AI visibility checker for any business

What you are measuring

The unit is a mention. You choose buyer-style questions for your industry and place. You ask them on ChatGPT, Grok, Gemini, and Perplexity. Mentioned means your public business name appears in that reply. Absent means it does not. A third outcome is common enough to write down on its own: the reply points at other businesses, or at a directory, and never reaches you. That is the whole measurement. It is smaller than “brand awareness” and concrete enough to explain to a partner in one sentence. It is also comparable over time if you keep the questions stable.

A checker should be modest about what a sample can prove. Four assistants and a set of questions are not a census of every person who might ask an AI product about you. Answers change when the wording changes, when browsing is on or off, and from one week to the next. A useful report says that in plain language. It also keeps a by-name question out of the recommendation count. “What do you know about us?” tells you whether the assistant can describe the business. It is a useful extra row. It is not a recommendation, and folding it into the tally hides a miss.

What a checker should show

Judge a tool by the questions it shows you and the names it returns. If the prompts stay hidden, or a single score has no replies behind it, you cannot decide what to fix. MentionMeter’s free Snapshot says whether you were mentioned. It is not a 0–100 index. The $39 report adds the other businesses named in the replies, the review and listing sites you are missing, and a 30-day fix list. The example report on the homepage is labeled example data and uses fictional names. It shows the layout you will be reading. It is not a customer’s result, and it is not the real check printed on this page.

How to read a result

Read a result in a fixed order. Start with the total, then the assistants, then the names that appeared instead of you. The card below is a real business we tested. The name stays off the page. Zero of 28 means the business was absent from every recommendation answer in that pass. The four rows show the miss was not confined to one product. “National lenders” and “directories” are generic labels for what filled the replies. On your own check those labels become the actual businesses and sites, because those are the names you can respond to.

Then read the limits, which belong on the card as much as the number does. One pass. Answers can change. A restaurant, a software product, and a clinic will be compared with different companies. The number does not transfer. The shape of the reading does: were you named, who was named, and which sources were cited. If the assistants can describe you when asked by name and still leave you out of the buyer questions, you have a recommendation problem, not a “they have never heard of us” problem. Those two problems lead to different work.

How MentionMeter runs the check

Here is how a MentionMeter check is run. You enter the business name, a website if it helps separate you from a similar name, a city or the word online, the industry in your own words, and an email. The questions follow that industry and place. With a city, a buyer-style question looks like “best [industry] in [city].” Online, the city drops out. The Snapshot is emailed after the questions run. The form does not ask for a card. It does not ask for a login to your site, to Google, to ads, or to analytics.

The $39 payment is once. The $19 monthly re-check is optional. It repeats the same kind of buyer-style questions so you can see whether the replies changed after you edited pages and profiles. It does not start on its own. You can look at the pricing on the homepage before you decide, and you can inspect the example report before you submit the form.

What to do with the output

What you do next depends on the sheet. If you were mentioned, note which wording produced it and keep the pages that support that answer. If you were absent, treat the other names and the cited sites as the work list. Align your public name, publish a page that answers the lost question in plain text, and show up on the sources the replies used. Then run the check again. A checker is a measurement. The repairs happen on your site and on the profiles a stranger can already open.

Use a checker when you want the four assistants asked the same way and a record you can keep. Use a handmade log when you want to learn the prompts yourself this afternoon. Either path fails if the questions are about your brand instead of about the buyer’s job. Write the buyer’s question first. Then see whether your name is in the answer.