Your business getting mentioned in chatGPT is the first step, but what if that mention is a negative one? If the answer raises concerns about your services or refund policy, the person reading it now has a trust issue. Counting showing up as a mention (like a lot of tools do) tells you very little about the impression it leaves.
This is where reputation management fits into answer engine optimization (AEO). Alongside tracking whether your business appears, you have to read what the answers say about it. Questions like "Is [company] reliable?" and "How does [company] handle refunds?" help you see what a potential customer might learn before buying.
What your visibility score leaves out
We covered the reporting problem in Every AI Visibility Tool Is Lying to You. A visibility score depends on the questions selected, how often they run and how the answers are scored. To judge the result, you need to see the actual answers and the conditions that produced them. Reputation adds another question: how does your business come across in those answers?
Imagine an answer that names your business, links to your website and then warns that your refund terms are unclear. You have a mention and a citation to your site. You also have a concern that a potential customer may want settled before buying. A report that stops at the first two observations leaves out the reason they might hesitate.
Sentiment helps you separate a recommendation from a warning or a mixed assessment. An answer that simply lists your address is factual; it should not be treated as criticism. Our article on Canonry's sentiment analysis explains how we keep those distinctions, separate branded questions from category questions, and retain the sentences behind a verdict. A label is useful when you can inspect what it was based on.
Ask about the decision to buy
Category questions such as "best [service] in [city]" help you see whether your business is among the options. Once someone has heard of you, their questions get more specific. They may want to know whether you are reliable, how you compare with a competitor, or what happens if something goes wrong.
Build those questions from what customers actually ask you. If refunds come up in sales calls or support messages, include a question about your refund policy. If customers regularly compare you with one competitor, check that comparison. You will get a more useful view than a set of broad category prompts alone.
Keep those groups separate when you review the results. Being absent from a list of options is a discovery problem. Being named in an answer that questions your service is a reputation concern. Each deserves attention, and each gives you different work to do.
The source tells you what needs attention
ChatGPT can search the web and link to sources. OpenAI warns that its search results and citations can be incomplete, outdated or incorrect. When an answer repeats a complaint from a review or Reddit thread, open the cited page and read the replies. You need to know whether the answer represents the discussion fairly and whether the issue is still unresolved.
An old complaint with a later resolution tells a different story from a recent complaint your business has never answered. If the answer leaves out the resolution, you have identified missing context. If the problem is still happening, the answer has brought a customer issue to your attention. The same care applies to praise: one favorable review is evidence of that person's experience, not a promise about every customer.
Ratings can also be misunderstood. The Better Business Bureau (BBB) bases its letter grade partly on complaint history and how your business responds. Customer reviews do not count toward that grade. If an AI answer cites your BBB rating, read the reasons on your business profile before treating the grade as a customer satisfaction score. Unanswered complaints are something your business can address directly.
Turn the concern into a concrete change
Suppose an answer says your business does not offer refunds. Compare that statement with the policy it cites. The answer may have missed an exception, or your page may actually say what the answer reports. An accurately described policy that customers dislike calls for a business decision. A confusing or incomplete page calls for clearer information. The answer can also be wrong even when your policy page is clear.
Your policy should explain the deadline, any costs and the exceptions. Make business and contact details easy to find too. Our guide to getting your business cited by AI covers the related work on clear pages, consistent listings and crawl access.
For a valid complaint, reply where it was posted and explain what you can do to resolve it. Identify yourself if you join a community discussion. Repeated complaints about the same issue should go to whoever can fix the service or process behind them. A public reply is useful when it gives the customer a resolution or a clear next step.
This is the work we describe in AEO Does Not Need Another Rank Tracker: connect the answer to its evidence, then decide what your business needs to change. The result might be an updated page, a response to a customer or an operational fix.
After the change, repeat the same questions with the same engine, location and app or API. Save the answers and their sources across repeated checks. AI answers vary, and a browser answer can differ from an API result, so one better response does not prove that your edit caused it. Look for whether the concern keeps appearing and whether the answer reflects the current information about your business.