When AI Speaks for Your Brand
By Anis Hammouche·July 6, 2026
When a customer asks an AI who you are and whether you can be trusted, the AI answers, not you. It can get it wrong: an old address, an outdated price, a namesake, or your name missing in favour of a competitor. You cannot dictate its answer. You can read it, though, and fix the pages and listings it relies on.
Take Malik, a self-employed heating engineer (an invented character, like the two other cases in this article). A customer in his town asks an AI: "Who is a reliable heating engineer near me?". The AI names Malik, with the address of his old workshop and opening hours he dropped when he moved. The customer drives over, finds a closed door and calls someone else. Malik will never know why he lost that job.
Why an AI mistake about you goes unnoticed
Before, a customer who heard about you typed your name into a search engine. They landed on your site, read a few reviews and made up their mind. They saw your sources and did the sorting themselves.
With an AI-generated answer, they read a summary of your pages, blended with other sources and reworded. That summary can be accurate. It can also be incomplete or dated, and the customer receives it as an answer, not as one opinion among several.
The risk moves. A mistake on your own site, you fix in a minute. A mistake in what an AI says about you, you do not see go by unless you go looking for it.
I have no measure of how often these errors happen, and I would rather not invent one. What I describe here is a real risk that is easy to check, not a statistic.
Three businesses, three errors: what the AI answers and why
The three cases are invented. They show three different errors and what most likely explains them.
| Business (invented) | What an AI may answer | Likely explanation | What to fix |
|---|---|---|---|
| Malik, self-employed heating engineer | Old address, old opening hours | Google listing never updated after the move | The listing, then the address everywhere it appears |
| Atelier Nova, four-person web agency | A mix-up with another company with a similar name | Name, description and activity differ from one page to the next | One identical presentation everywhere, plus organisation markup on the home page |
| Sonia, freelance fitness coach | Three competitors in her city, not her | A one-page site that says neither what she does nor where | One clear page: what, for whom, in which city |
The first case is bad information: an opening time, a price or an address that is no longer current. The second is confusion, which can go as far as a fake site: pages that imitate an existing brand do exist, and an AI with no clear reference point may recommend the wrong one. The third is the competitor: your name is not cited because your information is less clear than the others', not because you work less well.
The test: ask two AIs what they know about your business
The test costs nothing and takes the time to read an answer. Open a mainstream AI tool, ask the first question, read everything. Then repeat on a second AI: the answers differ, and the gaps are what teach you something.
What can you tell me about [name of your business] in [your city]?
Is it trustworthy, what do they offer, at what prices, and what is their official website?
Give me your sources.
The second question measures your place against the others.
Who do you recommend for [your trade] in [your city]?
Give me three names and say why you cite them.
Write each gap in a log. Here is a template to copy, one per question you ask.
Date: [day of the test]
AI used: [name of the AI]
Question asked: [exact text]
What the AI answers: [summary]
What is true: [the reality]
Likely source of the error: [your listing, your site, a directory, no idea]
Fix to make: [action and where]
An AI that cites sources saves you half the investigation, but those sources can be invented too, which is the subject of AI Invents Figures, Sources and Terms. Open the links before you believe anything. To compare your place with your competitors' on one question, the method in Run a First Market Study With AI in One Hour works here too.
Four fixes that are up to you
What you steer is the material the AI reads. Google says it plainly for its AI answers: there are no additional technical requirements, no special markup to add, and the SEO fundamentals are what counts.
Your business listing. If you have no Google listing, you can add or claim it for free, and you then control what shows up for your business. In the same document on AI answers, Google also recommends checking that your listing information is up to date. That is exactly Malik's fix.
The same information everywhere. Your name, your activity, your address and your contact details should be identical on your site, your listing, your social profiles and your directories. Scattered or contradictory information is where an AI gets things wrong or mixes you up with someone else. For Atelier Nova, that means one description, copied word for word.
Organisation markup, if you have a website. Google explains that Organization structured data on the home page can help it better understand your organisation's administrative details and tell it apart from others. No property is required. It does not force any AI to cite you, it only helps with reading.
Light monitoring. Repeat the test from time to time, and every time you change address, price or offer. Also search your name on its own: if a site passes itself off as you, keep a dated screenshot before you react. Spotting a fake site early beats hearing about it from a customer.
| Fix | When to do it |
|---|---|
| Up-to-date business listing | Once, then whenever your address or opening hours change |
| Same information everywhere | Once, then whenever your name, offer or price changes |
| Organisation markup | Once, on the home page |
| Test and search for your name | From time to time, and after every change |
Frequently asked questions
Can I force an AI to say exactly what I want about my business? No, and it is best to accept that from the start. You do not write an AI's answer the way you write a page on your site. What you steer is the clarity and consistency of your official information. An AI that reads clear sources gets things wrong less often, with no guarantee.
Is my small business really concerned? Yes, as soon as a customer can type your name somewhere. The risk depends on how clear your information is, not on your size. The basic test stays the same: ask an AI what it knows about you.
What should I do if the AI says something false? First, pin down the exact error and its likely source, since the AI repeats what it finds. Correcting the information at the source, on your site or your listing, is the most useful move. If a fake site impersonates your name, the response falls under brand protection, and a lawyer can help. In every case, start by writing down what is being said.
Do I need a tool or a provider to monitor all this? Not to start. The first test is one you run alone with a mainstream AI tool. A tool or a provider becomes useful when you want to track ten questions across several AIs every month. That repetitive follow-up can be handed to an automation, but I have not tested it myself on a real brand, so I will not give you a result.
The next step is Google Rolls Out AI Answers in France: Prepare Your Visibility Before Summer, which details what Google shows your customers. I share this kind of test in my newsletter, you can sign up right below.
