
Quick answer: how do I check what ChatGPT says about my business?
Open a new chat in a private window and run 10 fixed prompts about your service, prices, area and name. Score each answer on four yes or no checks: mentioned, recommended, cited with a link and facts correct. Record the results in a sheet, repeat monthly and compare. The test takes about 15 minutes in ChatGPT alone.
To check what ChatGPT says about my business, I run the same 10 questions in a fresh chat, score each answer and log the results in one sheet. The whole test takes about 15 minutes, costs nothing and tells you whether AI tools mention you, recommend you, link to you or get your facts wrong. It is the first step before any AI search optimization in London, Ontario, since you cannot fix what you have not measured.
This guide gives you the 10 prompts, the rules for running them fairly, a scoring method and a tracking sheet. It does not repeat how ChatGPT picks its sources or the steps to improve your standing. Those are in our ChatGPT optimization guide. This post is about the test itself.
Check what ChatGPT says about your business before you optimize
You should test before you optimize, because the result tells you which problem you have: invisible, mentioned wrongly or outranked by a named competitor. Each problem has a different fix, and guessing wastes a month.
Business owners in London often assume that a strong Google ranking carries over to AI answers. It sometimes does and sometimes does not. A dental clinic on Richmond Row can sit in the Google Map Pack and still never be named when someone asks an assistant for a dentist near Western University. The test shows which case you are in.
A baseline also protects you from false progress. If you start work without recording today's answers, you cannot tell in three months whether a change came from your effort or from the tool changing its behaviour. Fifteen minutes now gives you a number to compare against.
What the test can and cannot tell you
The test shows how assistants answer your questions on the day you ask. It does not prove how often anyone asks them, and it does not prove that a mention turns into a call. Answers also change from one session to the next, so treat one run as a sample and the monthly pattern as the signal.
The 10 prompts to run: service, price, location and comparison questions
The 10 prompts cover the four ways a customer asks an assistant about a local business: by service, by price, by place and by comparison. Use your own trade and neighbourhood in each. The table shows the pattern with a fictional physiotherapy clinic in Old North as the example.
| # | Prompt pattern | Example | What it tests |
|---|---|---|---|
| 1 | Closest-fit service search | Recommend a physiotherapy clinic in London, Ontario | Whether you are named at all |
| 2 | Service plus neighbourhood | Physiotherapy clinics near Old North London Ontario | Whether you appear for your own area |
| 3 | Service plus neighbouring town | Physiotherapy in St. Thomas or Strathroy Ontario | Reach across your service area |
| 4 | Problem-based question | Who can help with a sprained ankle in London Ontario | Match to customer problems |
| 5 | Price question | How much does a physiotherapy session cost in London Ontario | Whether your prices or ranges are found |
| 6 | Trust question | Is [your business name] a good physiotherapy clinic | What it says about you by name |
| 7 | Fact question | What are the hours and address of [your business name] | Accuracy of the basic facts |
| 8 | Comparison question | [Your business name] compared with [a competitor] | How you are described next to a rival |
| 9 | Urgent or after-hours question | Physiotherapy open Saturday London Ontario | Hours and availability data |
| 10 | Reviews question | What do people say about [your business name] reviews | Whether reviews are quoted and where from |
Prompts 1 to 5 test discovery, meaning whether a stranger finds you. Prompts 6 to 10 test reputation, meaning what the tool says once it knows your name. Both matter, and many businesses pass one half and fail the other.
Write your own versions first
Before you start, write your 10 prompts in a document and keep the wording fixed. Changing the wording each month ruins the comparison. Use the phrases your customers use, not the ones you use. A customer says roof leak, not envelope remediation.
How to check what ChatGPT says about your business fairly: logged-out, new chat, same wording, three assistants
A fair test uses a new chat, no personal history and identical wording every time. Assistants personalize when they can, so a test run from your own logged-in account mostly shows what the tool knows about you, not what a stranger sees.
- Use a clean session. Open a private browser window or use a temporary chat feature, and sign out where you can. If you stay signed in, expect your own history to colour the answers.
- Paste the prompt exactly. Do not add context such as your name or your website unless the prompt itself includes it.
- Start a new chat for each prompt. A follow-up inside the same chat inherits the earlier answers.
- Run it in three assistants. ChatGPT is the main one, and adding Perplexity and Google's AI answers shows whether a pattern is general or specific to one tool.
- Copy the answer and its sources. When the tool shows links or citations, record them. The sources show where it found its information.
- Note the date and the mode. Record whether the tool searched the web or answered from memory, since the two behave differently.
Thirty answers, 10 prompts across three tools, sounds like a lot, but each takes under a minute to paste and copy. If time is short, run all 10 in ChatGPT first and add the other two tools the following week.
Why location matters in the test
Run the test from London, or add the city to every prompt. Assistants that use your device location may return different answers from a laptop in Toronto than from one in Masonville. Writing the city and neighbourhood into the prompt removes that variable.
How to score each answer: mentioned, recommended, cited and accurate
Score every answer on four yes or no questions: were you mentioned, were you recommended, was a link to your site shown and were the facts correct. Four simple checks are quicker and more repeatable than a rating out of ten.
| Check | Yes means | No means | Points |
|---|---|---|---|
| Mentioned | Your business name appears in the answer | You are absent | 1 |
| Recommended | You are named as a suggestion, not just listed in passing | Named only as an example or after competitors with no endorsement | 1 |
| Cited with a link | Your own website or profile is shown as a source | Sources are competitors, directories or none | 1 |
| Facts correct | Name, address, phone, hours and services are all right | Any fact is wrong or out of date | 1 |
Each answer scores from 0 to 4. A total across 10 prompts gives you a score out of 40 per assistant. Do not chase a number. The pattern matters more: which prompts score zero, and which checks fail most often.
Flag wrong facts separately
A wrong fact deserves its own note, since it can cost a customer on the day. Write down exactly what was wrong, which assistant said it and which source it came from. A wrong closing time or an old address usually traces to a directory listing, and that is fixable.
What the common results mean and the next step for each
Most results fall into one of four patterns: invisible, mentioned wrongly, a competitor named instead or doing well. Each pattern points to a different next step.
| Result | What it usually means | Next step |
|---|---|---|
| Invisible: never mentioned | Little for the tool to find or trust about your business | Build the basics: clear service pages, a complete Google Business Profile and consistent listings |
| Mentioned with wrong facts | Outdated or conflicting information on a source it reads | Find the source in the citations and correct it, then align all listings |
| Competitor named instead | The competitor has more consistent or fuller information | Compare what they publish with what you publish, then close the gap |
| Mentioned and recommended, no link | The tool knows you but does not use your site as a source | Strengthen your own pages so they answer the exact question |
| Strong across most prompts | Your information is clear and consistent | Keep monitoring monthly and protect what works |
If a competitor is the one being recommended, read our post on why ChatGPT recommends your competitors before deciding what to change. It explains the reasons and keeps this post focused on measuring.
If your problem is that you are named but never linked, our guide to getting cited by AI in London, Ontario covers how sources are chosen. For the wider picture of how this fits with search, see AI SEO in London, Ontario.
A worked example: scoring one answer from start to finish
Scoring one answer takes under two minutes once you know the four checks. The example below is illustrative, not a real client result, and shows how a typical answer is read.
Say a fictional plumbing company in Byron runs prompt 2, plumbers near Byron London Ontario. The assistant lists four companies. The plumber's name is not among them, so Mentioned scores 0, and the three remaining checks score 0 as well because there is nothing to judge. Total for the prompt: 0 out of 4.
Now say the same company runs prompt 7, the hours and address question. The assistant names the company, gives the right address and a wrong closing time, and cites a directory listing instead of the company website. Mentioned scores 1. Recommended scores 0, since a plain fact answer is not a recommendation. Cited with a link scores 0, because the source is a directory. Facts correct scores 0, because of the closing time. Total: 1 out of 4, plus a note that the directory shows the wrong hours.
That single note is worth more than the score. It tells the owner exactly which listing to correct, and it takes ten minutes to fix. Most of the value in this test comes from notes like that one, not from the totals.
A simple tracking sheet: columns and a monthly cadence
A single spreadsheet with one row per prompt and per assistant is enough. Add a new tab or a new set of columns each month, so earlier results stay visible.
- Column A: date of the test. One date per run.
- Column B: assistant. ChatGPT, Perplexity or Google.
- Column C: prompt number. 1 to 10, matching your written list.
- Columns D to G: the four checks. Mentioned, recommended, cited, facts correct, each as 1 or 0.
- Column H: competitors named. Write each business name that appeared.
- Column I: sources shown. Paste the domains the tool cited.
- Column J: notes. Wrong facts, odd wording and anything surprising.
Run the test once a month on the same day. Weekly runs add noise, and quarterly runs miss changes. After three months you will have enough rows to see a trend, and a sorted view of the competitors column shows who keeps appearing in your market.
What to do with a sheet that never changes
If three months pass and nothing moves, the test has done its job: it has told you that waiting is not a strategy. At that point, the question becomes whether to change the content and listings yourself or to bring in help.
When to fix it yourself and when to ask for help
Fix it yourself when the problem is a wrong fact on a listing you control. Ask for help when the problem is absence, since that points to a gap in your content, schema or authority that takes ongoing work.
- Do it yourself: correcting your address, hours or phone number on your Google Business Profile and directories, and updating your own About and service pages.
- Do it yourself: asking recent customers for reviews, since fresh reviews give tools current information to read.
- Get help: you are invisible on 7 or more of the 10 prompts after you have fixed your listings.
- Get help: the sources the tools cite are all pages you do not control.
- Get help: you need schema written, content restructured or monitoring set up every month.
Bradlex SEO publishes an AI search add-on range of $150 to $400 per month, or $300 to $800 as a standalone project, and a free audit shows whether it is worth it for your business. If you want a second pair of eyes on your test results, request a free SEO audit and send us your sheet. You can also call (647) 692-2131.
The test works the same in St. Thomas, Strathroy, Woodstock and Ingersoll. Only the place names in your prompts change.