Ask ChatGPT what a brand is known for and you get an answer with no byline, no editor, and no disclosure of who shaped it. That answer is confident. It is often accurate. And it is increasingly the first impression a huge number of people form about a company, a founder, or a product category, before they ever read a press release, a review, or a news article written by an actual human being. Nobody hired these systems to be reputation managers. They have become them anyway, and most companies have not noticed.

I have spent my career in a business built on the premise that reputation is something you actively manage: through media relations, through crisis response, through the deliberate, patient work of shaping how the right people talk about a company at the right moments. That premise has not changed. What has changed is that a new, largely invisible layer now sits between a company and the people forming opinions about it, and that layer runs on different rules than the ones I spent twenty years learning.

The engines are already describing your brand

Research tracking how AI systems characterize specific brands has started to surface some uncomfortable patterns. Comparative studies of how engines describe rival brands in categories like beauty retail have found that AI answers frequently default to a small number of well-documented narratives about each brand, narratives that were shaped years earlier by whichever companies did the deliberate work of publishing structured, citable, well-sourced content about themselves. The brands that did not do that work do not disappear from the answers. They simply get described in whatever terms happen to dominate the secondhand coverage the engines could find, for better or worse.

This is a different problem than traditional reputation management was built to solve. A negative news cycle used to fade, measurably, as new stories pushed old ones down a search results page and eventually out of most people's attention entirely. An AI engine does not experience time that way. It retrieves from whatever corpus it was trained on and whatever it can access at query time, and a well-documented narrative from three years ago carries essentially the same retrieval weight as one from three weeks ago unless something newer and more substantive has displaced it. Reputation, in this layer, is closer to a standing record than a fading headline.

Real people have found this out the hard way

Public figures and executives have already discovered how literal this problem can get. When an AI system is asked to describe a public figure's business record, its politics, or a controversy they were involved in years earlier, it often produces an answer built from whatever secondary sources dominate its training data, sources the person in question may never have had a chance to respond to, correct, or contextualize at the time they were published. The correction cycle that used to exist, where a person could give an interview, publish a rebuttal, or simply wait for a news cycle to move on, does not map cleanly onto a system that is synthesizing an answer from a frozen snapshot of the internet.

This is not a hypothetical concern reserved for celebrities. It applies just as directly to a founder whose only substantial press coverage is a single unflattering article from years ago, a company whose Wikipedia page has not been updated since an acquisition, or an executive whose most detailed profile online is an old, since-resolved lawsuit. The AI engine does not know the story has moved on unless something has actively told it so, in a form the engine can retrieve and trust.

What actually shifts the record

I want to be precise about what does and does not move an AI engine's characterization of a person or a company, because the wrong intervention here can waste real time and money.

Volume alone does not do it. Publishing ten thin blog posts about how great a company is will not meaningfully shift how an engine describes it, because the engines are weighting substance and third-party corroboration, not repetition. What has shown measurable effect, based on research tracking these patterns across several categories, is a combination of three things happening together: structured, factually dense content that names specific people, dates, and outcomes rather than vague characterizations; independent, credible third-party validation, whether that is trade press, an analyst report, or a named expert's public commentary; and sustained cadence rather than a single push timed to a launch or an anniversary.

The Wikipedia entry matters more than most communications professionals currently treat it. Structured, well-sourced Wikipedia content is disproportionately weighted by several major AI engines as a canonical reference point, which means an outdated or thin entry is not a minor cosmetic issue. It is actively shaping what an AI system tells millions of people about a person or a company, every single day, whether anyone at that company has looked at the page in five years or not.

The correction problem is real and mostly unsolved

Here is the part of this that should concern anyone whose job involves protecting a reputation. The traditional playbook for correcting a false or outdated narrative, the op-ed, the clarifying interview, the follow-up story, still works for human readers who encounter it directly. It does not automatically propagate into how an AI engine will describe the same subject six months later, because that requires the correction itself to become part of the citable, retrievable record the engine draws from, not just a piece of content that exists somewhere on the internet.

This means the discipline of reputation management now includes a step that did not used to be necessary: actively verifying, on a recurring basis, how the major AI engines currently describe a person or a company, and treating any material gap between that description and reality as seriously as a factual error in a major news story. Most executives have never done this. Most communications teams have not built it into their standing workflow. The ones who have are finding outdated, incomplete, or flatly wrong characterizations sitting in plain sight, quietly shaping how a growing share of the public forms its first impression.

The generational split nobody is planning for

There is a demographic dimension to this that deserves more attention than it gets in most boardrooms. Younger consumers and younger professionals are, by a wide and growing margin, more likely to ask an AI system a direct question about a brand or a person than to read a traditional article about them, and that gap widens every year as a new cohort enters the workforce having never known a research process that did not start with a conversational query. For a company or an executive whose reputation strategy still centers on securing the right feature story in the right outlet, this is not a minor shift in distribution channels. It is a shift in which audience is even reachable through the old method at all.

I do not think this means traditional media relations becomes irrelevant. Credible journalism still shapes the underlying record that AI systems draw from, which means earned media remains one of the most effective ways to influence what an engine eventually retrieves. What it means is that earned media has to be understood, now, as an input into a second, largely invisible layer of reputation formation, rather than as the finish line it used to represent. A brilliant profile in a major outlet that never gets structured, cited, or reinforced in a form the engines can retrieve will shape the opinions of everyone who reads it directly, and almost nobody else.

The uncomfortable truth

Reputation management has always been about controlling a narrative that is, ultimately, out of any single company's or person's full control. What is new is the intermediary. A journalist could be persuaded, corrected, or held accountable through a correction. An AI system cannot be pitched a story or invited to lunch. It can only be given better material to retrieve, published consistently, sourced credibly, and structured in a way the system can actually use. The companies and executives who understand that distinction early are going to spend the next five years quietly shaping how they are described to an audience that increasingly does not read articles at all. It reads answers, generated in a fraction of a second, about a reputation that took years to build and that almost nobody is currently checking.

Reputation now compounds inside AI engines the same way it once compounded in the press. 5W runs AI Search (GEO) programs for brands across consumer, B2B, financial services, healthcare, and technology, building the machine-readable footprint that gets brands cited, not just ranked. Learn more at https://www.5wpr.com/practice/geo-optimization.cfm.

Ronn Torossian is the founder and chairman of 5W Public Relations, one of the largest independently owned public relations firms in the United States.