The communications director arrives Monday morning, opens the monitoring dashboard, and finds everything looking reasonably calm.
Mentions are within range. Social conversation is quiet. No headline requires a meeting with Legal before the second cup of coffee. Sentiment sits in that reassuring shade of green everyone likes to put on a large screen.
Then someone asks a much less sophisticated question:
“What does ChatGPT say about us?”
Ten seconds later, a summary appears that no one in the room approved. It blends the official company description with a three-year-old controversy, two recent news stories, third-party opinions, and a conclusion about the firm that appears in none of its own materials.
The dashboard was still green.
The reputation, perhaps not.
Two years ago it still made sense to explain that artificial intelligence could draft replies, score sentiment, personalize messages, spot trends, and produce content. All of that remains true. It is also the least interesting part.
The change worth watching is not that a company can write a press release in thirty seconds. It is that a growing share of customers, journalists, candidates, investors, and other publics can form an opinion of that company without ever reading the release.
AI reads it for them.
And it reads everything else.
From tool to intermediary
For a long time, corporate communications ran on a fairly stable model. The organization built a message. It sent that message out through owned channels, the press, advertising, social platforms, or other people. Then it watched what happened.
There was never total control. There were journalists, search engines, algorithms, critics, angry customers, and the man on Facebook who has been typing in all caps since 2009. But the raw material still moved through the world in recognizable ways.
Generative AI changes that architecture. It does not merely decide what we see first. It can take information apart, cross it with other sources, summarize it, and put it back together before handing it over as an answer.
That changes the job.
Ask which firm to hire, which hotel to book, which company looks trustworthy, which product has better guarantees, or which brand is in reputational trouble, and a system can now return a finished answer without anyone visiting ten sites, opening five tabs, or reading a single newsroom page.
The organization is no longer only the sender.
It has become raw material.
And communications increasingly begins before the click.
Reputation now has a synthetic layer
This may be the hardest shift to absorb, because it happens outside the channels we own.
For years, reputation work meant asking what the press was saying, what showed up on Google, what was happening on social media, and what people remembered after a campaign.
Now there is another question: What does an AI system reconstruct when someone asks about us?
That is not a technical curiosity.
Recent research on AI search has found that different models can build materially different reputational profiles of the same company, because they do not use the same sources or weight them the same way. There is no single “AI version” of your brand. There may be several.
And showing up is not enough.
Burson has documented a second, more uncomfortable distinction: visibility is not credibility. A company can appear often in generative answers and still fail to sound convincing to the people reading them.
We used to fight to be found. Now we also have to be understood — and believed.
Communications is no longer only a distribution problem. It is an architecture problem: what public information exists, where it lives, which sources support it, what contradictions we have left scattered across the internet, and what version of the company a system can rebuild without ever sitting in one of our meetings.
Synthetic reputation does not replace real reputation.
It interprets it.
And a bad interpretation can travel very fast.
Producing more is no longer the advantage
In 2024 it still felt remarkable that a model could write copy, generate images, summarize documents, or turn one piece into several formats. Today those capabilities fit inside a monthly subscription.
For decades, a large share of the cost of communications sat in production: writing well, designing, editing video, analyzing large volumes of information, personalizing at scale. All of that took time, people, and budget. Much of it no longer does.
That is good news. It is also a warning. When something stops being scarce, it stops being much of a differentiator. If every organization can generate fifty headlines, a hundred images, and twenty versions of a campaign before lunch, the advantage cannot be producing a fifty-first.
Value moves to what cannot be generated in bulk: what deserves to exist, for whom, with what intent, which information should prevail, and what tone actually belongs to the organization. It also moves to what should not be published.
Execution gets cheaper. Judgment gets more expensive.
The risk is not only that AI will produce bad content. It can also produce content that is extraordinarily competent. The internet can fill with messages that are perfectly able, perfectly interchangeable, and easy to forget.
Abundance does not end the need for communications.
It ends the premium we once paid merely to produce it.
The shift that matters now is not only one of quality. It is a change of verb. In 2024 we talked a great deal about chatbots: a question, an answer, and that was that. An agent does not only reply. It can research, compare, use tools, and complete a process within a set of permissions. The next visit to your website may not be human. The next conversation about your company may take place between two systems, before anyone on your side has had a chance to deliver the polished speech about “end-to-end solutions.”
A prettier page will not fix that. The page will still matter. So will the story and everything that makes a brand recognizable. Underneath it, though, you need a second kind of legibility: structured information, services described without ambiguity, data that does not contradict itself, and sources a system can consult without having to guess.
Inside organizations, a less elegant paradox is playing out. Everyone says they are transforming. Far fewer are. BCG’s global CMO survey this year puts numbers on it: 96 percent say AI is transforming their function end to end, yet 42 percent still use it as an assistant for discrete tasks. Only about a third have moved in earnest to agentic workflows. Eight percent are already running campaigns with several autonomous agents.
Both things can be true. They also describe the moment. Buying licenses is easy. Changing how an organization works is not. The tool usually trips over the same wreckage: two departments keeping different versions of the same facts, a message no one can say who approves, the important knowledge locked in three people’s heads, and seven ways of describing the same service. AI does not create that mess. It finds it. And if we let it, it automates it.
That is why the work that matters is no longer generating ten posts before the meeting ends. It is the system underneath. What information we treat as authoritative, what an agent may publish without review, what must go to a person, and who answers when the system gets it wrong. Someone has to keep the authority to say no.
A crisis no longer begins with the headline
Crisis communications is changing too.
The old model, simplified, was to spot a signal, judge its severity, prepare a response, and send it out. That is still necessary. But the signals now arrive earlier, and from more places.
Narrative-intelligence systems can find patterns in scattered conversations. Generative engines can pull up an old controversy when someone asks a new question. A false claim can enter community sources and later return dressed as an automated recommendation.
Gartner has put reputational risk intensified by AI search, falling trust, and fast-moving narratives among the communications priorities for 2026.
The danger is not only that someone says something false about a company. It is that the falsehood becomes another piece of material from which other systems rebuild the company. At that point the crisis statement arrives too late if everything underneath it is still a mess.
Reputation begins long before the statement: in the data, the sources, the consistency, what other people say, and in everything the organization never explained because, for years, no one thought it needed explaining.
What remains human
There is a familiar temptation when writing about these tools: to close by saying “the human touch will matter more than ever.”
It is comforting.
It has also become one of the most worn templates of the AI years.
I would put it differently. We do not need people in the loop for sentimental reasons. We need them because someone has to answer for what happens.
A machine can propose a strategy, write the message, spot a trend, weigh thousands of signals, and even carry out part of the response. That is no longer the argument. The argument is who decides what the organization considers true, what it will defend, and what consequences it is prepared to own.
That is not nostalgia.
It is governance.
In 2024 the large question seemed to be how much artificial intelligence could help corporate communications. Two years later, the question has changed. AI already helps. A great deal. What we have to decide now is the place we give it inside the systems that build reputation, interpret information, and act in an organization’s name.
We used to decide what to say. Now we also have to decide what we will allow machines to understand, recommend, and do on our behalf. The Monday dashboard can stay green. Someone still has to answer for what ChatGPT just said about the company.
