With AI visibility tracking, you ask answer engines the same buyer questions on a regular schedule and record what they say about your business. Each month you document whether ChatGPT, Gemini, Perplexity, and Google AI Overviews mention you, how they describe your services, and which pages they cite. A fixed list of prompts turns a one-time check into a measurable trend, and an ordinary spreadsheet is enough to begin.
You may have completed a quick version already, with the 10-question check in how to show up in ChatGPT answers or the AI section of a brand audit. A single check captures one day, and ongoing tracking shows whether improvements to your website and media coverage are changing the answers.
What is AI visibility?
Visibility in AI answers is how often, and how accurately, answer engines mention your business when someone asks a question you could answer. It matters because a potential client can read your name in an AI answer and never visit your website.
Pew Research Center studied the browsing data of 900 US adults from March 2025, and 18% of their Google searches produced an AI summary. When a summary appeared, people clicked a traditional search result in 8% of visits, compared with 15% on pages without one. Only 1% of visits to a page with a summary included a click on a link inside it. An answer in AI search results can therefore be a buyer’s entire first impression of you, and your analytics reports will never reveal what it said.
How is AI visibility measured?
I monitor four metrics, and each one is calculated from the same log of answers:
- Mention rate is the share of answers that name your business.
- Citation rate is the share of answers that link to a page on your own website.
- Accuracy is the share of your mentions that describe your services and location correctly.
- Competitor share is how often each rival business appears across those same answers.
Keep a separate count of AI mentions for every engine you test, because a business can appear in one engine every week and never in another. A single combined number across engines would conceal that difference. Google says even its own AI Overviews and AI Mode “may use different models and techniques, so the set of responses and links they show will vary.”
Record a baseline before you change anything on your website. OrangeBag, a commercial laundry company, saw its share of AI answers in Semrush tracking go from 16% to 49.73% within two weeks of its site rebuild. Without the 16% starting figure, that improvement would have had no reference point.
Step 1: Build a fixed set of buyer prompts
Begin with the questions from your initial check, and organize them into four groups according to the buyer’s motivation:
- Problem prompts describe the frustration, such as “how do I fix six months of messy bookkeeping.”
- Category prompts request a type of provider, such as “best bookkeeper for a small dental practice.”
- Comparison prompts evaluate two options, such as “bookkeeper vs accountant for a small business.”
- Branded prompts include your business name, such as “is [your business] a good bookkeeper.”
Aim for 10 to 15 prompts in total, with a minimum of two in each group. For a local business, add a second version of each category prompt that includes your city.
Once the list is final, freeze the wording, since changing one word creates a different question whose answer cannot be compared with last month’s. Number each prompt in a document, and copy from that numbered list for every run.
Step 2: Keep your testing conditions consistent
Answers can vary depending on who is asking and from what location, as I explain in my post on how ChatGPT finds its sources. Hold every variable you can control steady from month to month:
- Use a Temporary Chat in ChatGPT for every run, so your saved memories stay out of the answers.
- Work from the same city each month, and document the location in your log.
- Check Google AI Overviews and AI Mode with the same setup each month, ideally in a private browser window.
- Record the model or mode name each engine displays, since a model update can change the answers.
Perplexity is a practical starting point when you want to see sources, because its help page says it searches the internet in real time. According to that page, each answer includes numbered citations linking to the original sources.
Step 3: Repeat every prompt three times
One prompt can return a different answer on each run, so an isolated result can make a random variation look like a trend. Ask every prompt three times in each engine, open a new conversation for each run, and count every run as a separate answer. At 10 prompts, four engines, and three runs, one month produces 120 answers.
Step 4: Log every answer in a standard format
Copy these columns into a spreadsheet, and give every answer its own row:
- Date and engine: [date], [ChatGPT, Gemini, Perplexity, AI Overviews, or AI Mode].
- Prompt and run: [prompt number from your fixed list], [run one, two, or three].
- Named my business: [Yes or no], plus my position if the answer is a list.
- Description: [the sentence about my business, copied word for word].
- Accurate: [Yes, partly, or no], with the specific error noted.
- Competitors named: [each business, in the order they appear].
- Pages cited: [each URL], marked as mine or someone else’s.
Save each description exactly as the engine phrased it. A month later, that precise wording reveals whether an engine has absorbed your new services page or is still repeating an outdated directory listing.
Step 5: Calculate the monthly rates and compare
Harbor Books is a two-person bookkeeping firm I made up, and its October run of 10 prompts, four engines, and three runs produced 120 answers:
- Harbor’s name appeared in 18 answers, a mention rate of 15%.
- Its own pages were cited in six answers, a citation rate of 5%.
- Accurate descriptions accounted for 12 of the 18 mentions, or 67%, and the other six listed a payroll service the firm discontinued last year.
- A larger competitor appeared in 54 answers, or 45%.
The owner rewrote the services page and removed payroll from three directory listings the engines were repeatedly citing. In December, the same prompts, engines, and runs produced 120 new answers, and 33 of them named Harbor Books, a mention rate of 27.5%. Its pages were cited in 15 answers (12.5%), and 30 of the 33 mentions were accurate, an accuracy rate of 91%.
Interpret small movements cautiously when you compare months. A difference of two or three mentions out of 120 is under 3%, a change that ordinary run-to-run variation can produce on its own. An increase from 18 to 33 over two months is a more meaningful shift, and the accuracy improvement suggests the corrected pages were reaching the answers.
Step 6: Choose a schedule you can maintain
Run the complete set on the same day each month, such as the first Monday. After a significant change like a site rebuild or a major article about your business, schedule one extra run two weeks later.
Revisit the prompt list every quarter, and add new prompts as your services evolve. Continue running the original prompts as well, so your trend line retains a fair comparison.
How to check AI visibility of a website in Search Console
Google Search Console provides partial information about AI Overviews and AI Mode. According to Google, sites that appear in those features are included in your overall Search Console traffic and reported “within the ‘Web’ search type” of the Performance report. The Search Console Help Center adds that an AI Overview occupies a single position in the results, and selecting an external link inside it is counted as a click.
Those clicks and impressions are combined with ordinary results, and the report keeps no record of what an AI Overview said about you. I treat Search Console as a secondary signal beside the prompt log. In the Performance report, open the Pages tab, select a page that answers one of your tracked prompts, and compare its monthly impressions and clicks with your mention rate.
How to improve AI visibility once you can see it
Each pattern in your log requires a different response:
- Inaccurate descriptions need a correction at the source, and the worksheet in where ChatGPT gets its information helps you trace each error to the page responsible for it.
- Missing mentions call for a review of the pages cited in your competitors’ answers, and the steps for earning a place on those pages are in how to show up in ChatGPT answers.
- Wikipedia articles that appear repeatedly among the cited pages are worth evaluating with my Wikipedia readiness worksheet before you invest time in one.
- Thin branded answers point back to your own website and profiles, which should state clearly who you are and which clients you serve.
For Google’s AI features, the pages that already perform well in traditional search are the ones to improve first. Google’s AI features page says “there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
AI visibility tracking tools vs a spreadsheet
Dedicated AI visibility tracking tools run your prompt list automatically on a schedule and turn the results into AI visibility reports. Some vendors call them LLM visibility tools, after the type of AI model that generates the answers. Automation can save hours once your list grows, and until then a spreadsheet does the same job at no cost. Before you sign up for one of their paid plans, ask the vendor these questions:
- Which engines does it monitor, and does it report AI Overviews and AI Mode separately?
- How many times does it run each prompt, and at what frequency?
- Can you read the full answer text, or only an AI visibility score?
- If it includes sentiment analysis, can you see the sentence behind each rating?
- Is there a location and language setting for each individual prompt?
- Does it display every cited URL for each answer?
A free AI visibility checker can be a useful first look, provided you treat its result as a single day’s snapshot. Whichever tool you choose, keep your own prompt list and a copy of your AI visibility data, so you can switch tools later without losing your trend line.
Frequently asked questions
What is AI visibility?
AI visibility is how often and how accurately answer engines such as ChatGPT, Gemini, Perplexity, and Google AI Overviews mention your business. It covers whether you are named, how you are described, and whether your own pages are cited as sources.
How do you track AI visibility?
Ask a fixed list of buyer prompts in each answer engine once a month, three times each, and log every answer in a spreadsheet. Record whether you are named and described accurately, along with the competitors and cited pages, and compare the totals from month to month.
How is AI visibility measured?
It is measured as percentages across every answer in your log, calculated separately for each engine. Mention rate is the percentage of answers that name you, and citation rate is the share that link to your website. Accuracy is the share of your mentions that describe you correctly.
How do I check the AI visibility of my website?
In Google Search Console, AI Overviews and AI Mode traffic is included in the Web search type of the Performance report. It shows clicks and impressions for your pages, and a monthly prompt log across ChatGPT, Gemini, and Perplexity documents what each answer said about you.
How do I improve AI visibility?
Start with your log, because it identifies whether your problem is an inaccurate description or a missing mention. Inaccurate descriptions are corrected on the page the engine cites. Missing mentions call for clearer pages on your website and coverage on the external websites that appear in your competitors’ answers.
My 90-Day Growth Engine includes AI visibility tracking from the start, with a baseline of where you appear today. Run your first month of prompts, then bring the log to a 30-minute call, and I will tell you what the numbers say.







