I. The image you can't see
Most marketers have already looked up their brand in ChatGPT. Once. Out of curiosity. And then moved on.
What they didn’t ask: “Which provider would you recommend for X?” — without mentioning their own name. Or: “What are the pros and cons of this type of product?” Or: “Who is the best choice for someone in my situation?”
Those are the questions users actually ask. And they’re precisely the questions where your brand either shows up or disappears — without you ever seeing it.
Because AI models form an image of your brand whether you want them to or not. Based on everything that has been written about you. By you, but also by customers, journalists, comparison sites, forums and anyone who has ever published anything about you.
That image is shared with millions of users. Every day. In response to questions you never asked.
In my previous posts I described what LLM Visibility is and how AEO is rewriting the rules of SEO. Those posts were about visibility: being in the answer or not. This piece is about something subtler — and strategically at least as important: what gets said when you are in the answer.
II. How AI forms an opinion about you
Online reputation management has existed as long as the internet. Managing reviews, handling negative press coverage, steering Google results — familiar disciplines. They’re no longer sufficient.
Classic tools monitor what people see when they actively search for you. You manage the visible layer and can intervene, respond, counter.
AI reputation works differently. An AI model forms an image based on a training process that took place months ago, supplemented by live sources it retrieves at the moment of the query. That image isn’t a single result you can contest — it’s the condensed output of thousands of sources the model has processed and weighted. These include news sites and review platforms, but also Reddit discussions, industry reports, comparison sites and forums that most reputation teams never map.
And there’s this: with Google, a user gets ten results they can weigh and compare themselves. With AI, they get a pre-selected synthesis. The model has already decided which information is relevant and how the whole picture is framed. The user no longer filters — the model filters for them.
III. The four signals behind your AI reputation
How does a model build an image of your brand? Based on recognisable signals.
Tone of external mentions. Models implicitly build an image of brands based on recurring positive, negative or neutral contexts in external sources. If the general tone about your brand on independent sites is “expensive but solid,” that image will translate into how the model talks about you — even if you describe yourself in your own communications as “affordable and accessible.”
Authority of sources. A mention in a respected trade publication carries more weight than a post on an obscure forum. But that forum post doesn’t disappear — it’s simply weighted less. If there are far more negative than positive mentions in circulation, that volume ratio can colour the overall picture, even when the most authoritative sources are more positive.
Frequency of associations. Are you consistently associated with the right concepts? A brand that repeatedly appears in the context of the right product and “reliable” builds stronger entity value than a brand that’s mentioned incidentally without a clear pattern. A bank associated with “mortgages and security,” a software company with “ease of use and support,” a retailer with “fast delivery and fair pricing” — the pattern is the same, the sector differs.
Your own content as one voice in a choir. What you publish yourself counts — but less than what others write about you. Publishing consistently and informatively helps, but doesn’t drown out the external information flow.
IV. Four states — and most brands are in the wrong one
Not every brand has the same problem. Depending on what AI models say about you, you’re in one of four states.
Invisible. You’re rarely or never mentioned in answers about your sector or products. You exist in AI, but not in the orientation phase of buying decisions. This is the most urgent problem — and the least visible, because you won’t see it in your own analytics.
Flatly visible. You’re mentioned, but generically. Without distinction, without recommendation, without any reason for someone to choose you.
Suppose someone asks an AI model: “Which provider would you recommend for a mortgage?” or “Which accounting package is best for an SME?” The answer about a random player: “This is one of the larger players in the market with a broad range of products.” Factually correct. Strategically worthless. And not by coincidence: models tend to choose the safest, most neutral path when there isn’t enough distinctive data available. Being flatly visible is usually not an accident — it’s the result of an information vacuum the model fills itself.
Incorrectly visible. You’re mentioned, but the information is wrong. Outdated pricing, discontinued products, incorrect associations. Hallucinations are a real risk — but being incorrectly visible goes beyond hallucinations. It’s any form of inaccurate representation, from subtly outdated to outright wrong.
Strongly visible. You’re mentioned, correctly described and distinctively positioned. The model associates you with the right products, the right strengths, the right context. Users who encounter your name in an AI answer get a reason to look further.
Most brands think they’re in state four. Most are in state two.
V. Correct but colourless — the most underestimated position
Of the four states, being flatly visible is the most underestimated — and the most common.
An AI model that describes you as “one of the larger players in the sector with a broad range of products” isn’t saying anything wrong. It’s not hallucinating. It’s not lying. But it’s also giving no reason to choose you. You’re correct but colourless. Present but interchangeable. Visible but not winning.
And here’s the paradox: you invest heavily in brand positioning — in values, tone of voice, key messages — and then someone asks ChatGPT about your brand and the answer sounds flat. Why? Because AI models don’t only process your own communications. They average across thousands of sources. Your carefully crafted brand messages get diluted by everything others write about you — the customer who had a bad experience on a forum, the journalist who mentioned you as an afterthought, the comparison site that described your product correctly but colourlessly.
Brand positioning is what you say about yourself. AI reputation is what the model decides based on everyone.
Those are two different things. Most organisations only manage the first.
VI. Steering something you don't fully own
You can’t directly reprogram an AI model’s opinion. But you can influence the information sources it draws from.
Measure what’s there now. Test across multiple models — ChatGPT, Gemini, Perplexity, Claude. Ask not just brand-focused questions but also comparative and product-focused ones. What you want to know: am I mentioned and in what context? Is the information accurate? Which of the four states am I in?
Identify the discrepancy. Compare what the model says with what you want the model to say — not as a communications objective but as an information objective. Which facts, contexts and associations need to be more prominently present in the sources the model processes?
Strengthen the external information flow. Publish more informative content that answers specific questions and consistently establishes the right associations. Actively work on external mentions. Here’s an insight that surprises many: for AI reputation, it’s not just the link that counts — it’s the textual association itself. A brand mention in the right context can therefore carry value even without a hyperlink. PR and SEO are no longer separate disciplines.
Actively manage reviews and community. What users write on public platforms counts — for potential customers and for AI models alike. Encouraging positive experiences, proactively addressing negative ones: it’s no longer a nice-to-have.
Keep structured data up to date. Accurate, current markup is your first line of defence. Structured data gives models an explicit, machine-readable source of facts — one of the clearest factual anchors you can directly control.
Be consistent over time. AI reputation isn’t built in a month. Organisations that consistently publish the right information and build the right external presence see the difference over quarters, not weeks.
VII. A new discipline — organisationally too
Classic reputation management focused on response: managing a crisis, handling negative publicity. It was a defensive discipline.
AI reputation management requires proactive information structuring. Not waiting for something to go wrong, but continuously working on the quality, consistency and completeness of the information flow about your brand.
In practice, this calls for an organisational shift. Communications and SEO/AEO work together, not alongside each other. AI monitoring becomes not the final layer of communications, but an input for what communications and PR prioritise. The press releases communications writes, the FAQ pages the web team manages, the reviews customer service follows up on: they are all inputs for the brand’s AI reputation. But that collaboration doesn’t emerge naturally in organisations where those disciplines have historically operated in separate silos.
Content is evaluated on information architecture, not just creative quality. Is the answer to the question clear and findable? Is the information accurate and current?
Monitoring goes beyond social listening. What AI models say about you is at least as relevant as what people say on social media — because it influences the orientation phase of buying decisions before the user has made a single click.
VIII. Five steps to start with today
Test your own AI reputation. Open ChatGPT, Gemini and Perplexity. Ask five questions about your brand — vary between a general brand question, a product question, a comparison question, a recommendation question and a “best choice for” question. Read the answers critically. Which of the four states are you in?
Compare with the competition. Ask the same questions about your most important competitor. Who is framed more positively? Who is recommended more often?
Write down the discrepancy. What do you want the model to say versus what it says now? That’s your strategic starting point — not as a campaign brief but as an information plan.
Start with the easiest wins. Updating structured data, expanding FAQ pages, correcting outdated information: steps you can take quickly and that have immediate impact.
Make it a rhythm. Monthly monitoring, quarterly analysis of discrepancies, continuous improvement of the information flow. AI reputation is not a project. It’s an ongoing process.
AI reputation is a new domain — but not a radically different discipline. It builds on what good communications professionals have always done: getting the right information, in the right way, to the right place.
What changes is the scale, the speed and the impact. An AI model that provides millions of users with answers has more reach than any campaign. And that model draws on sources that are largely outside your direct control.
The brands that take this seriously now are building a reputational advantage that will be hard to close. Not because they understand the technology better. But because they understand that reputation in the AI era is no longer only about what you communicate.
The question is not just whether AI mentions your brand. The question is what image it leaves behind when it does.
