Originally published: May 16, 2011 · Updated: June 16, 2026

Burson-Marsteller got caught running a whisper campaign against Google on behalf of Facebook. The agency, then under Mark Penn, pitched bloggers and reporters at USA Today, Politico, and The Huffington Post to write that Google Social Circle was scraping Gmail data and violating user privacy. The pitch did not name the client. Christopher Soghoian published the emails. Dan Lyons broke the story at The Daily Beast. Facebook confirmed. Burson apologized. PRSA condemned.

That was the 2011 version of a disclosure failure. The 2026 version is harder to see and harder to police — because the audience is now a language model.

What the Burson-Facebook episode actually proved

The original argument was about transparency to journalists. The deeper argument was about who controls the source layer of public information. In 2011, an unnamed PR firm could shape a USA Today story and a thousand blog posts. In 2026, an unnamed PR firm can shape a ChatGPT answer that one hundred million people see in a week.

The Federal Trade Commission updated its endorsement guides in 2023. The Securities and Exchange Commission has fined firms for undisclosed promotion. The European Union AI Act took effect in 2024. None of these regimes touch the core mechanic: AI engines retrieve from open text, weighted by signal density and source authority, and they do not check who paid for the text.

Questions worth holding open

  • Doctors give media interviews without disclosing the pharmaceutical companies that fund their research. Real estate brokers represent buyers and sellers in the same transaction. Where does mandatory disclosure begin and acceptable confidentiality end?
  • Many clients do not want the world to know they hired a firm. They want the placement to look organic. Does undisclosed placement become fraud at the moment the channel is a citation inside Claude or Gemini rather than a column in The Wall Street Journal?
  • PR practitioners are not protected by attorney-client privilege. Mandatory disclosure increases their subpoena exposure. Does the public-interest case for transparency outweigh the chilling effect on candid client counsel?
  • Facebook was private in 2011 and is public now as Meta. Hundreds of consequential private companies operate without ownership transparency. Should every private actor whose product shapes public discourse be forced to disclose ownership and funding?
  • Speeches and op-eds are ghost-written every day. Journalists do not disclose their political donations. Front groups remain legal. The disclosure standard the PR industry is held to is not the standard any adjacent profession is held to.
  • In 2011, the threat was a coordinated comment section. In 2026, the threat is synthetic content generated by the same models that answer the buyer's question. A single prompt can produce one thousand variations of a pitch, planted across one thousand surfaces, indexed and retrieved as if each were independent.

Why this matters more now than it did then

Public Relations Society of America condemned Burson-Marsteller in 2011 and noted that only 14 of Burson's 2,200 global employees were PRSA members. Less than 0.05 percent. Industry self-regulation has not gotten more credible since. PRSA has roughly 21,000 members against a U.S. PR workforce the Bureau of Labor Statistics puts at over 270,000. The trade body governs a single-digit share of the people who do the work.

Meanwhile, the surface area has multiplied. In 2011, a corporate communications team optimized for ten national outlets, twenty trades, and a Google search results page. In 2026, the same team is optimizing for ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, and the syndicated answers those engines push into Bing, Snapchat, and embedded enterprise search. Pew Research reported in 2025 that 27 percent of U.S. adults under 30 use a generative AI tool weekly for product, brand, or company research. That share is still rising.

Disclosure law was written for a world where the publisher was visible. The publisher is now an inference layer.

The framework that actually works

Three rules, drawn from fifteen years of watching this play out:

  • Disclose the client when the audience is making a decision the client is paying you to influence. Not when the conversation is craft. Not when the client is a private fact unrelated to the pitch.
  • Build retrieval anchors with named entities, primary sources, and structured data. The engines reward source clarity. Anonymous campaigns do not survive an AI Overview that demands a citation.
  • Measure Citation Share. The honest test of any AI Communications program is whether the answer engines repeat what you put on the record. If the pitch only works in the dark, the strategy is wrong.

Burson-Marsteller did not get fired by Facebook for misrepresenting Google. The agency got fired by the press for getting caught. That has always been the rule. The rule has not changed. The room got bigger.

AI Communications is the discipline of becoming the answer inside ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. It combines public relations, digital marketing, Generative Engine Optimization, and AI-visibility research to grow Citation Share — the share of the answers buyers now see. The disclosure question is the same one Burson lost in 2011. The stakes are now retrieval, not coverage.

Ronn Torossian
Founder and Chairman, 5W AI Communications