AI Recycled a Dead Promotion and Sent Customers to a Client’s Doors

young man standing in a restaurant cash register queue

Every crisis communicator has handled misinformation: false rumors, twisted facts or one-sided stories. Now there's a new generator of fake news: LLMs. The very ones who became the new PR audience as AI search provides a main avenue for how consumers discover and engage with brands.

That is the scenario our team faced earlier this year, and it exposed a gap in crisis preparedness that most communications professionals have not yet closed.

The Setup

A quick-service restaurant client ran a well-received Teacher Appreciation and Nurses Week promotion in 2025 which included free food, strong coverage and tons of community goodwill. In 2026, the client chose not to repeat it. AI did not get the memo.

Somewhere, a journalist (or 50) asked an AI tool for freebies and discounts tied to Teacher Appreciation and Nurses Week. The tool served up a list, and outlets ran with it.

Not one person contacted the media team, checked the brand's website or made a single call before publishing. Teachers and nurses showed up at locations nationwide expecting a deal that had not existed in over a year. The brand absorbed the fallout.

The Channel Decision That Mattered Most

The instinct in any crisis is to respond quickly and publicly. Social media feels like the obvious lever. The PR team pushed back on that idea because posting mid-week to clarify that a brand is not giving away free food to teachers and nurses, while accurate, gives brutal optics of a brand distancing itself from two of the most respected professional communities in the country. There’s just no good way to say it.

Instead, the PR team deployed a new tactic, posting a clear, factual statement about the outdated promo on a non-indexed page on the brand's website where LLMs could find it and share the information in search responses. Local franchisees were briefed so frontline staff were ready with responses before customers arrived and the brand asked local operators to honor the offer (where feasible) to protect guest relationships. The PR team spent hours contacting newsrooms to ask for stories to be updated.

Disappointingly, media that ran the false information did not issue any corrections, but rather quietly removed the brand’s name from the stories. No editor's note, no notification to readers, no acknowledgment to the brand or to us. No accountability from the media that got it wrong.

What This Cost, and What It Revealed

The costs were real: unplanned giveaways at the store level, PR hours diverted from proactive work to reactive cleanup and a media cycle that kept recirculating false information, feeding it back into the AI systems that started the problem. The brand paid for all of it without causing any of it.

This is a new risk category for every crisis matrix: recycled truth. It is not fabrication; it is legitimate information AI has stripped of its timestamp, published by media without verification. Brands that weather it best will have already built these three protocols:

  • Post-campaign content audits. Know where promotional content lives after a campaign ends, what is indexed, cached or still surfacing in search.
  • Expiration metadata and structured data hygiene. Work with digital teams to build time-sensitivity signals into promotional content so AI systems have a fighting chance at context.
  • Pre-built media remediation frameworks. Know exactly who to call and what to send the moment unverified content goes live. Formal corrections are unlikely; silent removal is the realistic best case and even that takes sustained effort. Speed is infrastructure.

The Harder Conversation

The PR industry spends enormous energy educating clients on media relations, messaging and crisis preparedness. It needs equal energy on what happens when the media itself becomes an unintentional vector of harm to represented clients and brands.

The verification tools exist. The choice to use them is what's missing.

For communicators, the mandate is clear: build these protocols now, before the algorithm recycles something a client did two years ago and sends it viral during a holiday week. This was not an isolated event. It was a preview.

Kristen Kauffman is Senior Vice President, Editorial at SPM Communications.