On a fairly ordinary Monday, our PR platform, Agentcy surfaced a funding news story on a national news outlet involving several prominent industry backers. The news had implications for several of our clients in the AI infrastructure sector, so we started discussing what we could do with it internally.
The problem was that the article appeared to have been taken down by the time we clicked through. It’s left me curious enough to spend longer trying to find the story than I would have spent reading it. Did the deal change? Did someone hit publish ahead of schedule, or, worse, had I imagined the whole thing before my morning coffee?
ChatGPT reassured me that I hadn’t imagined it, citing the national news outlet and linking to the same unavailable article. The link remained in Google’s search results over the next days.
Two days later, the actual news went live with the usual fanfare of a major announcement. Between Agentcy surfacing the story and ChatGPT citing the missing page, we had an extra 48 hours to sit with the news before it reappeared.
The cat has acquired a distribution network
Anyone who has regretted an old social post knows the internet’s memory. Between screenshots, forwarded messages and copied articles, deleting something has always involved a degree of uncertainty about where it might turn up next.
LLMs add another route through which published information can reach readers. AI crawlers searching the web can use credible third-party sources to produce an answer, letting someone encounter a story without ever visiting the original page.
In PR, we regularly tell our clients about the role earned media play in helping companies appear in AI answers. This was a direct example of that, but it’s also a double-edged sword.
In this case, a story published ahead of schedule has put news into AI answers before a company is ready to have the conversation, potentially reaching employees, investors and competitors. The cat got out of the bag and has acquired a distribution network.
How a claim can travel with its references attached
Death and taxes are fairly predictable. What happens after sensitive company news appears ahead of schedule is harder to know, especially when you don’t know who’s seen it or what they might do with it.
An employee expecting a direct briefing may ask their manager why they found out through a news alert, while an investor may call with questions before the business is ready to answer. Competitors could gain an early indication of the company’s plans and use that time to adjust their own.
The cleanup takes more than pulling the article, especially if AI tools repeat inaccuracies from it. People may continue to come across and act on those claims well after the original page has disappeared, leaving the business chasing explanations and rebuilding trust where the information has travelled.
A smaller margin for error
This may be the first instance we’ve seen ourselves, but it is unlikely to be the last. For communications teams, the margin for error narrows significantly when information can travel beyond an article before anyone notices a problem. Taking down a page does little to reassure employees or investors who expected to hear directly from leadership.
Agreeing on embargo times and checking publication settings have always been crucial for sensitive announcements, and AI makes those basics more important than ever. Teams now also need to account for AI answers in their crisis plans, because these tools can keep serving up claims after a publisher removes or changes the original story.
Technology will keep changing how information travels, giving communications teams more to account for in crisis planning. Equally, this makes the fundamentals of PR more important than ever, or risk putting your company’s reputation at risk before the team has time to respond.