The moment things shifted for a client of mine arrived without drama or warning. It came in the form of a short email, casually forwarded by a colleague, carrying a voice message that sounded as routine as any other. Executives send dozens like it while travelling. The voice was recognisably his, the tone familiar, but the comment he appeared to make about being ‘deeply concerned’ over an internal decision was something he knew he had never said.
He played it three times before admitting that he hadn’t said it because the call never happened. Yet the message felt eerily close to something he might have said, and that was the part that unsettled him most. It wasn’t obviously fake. It wasn’t obviously real. It sat in a grey space where doubt grows quickly.
This is what makes AI-driven impersonation so difficult to navigate. It doesn’t need to be perfect. It only needs to be plausible. And once something feels plausible, it can quietly shape perceptions long before anyone questions its authenticity.
I see this pattern repeatedly now. Executives receive voice notes that sound just familiar enough to pass casual scrutiny. Photos appear online that blend real features with small, invented details. Clips circulate in private groups, not to create a public scandal but to plant a seed of uncertainty among the people whose opinions matter most. These messages move quietly through informal channels, which makes them difficult to track and surprisingly influential.
The more visible form of harm sits in the headlines: deepfake scams, synthetic blackmail attempts, impersonation attacks on founders and CEOs. But the harm that disrupts careers today often comes from smaller distortions. An AI summary that misstates a person’s role. A biography that invents an investment or misinterprets a quote. A factual error repeated so many times that it becomes part of the person’s digital identity.
One client recently discovered that a widely used AI tool had combined two unrelated articles about him, then confidently produced a version of his career that he barely recognised. The summary looked authoritative, so people repeated it. Before long, the fictional details surfaced in introductions and briefing notes. By the time he noticed, the misinformation had already spread to several places online.
This is the part many people underestimate. AI hallucinations do not look like lies. They look like confident answers, and confident answers spread quickly.
Traditional reputation management is not equipped for this landscape. In the past, if something inaccurate was published, you contacted the source, corrected the error and moved on. Today, misinformation does not always come from a person. It comes from systems that aggregate, guess, stitch and reinterpret. A correction does not reach all the places the error has travelled. It does not reach the people who saw it privately. It does not reach the small professional circles where first impressions are formed.
This is why proactive identity monitoring has become essential. Not because people should be paranoid, but because their digital presence is now generated from far more sources than they control. Some clients I support now check their AI summaries with the same regularity they check their financial statements. They notice when a phrase sounds unfamiliar or when a new platform seems to know too much. They treat these small signals not as noise but as the earliest signs of distortion.
One founder spotted a fabricated detail on an obscure directory site he had never used. It claimed he had stepped down from a role he still held. A tiny error, easy to ignore. But a month later, the same claim appeared in an automated briefing document generated for a potential investor. The mistake had been copied somewhere along the line and repackaged as fact. Had he not caught it early, the misinformation could have influenced negotiations long before he realised what had happened.
This is what AI harm looks like in practice. Not always dramatic. Often subtle. A gradual bending of the truth until the real version of a person becomes harder to locate.
Businesses are beginning to adapt. Some are developing digital fingerprints for key individuals, mapping where their name, image and voice appear so they can detect when something falls outside the pattern. Others are introducing internal protocols that treat unusual messages or misaligned summaries as signals worth examining rather than dismissing. It is not fear that drives these efforts. It is recognition that identity today is porous, and that the boundary between what is real and what simply looks real is thinner than ever.
The rise of AI harm is not a temporary phase. It is reshaping how trust is formed. It is changing how quickly reputations can be influenced, and by what. The people who navigate this well are not the ones who react the fastest, but the ones who observe the small shifts early and address them before they gather weight.
The truth has not disappeared. It has simply lost its automatic authority. Now it requires protection, context and deliberate tending.
Those who understand this will be the ones best equipped for a world where personal identity can be reconstructed, convincingly and quietly, by anyone with the motivation to do so.
Guest Post
About the Author: Sarah Willis is the founder of a discreet digital identity and reputation protection service for high visibility individuals and UHNW families. She specialises in AI era risk, online privacy and the quiet signals that shape modern reputation.
