Check your own AI visibility: a simple monthly routine
You ask ChatGPT for the best bakery near you, and your shop shows up. Your business partner asks the same thing an hour later and gets a different list. Neither answer is wrong. AI tools build each reply fresh, and what they say depends on who is asking, where they are, how the question is worded and when it is asked.
That variability is the reason a single check tells you very little. What does tell you something is a steady routine: the same questions, asked the same way, once a month, written down. Over a few months you stop looking at snapshots and start seeing a direction.
All it takes is about an hour a month, the AI tools you already use and a spreadsheet.
Pick the questions your customers actually ask
Start with a short list of prompts, somewhere between eight and fifteen. Fewer, and one odd answer skews the picture; many more, and you won't keep it up.
Write the prompts the way a real customer would type them, not the way you describe yourself. Mix three kinds:
- Discovery questions, where someone has a need but no name in mind ("who does emergency furnace repair in Columbus?")
- Comparison questions, where they are weighing options ("what should I look for in a small business bookkeeper?")
- Brand questions, where they already know your name ("are reviews good for Northside Bookkeeping?", using your own business name)
Discovery and comparison prompts tell you whether AI tools think of you unprompted. Brand prompts tell you what they say once your name is on the table, which matters just as much.
Once you have the list, freeze it. The whole value of the routine comes from asking the same questions every month. If you want to test a new prompt, add it as an extra row rather than swapping out an old one.
Keep the conditions as steady as you can
You cannot control everything about how an AI tool answers, but you can remove the variables that are easy to remove.
Use the same tools each month. ChatGPT, Perplexity and Google's AI answers in search are a sensible starting set, since they are widely used. Use a logged-out or fresh session where you can, and check around the same point each month. If location matters to your business, note where you were when you ran the check.
The answers still won't be identical, but a change is now more likely to be real.
Record what you see, not what you hoped to see
Your spreadsheet needs one row per prompt per tool, with these columns:
- The date
- The tool
- The prompt
- Whether your business was named (yes or no)
- Where in the answer it appeared (first, in a list, or a passing mention)
- Which competitors were named
- Which sources or links the tool showed, if any
- Anything the answer got wrong about you
That last column is easy to skip and worth keeping. An AI answer that lists the wrong opening hours or an old service you no longer offer is telling you which outdated pages are still out there.
The sources column matters too. Perplexity and Google's AI answers often show the pages they drew from. Over a few months you will see which sites keep coming up for your topic, such as a local news site, an industry blog or a directory.
Read the trend after three months, not three days
One month of data is a baseline. Two months is a comparison. By the third month you can start asking useful questions.
Are you named in more prompts than you were at the start? Has your position in the answers moved from a passing mention toward the top of a list? Are the same competitors showing up every time, and if so, what do they have published that you don't? Are the cited sources ones you could reasonably be featured on?
Look across all your prompts. One prompt flipping from "named" to "not named" is normal noise; appearing in four prompts where you used to appear in one is a pattern.
A worked example: a two-person bookkeeping firm
Picture a two-person bookkeeping firm in Raleigh that works mostly with restaurants. In month one, they run twelve prompts across three tools. They are named in a few brand prompts, but never in discovery questions like "bookkeeper for restaurants in Raleigh." The cited sources keep pointing to a couple of local business blogs and a regional restaurant association's resource page.
That tells them where to spend effort. Over the next two months they publish two practical articles on their own site, one on handling tip reporting and one on month-end close for small restaurants, and they get both read beyond their own site. They also fix an outdated directory listing that AI answers kept repeating.
By month four, their spreadsheet shows them named in a couple of discovery prompts where they were absent before, sometimes near the top, sometimes not at all. The answers still vary from one check to the next. The direction is what they care about, and the direction is up.
Turn the findings into one action a month
A routine that only produces a spreadsheet will fade. Close each monthly check by choosing one thing to do before the next one:
Write the article that answers a prompt where you never appear. Update the page that an AI tool keeps quoting wrongly. Get an existing article in front of readers on the kinds of sites the tools keep citing. Ask a happy customer to mention you in a place where people discuss your industry.
One action, every month, tied to what you actually saw. That is how the trend line moves, and the next month's check tells you whether it worked.
Keep the routine honest
It is tempting to rerun a prompt until your name appears. Record the first answer instead. The routine only helps if it shows the truth, including months where nothing moved.
The aim was never to win one answer. It is to become the kind of business AI tools keep finding: one with clear, useful articles published and read in more than one place.
If getting your articles read across more sites is your next step, Launch Pages publishes your article and republishes it across a wide network of news and blog sites, each copy crediting and linking back to you. Learn more