Not long ago, a prospect researched you by typing your name into Google and skimming the first page. Today they ask an AI assistant, "Is this company any good, and who else should I consider?", and they trust the confident paragraph that comes back. That paragraph is your new first impression, and most brands have no idea what it says.
So what is AI saying about you? Whatever the web gives it to work with. Assistants build their answer from your website, your reviews and whatever else has been written about you. Clear, consistent material full of proof produces a flattering verdict. Thin or contradictory material produces a vague one, and sometimes a paragraph about somebody else.
AI is the new word of mouth
An assistant like ChatGPT does not hand back ten blue links for the buyer to judge. It writes one verdict in a confident voice. If that verdict is out of date, or built from a competitor's marketing, you have lost the deal before a human ever saw your homepage.
Think about how word of mouth used to work. A friend says "use these guys, they were great." You trust it because it is short and specific, and because your friend has no reason to lie. An AI answer lands the same way. It reads like advice from someone with nothing to sell, so buyers take it at face value. A lukewarm description from an assistant can kill a sale you never knew you were in.
Scale is what separates the two. A friend recommends you to one person. An assistant gives its version of you to every buyer in your market who asks.
You can't be in the room when AI describes your brand. So you'd better shape what it has to work with.
Where the AI's opinion of you comes from
Assistants are not forming opinions out of thin air. They draw on five broad sources, and each one is a place you can work.
- Your own website. What you say you do and who you do it for. If that is buried in slogans, the model has little to lift.
- Your business listings. Google Business Profile, industry directories and social bios. The contradictions usually live here.
- Reviews. What customers say about you, in their words, on sites the models can read. We cover this in why AI reads your reviews before it recommends you.
- Independent mentions. Articles, podcasts, forum threads, comparison pages. Anything written by people who are not you.
- Live search results. Since October 2024, ChatGPT has been able to search the web before it answers, and it decides when to do so based on the question. Whatever ranks for your name and category can end up in the reply.
Only one of those five is fully under your control. That is uncomfortable. It is also your opening, because most of your competitors only ever work on that one.
How to become the answer instead of the afterthought
- Be unambiguous about who you are. Say what you do and who you do it for, in the same words on your site and everywhere else the models read.
- Get described by sources AI can read. Reviews and guest articles are the raw material these systems work from. So are accurate directory entries.
- Structure your site for machines. Schema markup and a direct answer at the top of the page give an assistant something it can quote.
- Feed it real proof. Named results and specific numbers give the model concrete, favourable things to repeat about you.
Clarity means writing one plain sentence that describes your business and using it everywhere. "We build custom timber decks for homes on the Sunshine Coast" beats "We create outdoor living experiences", because a machine can repeat the first and can do nothing with the second. The principle behind this is covered in becoming an entity AI trusts.
Citations mean getting described by other people. Ask happy customers for reviews that mention what you did for them. Pitch a useful guest article to an industry publication. Check that you are listed accurately in the directories your buyers use.
Structure means pages that answer questions directly, with the answer in the first line and the proof underneath. Add an FAQ section built from the questions buyers ask you on the phone, then mark it up.
Proof means specifics. Years in business, service areas, turnaround times, guarantees, the types of clients you work with. Vague praise gives the model nothing. A concrete detail gives it a sentence to say about you.
A worked example: the builder the AI got wrong
Say you run a renovation company in Geelong. You moved out of Melbourne three years ago, stopped doing new builds and now specialise in kitchens and bathrooms. Business is good. Then a prospect mentions on a call that ChatGPT described you as "a Melbourne builder offering new homes and extensions."
So you go looking for the source, and it does not take long. Your old website copy is still sitting on a directory listing. A local business roundup from years ago shows your old address. Your Google profile is right, but LinkedIn says "residential builder." And your own homepage says "transforming spaces" without once mentioning kitchens or Geelong.
The model did its best with what it found, and its best was wrong.
The fix is methodical. Rewrite your homepage opening to say exactly what you do and where, then update every listing and profile to match it word for word. Ask your last twenty happy clients for a review that mentions their kitchen or bathroom. Publish a short guide answering the questions Geelong homeowners ask about bathroom renovations. Then run the same prompt every month and watch the description move toward the truth.
None of it is technical. It is housekeeping nobody did, because nobody was checking what the machines were saying.
Audit yourself today
Open your favourite AI assistant and ask it about your company and your top competitor. What it says, and whether it mentions you at all, may be the most useful brand audit you run this year. If you don't like the answer, good. You now have a list of what to fix.
Run it properly so the result is something you can act on:
- Use several assistants. ChatGPT, Gemini and Perplexity can give different answers to the same question. So can Google's AI results.
- Ask three kinds of question. The Geelong renovator would ask about the brand by name ("What do people say about Smith Renovations?"), about the category ("Who is the best bathroom renovator in Geelong?"), and for a comparison with a named rival ("Which of these two should I choose?").
- Write down what comes back. Are you mentioned? Is the description accurate? Which sources are cited? What does it say about your competitor that it does not say about you?
- List every error and gap. Wrong location, old services, missing specialties, no mention at all.
- Trace each one to a source. Most errors come from one specific listing or article. Fix it there as well as on your homepage.
- Repeat monthly. Save your prompts and run the same ones each month so you can see movement over time.
The mistakes that make the answer worse
- Fixing only your own site. If the directories and old articles still say the wrong thing, the model still has contradictions to deal with.
- Writing for keywords instead of facts. Pages stuffed with search phrases but light on plain statements give assistants nothing solid to repeat. See the three edits that get your business quoted for what to do instead.
- Buying fake reviews. It is dishonest, and the ACCC says it is against the law for a business to create fake or misleading reviews or to arrange for others to do so. It also teaches the model nothing true about you.
- Checking once and forgetting. The answers change as the web changes, so a single audit only tells you about the day you ran it.
What to do first this week
You do not need a six-month project plan to start. Five focused hours spread across the week will do it.
- Monday: run the audit. Ten prompts across three assistants. Paste the answers into one document so you can see the whole picture at once.
- Tuesday: write your one-sentence description. What you do, who for, where. Plain words. Read it aloud. If a stranger could not repeat it back, simplify it.
- Wednesday: fix your top five listings. Google Business Profile, LinkedIn and Facebook, plus your two most important industry directories. Same name, same description, same service area, same contact details.
- Thursday: rewrite your homepage opening. Put the one-sentence description at the top, followed by the proof: how long you have been doing it, who for, and what customers get.
- Friday: ask for reviews. Email ten recent happy customers. Ask them to mention the specific job you did and where. Specific reviews give the models specific things to say.
One week will not fix everything. It does remove the most common causes of a bad AI answer, and it gives you a baseline to measure against next month.
This is brand management now
For decades, brand management meant controlling your logo and your ads. Now a big slice of your reputation is written by a machine summarising everything else, and you cannot edit that summary. You can control the raw material: the homepage sentence, the Google Business Profile, the twenty reviews that mention a kitchen in Geelong.
That steady work is what our digital marketing team does for businesses who want to show up well when their buyers ask a machine for advice. Run the audit today. If you like what you hear, protect it. If you don't, you know where to start.
Frequently asked questions
How do I find out what ChatGPT says about my business?
Ask it directly. Start with your business name. Then try your category plus your city, and finish with a comparison against a competitor. Repeat the same prompts in Gemini and Perplexity, because each can answer differently. Note whether you are mentioned, how accurately you are described and which sources are cited. Save the prompts and run them again monthly to see if the picture improves.
Why does AI give wrong information about my company?
AI assistants build answers from what they can find online, including old directory listings and outdated articles. If your details differ across sources, the model may pick the wrong version or blend several together. Trace each error to its source and correct it there. Then make your business description match on your website and every listing.
Can I pay to be recommended by AI assistants?
You can now buy ads in ChatGPT, but you cannot buy the recommendation. In March 2026 OpenAI said it was extending its ad pilot to Australia for logged-in adults on the Free and Go plans, and it says ChatGPT's answers "remain independent and unbiased", with ads labelled and kept apart from the response. The way into the answer itself is to earn it with clear information on your own site, consistent listings and real reviews. Be wary of anyone promising guaranteed AI recommendations for a fee.
How long does it take to change what AI says about you?
It varies. Assistants that search the web live can reflect updated pages and listings once search engines have recrawled them. Answers that lean on a model's built-in training data change more slowly, often only when the model itself is updated. That is why ongoing work beats a single fix. Track it monthly with the same prompts so you can see progress.