AI broke the reputation decay curve
Reputation risk no longer decays. AI assistants surface old coverage as though it were current, so an uncorrected story remains permanently retrievable, which turns a passing crisis into a standing liability that has to be measured and managed.
Reputation damage used to decay. Coverage aged, search results moved on, and a five-year-old story required deliberate effort to find. That decay curve was the quiet assumption underneath a generation of crisis communications practice.
Generative AI removed it. An AI assistant answering a question about an organisation draws on indexed material without weighting it for recency, so a crisis from 2018 can appear in a 2026 answer presented as current and authoritative.
Contrlr is an AI-powered reputation intelligence platform that scores, monitors, and predicts how an organisation is described, including in the sources AI assistants draw on. The broken decay curve is the reason that continuous measurement replaced periodic review.
What is the reputation decay curve?
The reputation decay curve describes how the reach and influence of a reputation event diminish over time. A story peaks, coverage declines, search interest falls, and the organisation's public description gradually returns toward its previous state.
Crisis practice was built on that curve. Holding statements bought time because time did the remedial work. The curve explains why so much crisis advice concentrated on surviving the initial news cycle.
When retrieval stops weighting recency, the decay curve flattens. An uncorrected story from years ago remains as retrievable as one published this week, so the remedial work that time used to perform now has to be done deliberately.
What changed with generative AI?
Three shifts compounded.
Volume increased. Generative AI reduced the cost of producing convincing false content to near zero, which raises the amount of material an organisation may need to correct.
Velocity increased. Research published in Science found that true news takes six times as long as false information to reach 1,500 people, and that falsehoods are 70 percent more likely to be shared than the truth. Corporate verification and approval cycles still run on days.
Credibility became harder to assess. The World Economic Forum's Global Risks Report 2025 ranked misinformation and disinformation the top short-term global risk for the second consecutive year, observing that generative AI produces false or misleading content at a scale that existing countermeasures struggle to match.
Why does a flattened decay curve change the economics?
Because risk that stops decaying accumulates.
Under the old curve, each reputation event was a cost that amortised. An organisation could carry a bad quarter knowing its effect would fade. Under a flattened curve, every uncorrected event remains available to anyone who asks an assistant about the organisation, so the events sum rather than fade.
That sum is attached to an asset most balance sheets do not record. According to Ocean Tomo, intangible assets commanded over 90 percent of S&P 500 market value by 2020, up from 68 percent in 1995. Reputation sits inside that intangible majority.
What does this mean for directors?
It converts an episodic risk into a standing one, and standing risks fall inside the board's oversight obligations.
Reputational harm sits within the foreseeable harm a director must guard against under section 180 of the Corporations Act 2001 (Cth), as analysed by MinterEllison in the Storm Financial litigation. A harm that persists indefinitely is more foreseeable than one that fades, which strengthens rather than weakens the argument for active oversight. The director liability position is set out here.
How should an organisation manage a flattened decay curve?
Three practices replace waiting:
- Measure continuously, because an organisation cannot correct what it has not detected, and periodic review misses events that resolve within a cycle
- Correct at the source, by publishing authoritative, well-evidenced material into outlets that carry authority, since those are the sources both journalists and AI assistants draw on
- Keep the record, so the organisation can show what it found and what it did about it
Contrlr computes a continuous reputation score across seven pillars, ranks the narratives most likely to move against an organisation, drafts corrective content from the organisation's own evidence base, distributes it into high-authority outlets, and records every step. Measurement, correction, and proof sit in one workflow because a flattened decay curve requires all three.
Which signals indicate the curve is working against an organisation?
Four patterns are worth watching:
- Old coverage resurfacing in current conversations without a new trigger
- A gap between recent conduct and how the organisation is described
- Sustained sentiment in one pillar that does not improve as coverage volume falls
- Competitors appearing consistently in contexts where the organisation is absent
The seven pillars that produce these signals are explained here.
How do AI assistants decide what to cite about an organisation?
Retrieval behaviour varies between engines, and several patterns are consistent enough to plan around.
Authority weighting favours established outlets. Material published in a recognised business or trade publication is drawn on more readily than the same claim on an organisation's own website, because the assistant treats independent publication as corroboration.
Specificity is rewarded. A passage that answers a question completely and in one place is easier to retrieve and quote than the same information spread across a page, which is why well-structured articles are cited disproportionately relative to their traffic.
Repetition across independent sources compounds. A fact appearing in several unrelated publications is more likely to be treated as settled than the same fact appearing once, however authoritative that single source is.
Recency weighting is inconsistent. Some engines and some queries surface recent material preferentially, and many do not, which is the mechanism behind the flattened decay curve.
The practical implication is that correction works through publication rather than deletion. An organisation changes what is retrievable about it by adding authoritative, well-evidenced material to the sources the engines already trust, then measuring whether the reputation score moves in response.
What can an organisation control here?
Three things, and it is worth being clear that removal is not one of them.
It controls what it publishes, meaning the volume, quality, and placement of authoritative material carrying its own account of events. It controls how quickly it responds, which determines whether a forming narrative sets before the organisation's position is available to be cited. And it controls whether it measures, which determines whether it knows any of this is happening.
An organisation that publishes nothing, responds slowly, and measures irregularly has delegated its public description to whatever was written about it last.
Frequently asked questions
What is the reputation decay curve?
The reputation decay curve describes how the reach and influence of a reputation event fade over time as coverage ages and search interest declines. Crisis communications practice was built on that assumption, which generative AI has substantially weakened by surfacing older material without weighting for recency.
Why does AI make old reputation damage persist?
AI assistants answer from indexed material and do not consistently weight it for recency or check whether a story was later corrected. An uncorrected article from years ago can therefore appear in a current answer presented as authoritative, keeping the event in circulation indefinitely.
Can old negative coverage be corrected?
It can be diluted and contextualised rather than removed. Publishing authoritative, well-evidenced material into outlets that carry authority changes what the sources say about an organisation over time. Contrlr measures whether that work moves the reputation score, which is how the correction is verified.
Does this affect private companies as well as listed ones?
Yes. Any organisation that customers, employees, or counterparties research is affected, because the retrieval behaviour is the same regardless of listing status. Listed companies carry the additional exposure of a market pillar that prices the narrative.
How do we find out what is currently being said about us?
Start with a baseline measurement across all seven pillars. Contrlr provides the reputation score, continuous monitoring, and AI content writing free at launch, so an organisation can see its current position and the sources behind it before committing to anything. Request a score here.