled, since that is how most consumers use it for "near me" questions.
4. Run each prompt 3 times. Three runs per prompt smooths out random variation. For 25 prompts, that is 75 answers per platform.
5. Set location context. Include the city in the prompt rather than relying on device location, so results are reproducible.

Repeat the process on at least three platforms: ChatGPT, Google (AI Overviews and AI Mode), and Perplexity. If you serve a customer base that skews toward Microsoft products, add Copilot.

Step 3: Record What Each Answer Contains

For every answer, capture the same fields. A consistent log turns anecdotes into data. Use one row per run with these columns:

• Prompt text and category
• Platform and date
• Whether your business was mentioned (yes or no)
• Position in the list (1st, 2nd, 3rd, or unranked mention)
• Sentiment of the description (positive, neutral, negative)
• Accuracy of the details (address, hours, services, phone number)
• Competitors named
• Sources cited (websites, directories, review platforms)

The sources column is the most valuable and the most often skipped. When ChatGPT or Perplexity cites Yelp, your state bar directory, a local news article, or a competitor's blog, it tells you exactly which pages feed the answer. Those pages become your optimization targets.

Step 4: Calculate the Core AI Visibility Metrics

Once you have 75 or more data points per platform, calculate five metrics. These are the standard AI visibility metrics used to benchmark local businesses.

Mention Rate

Mention rate is the simplest measure. If your business appears in 12 of 75 ChatGPT answers, your mention rate is 16%. Track it by prompt category. Many businesses discover a 40% mention rate on branded prompts (where the name is in the question) and under 5% on discovery prompts, which is where new customers actually come from. Always report non-branded mention rate separately.

AI Share of Voice

AI share of voice measures how much of the total recommendation space you own. Count every business named across all answers for a category. If 10 different dental practices are mentioned a combined 200 times and yours accounts for 24, your AI share of voice is 12%. This metric shows competitive position, which mention rate alone cannot. A 20% mention rate sounds weak until you learn the top competitor sits at 22%.

Average Position

Position matters because users act on the first recommendation far more often than the fourth. Record rank only when your business is in a list. A business that averages position 1.8 across mentions is in a stronger spot than one averaging 4.2, even at the same mention rate.

Citation Rate

Citation rate applies to platforms that show sources, mainly Perplexity, Google AI Overviews, and ChatGPT with search. Being named is one outcome. Being linked is another, and it drives traffic. Track both separately.

Accuracy Score

Check five facts in every answer that mentions you: business name, address, phone number, hours, and primary services. Incorrect details, such as a closed location or a service you do not offer, send customers to the wrong place and damage trust. An accuracy score under 90% signals inconsistent information across the web.

Step 5: Benchmark Against Three Competitors

Raw numbers mean little without context. Pick the three competitors that appear most often in your logs and compute the same metrics for them. The comparison reveals your real position:

• You lead: Protect the position by keeping reviews and content fresh.
• You trail one competitor: Study the sources feeding their mentions and close the specific gaps.
• You are absent while competitors dominate: You have a foundational visibility gap, usually missing listings, thin website content, or too few reviews.

This competitor view also shows who is gaining ground. A competitor whose share of voice climbs from 8% to 19% over a quarter has made changes worth reverse-engineering.

Step 6: Diagnose Why You Are Missing

When your business is absent from answers, the cause usually falls into one of five categories. Use the sources column from your log to identify which one applies.

Most local businesses have two or three of these gaps at once. Start with whichever gap shows up most often in your sources column.

Step 7: Set a Monthly Tracking Cadence

AI answers shift as models update and as new content gets indexed. A one-time check goes stale quickly. Use this cadence:

• Monthly: Rerun the full prompt list (25 prompts, 3 runs, 3 platforms) and update every metric.
• Quarterly: Refresh the prompt list with new customer questions and add seasonal prompts, such as furnace tune-ups in fall or AC repair in spring.
• After major changes: Retest within 2 to 4 weeks of publishing new service pages, launching a review campaign, or fixing listings, so you can tie results to actions.

Expect movement to be gradual. Content changes typically take several weeks to be reflected in AI answers that use live retrieval, and longer for answers drawn mainly from model training data.

Manual Tracking vs. Automated Tracking

Running 225 prompts a month by hand takes three to five hours. That is workable for one location and painful for several. Here is how the two approaches compare.

Start manually so you understand what the data means. Move to an automated AI analyst once you want ongoing monitoring, larger samples, and alerts when a competitor overtakes you.

What Good Results Look Like for a Local Business

Benchmarks vary by category and market size, but these ranges are practical targets for a local business tracking non-branded prompts:

• Under 5% mention rate: Foundational gaps. Prioritize listings, reviews, and service pages.
• 5% to 20%: Emerging presence. Focus on content depth and third-party mentions.
• 20% to 40%: Competitive. Work on position and citation rate.
• Over 40%: Category leader in your market. Maintain accuracy and freshness.

Dense markets, such as personal injury law in Los Angeles, produce lower numbers for everyone because answers spread across many firms. Compare yourself to local competitors rather than to a universal number.

Turning Tracking Into Revenue

Visibility metrics only matter if they connect to leads. Add three simple checks to close the loop:

1. Ask new customers how they found you. Add "ChatGPT or AI assistant" as an option on intake forms and in phone scripts.
2. Check analytics referrals. Look for traffic from chatgpt.com, perplexity.ai, and Google AI features in your reporting tool, and tag them as their own channel.
3. Track branded search growth. Customers often hear a name in an AI answer, then search it directly on Google.

Together, these show whether rising mention rate and share of voice produce more calls, bookings, and form submissions. A website with a built-in form builder makes the intake side easier to measure, since every submission can include a "how did you hear about us" field. WorkspaceCMS includes the form builder on every plan, and booking and scheduling on Growth and Premium, so those source questions can sit directly in the booking flow.

Common Mistakes When Tracking AI Visibility

• Testing once and drawing conclusions. One answer is a single sample from a varying system. Use at least three runs per prompt.
• Using only branded prompts. Asking "Tell me about Smith Family Dental" measures recognition, not discovery.
• Ignoring the sources. The mention tells you where you stand. The citations tell you how to change it.
• Testing on a logged-in account with history. Memory and personalization inflate results for businesses you have discussed before.
• Tracking only ChatGPT. Google AI Overviews appear inside the search results most customers already use, so omitting them leaves out a large share of exposure.
• Changing the prompt list each month. Trends require identical prompts. Add new ones, but keep the core set fixed.

AI Visibility Tracking FAQ

How do I check if ChatGPT is recommending my business?

Open a fresh ChatGPT session with web search on, then ask 20 to 30 real customer questions that include your city and service, such as "best roofer in Charlotte for storm damage." Run each prompt three times, log whether your business is named, its position, and which competitors and sources appear. Repeat monthly to see trends.

What is AI share of voice?

AI share of voice is the percentage of all business mentions in a category that belong to you across a set of AI answers. If competitors and you are named 200 times combined and your business accounts for 24 of those, your AI share of voice is 12%. It measures competitive position, not just presence.

How often should I track AI mentions?

Track monthly with a fixed prompt list, refresh the list quarterly, and retest 2 to 4 weeks after any major change such as new service pages or a review campaign. Monthly tracking catches model updates and competitor gains without producing noise.

Why does ChatGPT give different answers each time I ask?

ChatGPT generates answers probabilistically, and it combines training data with live web retrieval, so wording, timing, and context change the output. This is why a single test is unreliable and why tracking uses repeated runs and percentages rather than a single yes or no.

What is a good AI mention rate for a local business?

For non-branded discovery prompts, a mention rate of 20% to 40% is competitive for most local markets, and above 40% marks a category leader. Under 5% points to foundational gaps in listings, reviews, or website content. Always benchmark against the three competitors named most often in your own results.