Across fifteen years in the industry, I have heard the request put countless ways. It is rarely the first thing said. It arrives late in the introductory call, after the background story, the context and the careful framing, and it amounts to the same sentiment every time. “Make it go away.”
The methods vary, but three approaches recur in campaigns I have been asked to review after work elsewhere has gone wrong. Having unfavourable results removed. Swamping search results with filler content. Reinventing the person somewhere far from the problem. Removal, suppression and repositioning each have their place when properly executed. Applied as quick fixes, each can leave a client with a new problem as well as the original one. By the time I see them, the consequences are usually deep-rooted.
The three cases that follow are drawn from publicly reported events I came across whilst researching.
The Takedown That Rebounds
The first of the three is removal, the most direct and the most often requested method. Earlier this year a news outlet published an investigative piece about a digital firm that was publishing low-quality AI-generated content to its network of sites at scale. Days later the investigative article had vanished from Google, following a copyright complaint claiming it had copied an article from another site. The complaint turned out to be bogus, and Google restored the investigative piece the following day.
The investigative article was also updated to include additional content about the failed removal. The record of this removal attempt is now a further stain on the digital firm's search and AI results.
Worse was to come. The coverage flooded in, as per the pattern commonly referred to as ‘the Streisand effect’. Further news outlets wrote about the failed takedown. An unfavourable story that had run once, in front of a single readership, was republished across a growing number of titles, attached to something more memorable than the original allegation: that somebody had tried to have it removed.
One thing separates what happened there from a legitimate removal. Removal on proper grounds is ordinary legal process. Material that genuinely defames, that infringes copyright somebody actually owns, or that a regulator agrees should not surface against a person's name, comes down through a process built for it, and leaves behind a record of a legitimate request. The damage comes from removal on false grounds, because there the grounds are themselves the evidence. A backdated copy, an invented company, an order that was never made. The archive does not record that something was taken down. It records the attempt to do so.
Removal Attempts Leave a Footprint
The record outlasts the removal because it is published to a public ledger. A complaint sent to Google is forwarded for inclusion on the Lumen Database, a research archive at Harvard that records who sent it, when, what they claimed to own and which pages they wanted removed. It verifies none of it. Google then prints a line at the foot of the results page saying results have been removed following a complaint, and links to it. That line appears wherever the removed page would have ranked when a search is run for a principal or a company. Even a successful removal is not a clean absence. It is a visible gap with a footnote, and that footnote leads to a record.
The archive is also searchable, and fraudulent complaints are rarely sent one at a time. Lumen's own researchers traced a single campaign of around thirty-four thousand of them, sent under thirty different names against five hundred and fifty websites. In another case a journalist found his own article reported as stolen: somebody had copied it, backdated the copy, and then complained that his original was the forgery. Searching the archive for one unusual phrase from that complaint turned up a hundred and eighty more worded identically.
None of this is confined to copyright. Of seven hundred court orders sent to Google between 2012 and 2016 and later examined, close to two hundred were forged or highly suspicious. One man was jailed for nine months for forging a judge's signature.
Volume Alone Cannot Reach the Top
The second method is suppression: pushing the unfavourable material down the rankings. The problem is not the displacement itself, and it is not scale either: a large programme of considered, on-brand material is often exactly what the job needs where a judgment or a regulatory notice cannot be withdrawn. Building above it is the only answer available, and the question is never whether to displace but what you displace it with. What fails is displacement done cheaply: produced at speed, to no real brief, with nothing of substance behind it.
A presence built from credible, substantial material can hold a solid position over time rather than a short-term fix of only a few months, weeks or even days.
In many of the examples I have seen running at that kind of scale, the material is almost entirely AI generated. That is not an argument against the tools. These systems are useful and there is nothing wrong with using them intelligently. What matters is how. The difficulty begins when material is produced almost entirely by AI, with little human oversight, which is where a footprint is left open to scrutiny. Google does not currently say it penalises material for having been produced by a machine. What it acts against, under a policy introduced in 2024, is scaled content abuse: large volumes of regurgitated pages that say little beyond what is already available elsewhere, and exist to capture rankings rather than to inform anybody.
An analysis of around a million pages this year found that pages largely composed of machine-generated text were indexed less often and drew two to three times fewer impressions. A study of forty-two thousand articles across twenty thousand search terms found the first position around eighty per cent likely to be human-written and around ten per cent machine-generated, with the difference closing by the fifth position. Machine output converges. Ask a hundred writers for a profile and you get a hundred versions; ask a model the same question repeatedly and the answers share a register, a structure and a vocabulary. The first position is more likely to reward material that only somebody with actual knowledge of the subject could have produced.
You cannot unpublish a judgment. You can out-rank it, but only with material good enough to hold the position. A principal is not helped by new material sitting at position six while the unfavourable content holds first place. It holds first place because it is a primary source with a named author at a publication with a reason to be trusted. Taking that position requires material of comparable authority and relevance, neither of which follows from volume alone or from low-quality material.
That is Google's position today, not a guarantee for the future. Policies are rewritten without notice, and the law is moving faster than the policy: Europe's transparency rules, which applied from August, already require generative systems to mark their output in a form a machine can read. What passes today may not comply with the rules, or satisfy the algorithms, tomorrow, which is worth considering when building a reputation designed to last.
The Answer Describes the Attempt
The third method is repositioning. Repositioning is ordinary and often necessary: people's interests broaden, their work moves on, and a profile that no longer reflects what somebody actually does is worth rebuilding. What fails is repositioning chosen for distance rather than truth - a field the principal has no standing in, picked because it sits far from the unfavourable material. Every fact in it may be true. The story it tells is not.
I went looking for a live example and did not have to look hard. A man with a regulatory history, now publishing content on an entirely different subject to his expertise, with the new material ranking well against his name. By the measure most campaigns are sold on, with the unfavourable content no longer visible on Google's first page, the job appeared successful.
I then asked an AI answer engine about him. The summary presented the new positioning alongside the unfavourable material and described the new material as an attempt to move away from it. It told the reader, in effect, that somebody had been at work on this and that the result looked like a cover-up. The attempted change had become part of the summary, which is a worse outcome than having done nothing at all.
What a Long-Term Plan Looks Like
What works instead of a quick or cheap fix is a well thought out reputation strategy. Establish what the record actually says now. Identify the handful of pages actually carrying it, in search results and in AI answers alike. Make sure the corrective facts exist somewhere authoritative and citable. Then rebuild the rest of the record legitimately and patiently: the background, the career, the achievements, the decades that produced the position in the first place, in outlets that would naturally run it anyway.
That is still displacement. It is still personal brand building too. The difference is that the material is worth reading and true of the person, which is the only version that lasts.
What survives is material with an author who can be held to it, published somewhere that had a reason to publish it. Producing it has always taken time, skill and judgement, which is precisely why it still works.
Timing decides how much there is to fix. Left untouched, an issue sitting on a primary source gets cited by other sources and by the engines, spreading further as time passes.
A reputation built to last is not the one with the fewest unfavourable results on the day the engagement closes. It is the one that would survive somebody examining how it was built.