For two decades, reputation management had one job: own the first page of Google.
The mechanics barely changed. A brand or executive had a problem. A firm built positive content, optimized it, and worked it up the rankings until the negative results slid to the second page, where click data confirmed almost nobody looks. The industry called it suppression, and for a long time it was the entire product.
The New York Times investigation into Terakeet is the clearest public look yet at how that product is built. It also marks the point at which the model became a liability rather than a strategy. Here is why — and what is taking its place.
Why suppression became a liability
The Terakeet case makes the first reason concrete: suppression only works unobserved. The model depends on manufactured content not being recognized as such. Once the method is documented, the work becomes the headline, and a manageable problem becomes a permanent one.
The second reason is structural. Suppression is a tactic that functions inside one specific interface — an ordered list of links, with attention concentrated at the top. That interface is losing ground. AI Overviews now sit above Google's results. ChatGPT, Claude, Gemini, and Perplexity skip the list entirely. When a rising share of research begins with a tool that returns one composed answer rather than ten links, "move the bad result to position 14" stops being a meaningful objective.
How AI tools handle a query about you
The shift most of the industry has not absorbed: an AI system does not just rank pages and leave you to draw the conclusion. It tends to draw the conclusion itself. Asked about a person or company, it pulls from a range of sources, weighs them for authority and consistency, and composes an answer.
That changes the problem. Burying a negative result accomplishes little if the system still reads it. Planting a flattering profile accomplishes little if the system does not treat the source as credible. The lever is no longer position. It is which sources the system relies on, and what those sources say.
What replaces suppression
The replacement is not concealment. It is becoming a source the systems rely on.
This is what Generative Engine Optimization describes, without the jargon: building genuine, authoritative, well-structured, broadly corroborated information across the places AI systems read. The aim is to improve a brand's citation share — how often, and how favorably, the systems reference it when asked.
Unlike suppression, this work does not collapse when it is observed, because there is nothing to conceal. Making accurate, credible information easy for an AI system to find and trust holds up in daylight. That is the difference that matters.
The objection — and why it falls short
The standard pushback: Google still drives enormous traffic; the blue links are not dead.
True, and not the point. Google itself increasingly answers queries with an AI Overview placed above the links. For many users, the composed answer is read first, and often it is the only thing read. Optimizing for position three of a list that a growing share of users skip is optimizing for shrinking ground.
The lesson of the Terakeet case
The Terakeet story is being read as a story about ethics. It is also a story about timing. The firm invested 20 months and tens of millions of dollars mastering an interface that is being de-emphasized beneath it.
The takeaway for any brand watching: the reputation infrastructure that matters is not a campaign run during a crisis. It is a credible-source position built before one. Build the infrastructure before the crisis — not during it.
