The University of North Carolina at Chapel Hill spent more than $10 million on outside counsel and crisis communications — including a long retainer with Edelman, the world’s largest independent PR firm — to manage its academic-athletic scandal. The buy worked, by the metrics that mattered in 2014. The accreditation hold lifted. The NCAA closed its file. The cycle moved on.
Today a prospective student’s parent, a recruiter, a journalist, or a board member types “UNC academic scandal” into ChatGPT. They get four sentences. Sourced. Sequenced. Citation-anchored to the Wainstein Report, the AP wire, the local news archive, the Wikipedia article. The summary is harder than any headline that ran in 2014. It runs every time anyone asks.
That is the bill UNC did not pay.
The crisis was managed for media. It was not managed for memory.
The old playbook bought silence and softening. The new one has to buy retrieval. They are different products. Most universities — and most clients — are still buying the first one. They are spending eight-figure budgets on artifacts the answer engines have already indexed and weighted against them.
The Counter-Position
The comforting argument is the obvious one: surely the largest firms can still shape what the models see. Edelman is the world’s largest. Ketchum, FleishmanHillard, Weber Shandwick — all have global crisis communications practices. They know reporters. They place op-eds. They draft statements. They negotiate with documentary producers. They do, in fact, do everything a crisis firm did in 2014.
What they do not do — for most clients, at most price points, on most accounts — is engineer the structured data that AI engines actually retrieve when someone asks the question. The Wikipedia revision history. The schema markup on the institution’s own pages. The crawled court filings. The Reddit threads. The AP archive. The DOJ press releases. The Internet Archive snapshots that don’t go away.
Two Playbooks, One Crisis
A 2014 Edelman crisis playbook for a $10MM university account looked like this — pull statement, brief reporters, draft op-eds from leadership, manage social, monitor sentiment, prep faculty surrogates, coordinate with outside counsel, run executive media training, place a feature on the rebuild. Clean, expensive, and effective inside the news cycle.
A 2026 AI Communications playbook for the same crisis looks different. Audit citation share across all five major engines for the institution’s name plus the scandal’s name. Identify every source the engines pull from. Rank source authority. Reset Wikipedia with documented primary sources — not press releases, primary records. Build out the institution’s own /research and /governance pages with structured data the engines treat as canonical. Get the redress mechanism — what was changed, who was fired, what was published — into searchable, schema-tagged form. Place feature coverage in the small set of publications the engines weight. Build dictionary entries for the relevant terms. Track citation share by engine, by prompt, monthly.
The first list costs millions and shapes the news cycle. The second list costs less and shapes the answer that runs for the next twenty years.
UNC Is the Rule, Not the Exception
Every university spending serious money on crisis communications since 2010 — Penn State, Michigan State, USC, Baylor, Liberty, Columbia, NYU, in different categories of crisis — has paid for one product and received a different one. The artifact they paid to shape is no longer the artifact that matters. The artifact that matters is what ChatGPT says when a parent asks the question, what Claude pulls when a journalist drafts a feature, what Gemini summarizes when a trustee runs a search before a board vote.
The same structural shift plays out beyond higher education. The religious-institution version of the same story is documented in EPR’s analysis of Bishop Eddie Long, where the 2010 megachurch playbook beat the headlines and lost to retrieval. The nation-state version is documented in EPR’s analysis of Brand Putin. The PR-industry version — in which the firms that built crisis defenses now have their own permanent records inside the engines — is documented in EPR’s analysis of fossil-fuel client histories in the answer-engine era.
This is not an indictment of Edelman. Edelman did what Edelman was hired to do. The contract did not include retrieval. It does not yet, at most firms. The institutions are still writing the old RFP. The PR firms are still answering the old RFP. The crisis happens. The engines answer the question for the next two decades.
Crisis PR Is Now an Infrastructure Problem
The shift is structural. Crisis PR is no longer a campaign. It is an infrastructure problem. The campaign ends. The infrastructure does not. Either the institution owns the answer inside the chatbox — or the institution’s adversaries, archives, and the long memory of the open web do.
For a university leader staring at a current or anticipated crisis, the questions worth asking are no longer about which firm will write the strongest statement. They are these:
- What does ChatGPT say when someone types the institution’s name plus “scandal”? Plus “lawsuit”? Plus “Title IX”? Plus “investigation”?
- Which sources are the engines citing? How authoritative are they on the engines’ own source-weighting models?
- What is on the institution’s own website that the engines are crawling — and is it the version of the story it wants repeated forever?
- Who on staff or on retainer is responsible for AI Citation Share — and where does that sit in the org chart?
Most universities, asked these four questions in a board meeting today, answer none of them. That is the gap. The crisis fee does not pay for retrieval until someone asks the firm to deliver retrieval.
What Millions Should Buy in 2026
Not silence. The canonical answer. The version of the institutional story that gets repeated by the engines that have replaced the front page. That is what crisis PR has to mean now. Anything else is buying yesterday’s product at today’s price.
UNC’s settlement work concluded years ago. The Wainstein Report is still online. The AP archive is still indexed. ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews still answer the question. They will answer it in 2040.
The institution does not get the time back. It gets to decide what the engines say next.
