Editor's Introduction
For most of the last two decades the object of online reputation management was a page. Ten results, ranked by a system that did not read, did not reason and held no view on whether the material in front of it existed for any purpose other than to occupy the position it held.
That system has been replaced by one that forms a view.
When a generative model is asked about a person it does not return a list. It returns an account. It reads across everything available, reconciles the sources against one another, and produces a single narrative in which each element has been weighed against the rest. In doing so it necessarily forms an impression not only of the subject but of the body of material surrounding the subject: how much of it there is, where it came from, when it appeared and whether it says anything.
That last capability is the one the industry has not yet absorbed, and it is the subject of this volume. The exposure is not confined to obvious fabrication. It extends to the entire displacement model on which a great deal of this industry still runs.
The Case That Prompted It
We were approached some time ago about a difficult reputation problem, prepared a proposal, and were not engaged. That is a routine outcome and we thought no more about it.
We looked at the search results some time later out of professional interest, as we often do.
What had been built was a second identity. New sites under the same name, carrying a biography unconnected to the actual record, describing a different career and a different set of interests. The apparent intention was dilution: enough same-named material to crowd out what was already there. We said at the time that we regarded the approach as high risk and that it was not something we would have undertaken. That was the end of our involvement.
We have since seen where that leads. An AI answer on the subject did not simply restate the original history. It described the newer profiles as self-published material that appeared intended to downplay a prior record, and characterised the alternative biography as an attempt to manage the subject’s online reputation.
The strategy had been read, understood, and written into the record as evidence of intent.
That is the unambiguous end of the spectrum, and it is easy to set aside as an unusual case. The more consequential point is that the faculty which recognised a fabricated identity does not require fabrication in order to operate.
Displacement Is Legible
Consider what a conventional suppression programme produces.
Twenty or thirty owned properties carrying variations of the same biography. Directory profiles across every platform that accepts one. Wire distribution that places an identical release on forty syndicating domains. Contributor placements on outlets that sell access. A personal site, a name-matched domain, a portfolio, a speaking page. All of it accurate, in the narrow sense that nothing in it is false. All of it created within a few months of one another. All of it published, directly or indirectly, by the subject.
To a search engine that was a set of documents to be ranked, and the strategy worked because ranking was the whole game. To a model asked who this person is, it is a single object with a describable shape, and the shape is the finding.
The signals are structural rather than semantic, which is why accuracy provides no protection against them.
Duplication. The same two hundred words of biography appearing across thirty domains is not thirty sources. It is one source, republished. Models weight corroboration heavily and are competent at recognising that apparent corroboration is an echo of a single origin. A footprint that looks broad from a distance collapses to a single point on inspection.
Syndication. Wire copy is identifiable to anyone who has seen it twice, let alone to a system trained on the entire genre. Identical text, identical headline, forty domains, one date. Many of these placements are additionally labelled as sponsored or as press releases in their own markup, which means the disclosure the client never reads is machine-readable.
Clustering. Where an entire footprint post-dates a single event, the sequence is legible and the inference is short. Nothing about the material has to be untrue for its timing to explain its existence.
Non-specificity. A great deal of commissioned reputation content contains no falsifiable claim of any kind. Visionary leadership, a passion for innovation, a commitment to excellence. Content that asserts nothing checkable carries no evidential weight, and where a body of material is uniformly unfalsifiable, the absence of substance is itself information about why it was made.
Silence. A footprint that is exhaustively positive across forty properties and does not acknowledge a documented public event anywhere is not a neutral omission. Real biographies are uneven. A uniform one is an artefact.
Procedural record. Suppression conducted through process leaves its own trail. Delisting requests are noted at the foot of the results. Takedown demands are published in full, with the removed URLs, in public databases that are indexed and read. Conflict-of-interest editing on Wikipedia is logged on talk pages which are themselves crawled and cited. Legal threats to journalists have a long history of becoming the story. Each of these is a public record of an attempt to remove a public record.
None of this is new evidence. All of it was visible to any competent analyst for years. What has changed is that the evidence now has a reader, and the reader answers questions from anyone who asks.
Truth Is No Longer The Test
This is the part worth sitting with, because it disposes of the defence most of the industry currently relies on.
The standard justification for a displacement programme is that everything published is true. No fabricated identity, no invented credentials, no false claims. Just a large volume of accurate positive material, published to move something down. That defence answers a question nobody is asking any more.
A model assessing a corpus is not only asking whether the statements in it are correct. It is accounting for why the corpus exists. And a body of truthful material that has no reason to exist other than to occupy space is structurally distinguishable from a body of material that exists because someone had something to say, or because an independent party judged it worth publishing.
Content built to rank looks different from content built to be read. It is more numerous, more uniform, more recent, less specific, less independently sourced and less varied in register. Those properties are not opinions about the content. They are measurable features of it.
The line no longer falls between true and false. It falls between material with a reason to exist and material with only a function.
A truthful biography syndicated sixty times is not a lie. It is legibly a suppression asset, and it will eventually be described as one. If suppression is the only objective of a programme, that objective is now discoverable from the programme’s own output, regardless of how honest the individual pieces are.
Why The Response Costs More Than The Event
A historic regulatory matter, however serious, is a closed event with a date attached. It can be contextualised, set against subsequent conduct, and allowed to age. Sophisticated readers, and there are more of them than most subjects assume, are capable of separating what someone did a decade ago from who they are now.
A visible concealment effort is not a closed event. It is present tense, it is voluntary, and it speaks to disposition rather than circumstance.
We work regularly alongside banks, family offices, law firms and corporate intelligence teams. Their tolerance for an old finding is considerably higher than clients expect. Their tolerance for evidence of an active attempt to obscure one is close to nil, because that evidence is not a fact about the past. It is a fact about how the subject behaves under pressure now, which is precisely the question every counterparty assessment is trying to answer.
An old finding tells a due diligence team what happened. A manufactured footprint tells them what to expect.
What Survives Scrutiny
The material that holds up under machine reading is the material that would hold up under a competent journalist, and it has a small number of characteristics.
It is independent. What a credible third party publishes about someone carries weight that self-published material cannot, and it is slow to obtain, which is exactly why it counts.
It is specific. Verifiable claims, real detail, positions that could in principle be checked and found wanting. Falsifiability is what gives a statement evidential value.
It is proportionate. Fewer, better, distinct pieces with individual purpose, rather than volume for its own sake. Thirty near-identical assets are weaker than three substantial ones, not stronger, and they now carry a cost the three do not.
It is coherent with the record. A footprint that engages with a subject’s actual history, rather than constructing an alternative one around its edges, does not create the contradiction that invites comment.
It has a reason to exist. Genuine contribution to a field, a real position on a real question, work the subject has actually done. Material of that kind survives because its purpose is not exhausted by its ranking.
None of that is suppression. Where displacement occurs it occurs as a consequence of having built something with independent value, not as the objective.
The Economics
It is worth being direct about why the cheaper quote is cheaper.
Volume publishing is a production task. It can be executed quickly by people who need no understanding of the subject, the sector or the institutions that will eventually read the output. There is no research, no editorial judgement, no verification and no assessment of risk, because the practitioner does not expect to be present when the consequences arrive.
Positioning is an editorial task. It requires knowing what is defensible, what is substantiable, what withstands a journalist or a regulator or a model, and what is better left alone. The cost difference reflects a difference in what is being made.
A suppression programme is not the same outcome at a lower price. It is a second problem placed on top of the first, and in the current environment it is the second one that ends up being described in the answer.
Pavesen Perspective: The Objective Has To Be Something Other Than Removal
The instinct behind suppression is understandable. Someone facing a permanent searchable account of the worst period of their professional life wants it gone, and there will always be a provider willing to sell that feeling.
What has changed is that the environment no longer merely retrieves. It interprets, and what it interprets includes the reputation effort itself. Google, ChatGPT, Gemini, Claude and Perplexity are not ranking documents about a person. They are constructing an understanding of one, and that understanding will accommodate both what the person has done and what the person appears to have attempted.
We do not regard this as making the work harder. It removes a category of practice that was never serious and rewards the firms that were doing something else.
A programme has to be built around a proposition rather than an absence. What is genuinely true about this person, what are they credibly known for, what have they actually done, and what material would exist for those reasons even if the search results were clean. Everything that follows from that question tends to survive scrutiny. Very little that follows from the other question does.
A career built over several decades should not be reduced to a single episode. It will not be rescued by burying it, and in the current environment the burial now sits alongside the episode rather than in front of it.
Where we begin is the Pavesen Audit: presence, perception and exposure, assessed across search and the AI systems together, so that any decision about what to do next is made against what is actually there.