How to check what AI says about your brand, and correct it
You cannot edit what an assistant says about you. You can check it the same way every month, and fix what it reads.
Ask the assistants your buyers use the same set of questions every month, in a fresh session that knows nothing about you, and save every answer word for word with the date and the tool. Score each one against your own approved description of the company. When an answer is wrong, find the page it stands on and fix that page or publish a clearer one, because you cannot edit the answer itself.
On 9 November Gartner reported that 42% of CMOs have taken steps to monitor and correct misinformation about their brands in AI-generated output. In the same survey of 408 senior marketing leaders, run in July and August, 61% said they had adapted content to improve machine readability.
So for four in ten CMOs, what AI says about the brand is already someone’s job. The easy version of that job is typing the company name into ChatGPT and screenshotting whatever comes back, which tells you about one afternoon. We run a monthly check on ourselves instead, and the first one found us missing from every who-to-hire answer we read.
Why does it matter what AI says about our brand?
Because buyers increasingly use AI tools to discover and evaluate brands, as Gartner notes, and a wrong answer can shape a shortlist before your team knows it was in the room. Gartner’s analyst Sharon Cantor Ceurvorst put it plainly: marketing leaders “can no longer take a passive approach to brand representation” in AI-generated experiences.
The risk is rarely a scandal. It is a description that is slightly off: a product you do not sell, a category you left, a competitor named where you should be. Nobody complains about those answers. They just quietly send the buyer somewhere else.
You cannot write the assistant’s answer. You can decide what it has to work with.
How do we check what AI says about us?
Ask a fixed list of questions, the same way, every month, and keep the answers. A one-off search tells you how one answer looked on one afternoon; a monthly read tells you whether anything you changed moved it. Four rules keep the read honest:
Use the buyer’s words Write the questions a buyer would type: what your category is, what your company is, who to hire for the job you do. The who-to-hire questions matter most, because that is where a buyer makes a shortlist.
Start every session fresh Log out or open a clean session with no memory, no saved instructions and no files about you. An assistant that already knows your brand is answering your context, not the market’s.
Keep every answer word for word Save the full answer with the date and the tool before anyone scores it. The transcript is the evidence; the score is only a reading of it.
Score three things Did it name you, did it link to you, and was what it said accurate against your own approved description of the company. Accuracy needs a yardstick, so write that description down first.
Ask every assistant your buyers actually use, and say plainly which ones you could not read. A tool that blocks you after three questions gives you three answers, not twenty, and a gap you write down is more useful than a number you guessed.
What did we find when we checked ourselves?
We were named in 7 of the 63 answers we could read, and all seven came from four questions: what a Brand Brain is, what a customer-owned Brand Brain is, who builds Brand Brains, and what Adaptius is. None came from the broader questions a buyer asks when hiring, such as who is a good brand AI consultant. We ran our first full read on 6 November: twenty questions, asked fresh, in Google’s AI Mode, Gemini, Perplexity and Claude (run from its command-line tool, not the chat app). Perplexity asked us to sign up after three questions, so we stopped there. We did not read ChatGPT or Grok this month, so neither is counted.
- The good: asked who builds Brand Brains, Gemini named us first and had both founders right.
- The gap: every who-to-hire question we read left us out.
- The error: one Google answer said we also make a no-code assistant builder. We do not. It came from a third-party software listing.
- The blank: Claude, asked what Adaptius is, said it could not find the company.
Six of the seven mentions were accurate. That is a small first sample from one month, read by us about us, so it is a starting line and not a benchmark. Next month we ask the same twenty questions and compare each tool with itself.
How do we correct a wrong answer?
Fix what the answer stands on, not the answer. Assistants that search build their answers from pages they find, so a wrong answer usually traces to a page: yours, or someone else’s about you. Google says its AI answers come from its Search index and link to supporting pages, and that optimizing for them “is optimizing for the search experience, and thus still SEO”, with no special file or markup required.
Find the source Open the links the answer cites. Our wrong answer led to a software directory listing, so the fix sits with that listing, not with the assistant.
Say it plainly on your own pages Put the one-paragraph description of the company in readable text on the pages that matter, in the same words everywhere, so every reader, human or machine, meets one version.
Answer the question you are missing If assistants never name you for a who-to-hire question, there is probably no page of yours that answers it. Write that page.
Read again next month Same questions, same tools. A change you made counts only when the next read shows it, and even then one month is a hint, not proof.
None of this guarantees what an assistant will say. It improves what it has to work with, which is the part you control.
Who should own what AI says about the brand?
Marketing owns the yardstick, and anyone disciplined can run the read. The hard part is not the asking. It is agreeing, in writing, what the company is, what it claims, and what it never claims, because without that every wrong answer becomes an argument instead of a fix.
We keep that description, with the proof behind each claim, in our Brand Brain: the approved brand thinking your people and AI tools read before they make anything, kept in files your company owns. The same files are the yardstick our monthly read scores against, and the source our own pages are written from.
Where should we start this week?
Write the paragraph you would want an assistant to give about your company. Then pick ten questions in your buyers’ words, ask three assistants in fresh sessions, save the answers and count.
If you are named and right Keep the questions and ask them again next month. You now have a starting line.
If you are named and wrong Follow the links to the page behind the error and fix it there, then check your own pages say the right thing plainly.
If you are not named at all Look at which questions you lose. A missing answer is usually a missing page, not a missing trick.
For the bigger picture, read what an AI brand strategy is, and for the rules behind the work, what AI brand governance means.
Where this comes from.
-
Analyst firm press release (survey)
Gartner · Gartner Survey Finds 42% of CMOs Are Monitoring Brand Misinformation in AI-Generated Content
Used: survey of 408 senior marketing leaders, July through August 2026; 42% of CMOs report taking steps to monitor and correct misinformation about their brands in AI-generated output; 61% adapted content strategies to improve machine readability and visibility in AI-driven experiences; the quoted line from Sharon Cantor Ceurvorst, VP Analyst. Not used: the 94% planning more human touchpoints.
-
Product documentation
Google Search Central · Optimizing your website for generative AI features on Google Search
Used: generative AI answers draw on Google's Search index and show links to supporting pages; optimizing for them is still SEO; no new machine-readable files, AI text files or special markup are needed. It describes Google only, not other assistants. Page footer shows last updated Jul 10, 2026; opened Nov 9, 2026.
