Ask ten Essex business owners what ChatGPT says about their company and nine will shrug. They have never looked. That is a problem, because customers in Chelmsford, Southend and Colchester are already asking these tools who to call, and the answers are not always flattering or even accurate.
The good news is that checking takes about an hour and costs nothing. Here is the routine we use when we first look at a local business.
Start With the Questions Customers Actually Ask
Do not begin by typing your company name. Begin with the question a stranger would type. "Who is a reliable electrician in Brentwood?" or "Best accountant for a small limited company near Basildon" are typical. Write down ten of these, mixing service, town and a specific need.
Then ask your own name directly: "What does [business name] in [town] do?" The two tests show different things. The first tells you whether you are recommended at all. The second tells you what the tool believes about you once it has heard of you.
Test More Than One Assistant
Run every question through ChatGPT, Google Gemini, Perplexity and Microsoft Copilot. They draw on different sources, so the answers often disagree. Perplexity shows its citations, which is useful because you can see exactly which web pages shaped the answer. Use a fresh browser window or a private session so your previous chats do not colour the result.
Also type a few of your questions into Google and look for an AI Overview at the top. For many local searches it now sits above everything else.
Record What You Find
Open a spreadsheet with one row per question and one column per assistant. Paste in the answer, or at least note whether you were mentioned, how you were described and which competitors appeared. Take screenshots as well. Answers change, and a dated record is the only way to know later whether anything you did made a difference.
It is tedious, but it is also the point where most owners see something surprising. A firm in Harlow might find it is described as a different kind of business entirely. You cannot fix that if you never saw it.
What to Look for in the Answers
We sort problems into four buckets.
- Absent: the tool never mentions you, even when you are plainly relevant.
- Wrong: old address, wrong phone number, services you stopped offering years ago, or a mix-up with a similarly named firm.
- Thin: you are named, but the description is generic and tells a customer nothing.
- Outranked: competitors appear and you do not, usually because they have clearer pages and more consistent listings.
Wrong is the most urgent. A wrong phone number costs real enquiries. Absent is the most common, especially for firms that rely on word of mouth and have a website last touched in 2019.
Trace Errors Back to Their Source
AI tools rarely invent details from nothing. They repeat what the web says. If an assistant lists your old Southend office, look for that address on your Google Business Profile, Companies House, Facebook, directories and your own contact page. One stale listing is usually enough to cause the mistake.
Fix the sources, not the chatbot. There is no form that makes ChatGPT forget something. What works is correcting the underlying pages and giving the tools clearer, consistent information to find next time.
Repeat It Every Quarter
Answers shift as models update and sources change. Put a recurring reminder in your diary, rerun the same ten questions and compare against your notes. Do not panic over a single odd result, because these tools can vary between runs. Look for patterns that repeat across sessions and across assistants.
If you would rather have someone else do the checking and the clean-up, that is part of what we do through our SEO service for Essex businesses, but the routine above works fine on its own.
The Bottom Line
You cannot manage what you have never looked at. Spend an hour asking the questions your customers ask, write down what comes back, and fix the pages that feed the mistakes. It is unglamorous work, and most of your local competitors are not doing it.