There was a time when a brand determined exactly what people saw. A catalog on the kitchen table. A flyer in the mailbox. A product page you wrote, rewrote and argued about in three meetings. Every word was yours.

Then someone, somewhere, typed a question about your product into an AI assistant. Is it any good? Is it worth the price? What do people really think of it?

The assistant answered in seconds, with confidence..You weren't in the room.

Here are questions you need to ask: what is AI saying about your brand right now? And would you agree with it?

#Two sides of the same problem

When I talk to marketing teams about AI search, the conversation usually splits in two. One half is about being seen. The other half, the one people bring up more quietly, is about being understood. They feel like separate problems. I think they're the same one.

Side one: you might not show up at all

Take Oreo, the cookie everyone knows. When people asked AI chatbots for cookie recommendations, Oreo appeared in only about 10% of the answers, Digiday reported. Part of the reason was technical: AI crawlers were blocked from its site. Once Mondelez fixed that, Oreo showed up around 70% of the time.

If Oreo can go missing from a cookie answer, what are the odds for a B2B software company, a regional retailer or a brand that launched last year?

Most marketers already feel it. In a survey by Scrunch and Scribewise, two in three marketers said their brand appears in AI answers less often than they'd like, and 59% said they can't confidently turn their AI visibility data into action. They built their GEO strategy decks, but knowing what to change on Monday morning is the hard part.

Side two: when you do show up, someone else is talking

Visibility is the problem everyone's measuring, but control is the one that should keep you up at night.

Before AI, you were in control of your brand, all the copy that was shared with your brand came from your decks, websites, catalogs, and flyers. You could update them when needed, and any updated content could be hidden away.

Today, a brand's own sites make up only 5 to 10% of the sources AI search references. Reddit ranks first or second among the sources cited by every major AI engine, from ChatGPT to Google's AI Overviews. In B2B, review platforms like G2 do a lot of the talking.

That means when someone asks what your product is like, the answer might come from a two-year-old thread, a frustrated review or a page you forgot you published. Real users have every right to their opinions. What I want to know is whether your side of the story makes it into the answer at all.

Let me ask you, then: are you still confident that the message you want to send is the one your customers see? Or has your brand become one more piece of noise they have to cut through?

#AI answers for your brand, and it reads your noise

Some of that noise lives outside your walls. Devon Page, founder of WorkflowMind, recently posted on LinkedIn about a strange new habit: people are creating social media accounts for their AI assistants, so the agents can comment and interact online on their own. Our CMO Anthony Poliseno picked up on it. Trust is the obvious casualty, he wrote. The quieter one is that you can no longer tell which signals come from buyers and which come from agents. Plenty of the noise is homemade, though. The FAQ nobody has touched since 2022. Three versions of the same product description. Last spring's pricing page, still live, still ranking. AI tools read every one of them with the same attention they give your newest launch, then repeat what they find to your customers.

At Hygraph, we see three kinds of noise that matter to any business:

• Online noise: more content and more voices than anyone can verify, with bots and AI-written posts sitting right next to real people.
• Organizational noise: every page that never got archived and every doc since day one, which your team digs through to work out how things actually run.
• Context noise: AI tools pulled off-brand by a dozen different prompts an hour.

And the stakes go beyond reputation. When Air Canada's chatbot invented a refund policy, a tribunal held the airline responsible for all the information on its website, chatbot included. You can't hand the blame to the AI.

#So where do you start?

You can't control the assistant. You can control what it reads.

That's the part I find hopeful. Oreo got back into the answers by letting AI read its site properly. Most brands are one honest audit away from a similar fix, because AI doesn't need better prompts. It needs better context.

Three places to begin:

• Ask before your customers do. Ask the major AI tools what they say about your brand and your products. Write down every wrong answer and trace it back to its source.
• Give every fact one home. Prices, specs and policies belong in one place, with an owner and a date they were last checked. Fix a fact once, and every channel and every AI tool that reads you gets the fix.
• Cut your own noise. Retire the outdated pages, merge the duplicates and archive what nobody owns. Less to read means less to misread.

#Would you recognize yourself?

Soon, many of your customers won't visit your website at all. They'll ask their phone, their car or an agent acting on their behalf. The answers will still come from somewhere.

The brands that win will make sure those answers come from knowledge they own: accurate, current and in their own voice. Every answer a brand gives should be one it would stand behind.

One last question. If AI introduced your brand to a stranger tonight, would you recognize yourself?

If you're working out how to get your content ready for AI, see how Hygraph helps digital and AI leaders.