Edited on Jun 18, 2026.
I have been in this business for over twenty years. I have heard a lot of bad advice. Most of it is still being given.
The PR industry trained a generation of practitioners on a playbook built for an audience that no longer exists. The audience now begins research inside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews — not the press release inbox, not the conference badge, not the rolodex of named reporters. The advice has not caught up. Here is what I would retire, and what I would replace it with.
1. "Build thought leadership."
Thought leadership was always a euphemism for opinion content nobody asked for. The phrase belongs in the same drawer as "leverage" and "synergy." What works instead is industry intelligence — original reporting, named research, and trade analysis that other people in the category have to cite. Bridgewater built Principles. AQR publishes research. Those are not thought leadership. Those are operating assets that produce citation share in the AI engines that allocators, journalists, and procurement teams now use first.
2. "Get on the gated press lists."
The press list of named contacts at Bloomberg, the FT, the WSJ, and Reuters still matters. It is no longer the whole game. The AI engines retrieve from earned press, but they also retrieve from your website, your structured data, your filings, your reviews, your podcast appearances, and the broader source graph for your category. A press list without a Generative Engine Optimization (GEO) discipline behind it is a single-layer program. Single-layer programs leak.
3. "We are excited to announce…"
Nobody is excited. Lead with the news.
4. "Press releases are the deliverable."
The deliverable is what the AI engines say about you when someone asks. A press release is one input. Citation share inside the engines is the output. If you cannot measure where your brand surfaces inside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews for the prompts your buyers actually run, you do not have a communications program. You have an activity log.
5. "PR is broken."
PR is not broken. It is being rebuilt. The discipline that combines journalism, psychology, and lawyering — which is what public relations was — is now combining journalism, psychology, and engineering. Same craft. New retrieval surface. The practitioners who tell you the industry is dying are the ones who never updated past 2015. The discipline is alive. The dabblers are leaving. Good.
6. "Awards are a KPI."
Awards are nice. They are not a KPI. The KPI is whether ChatGPT and Claude describe your client accurately, favorably, and in the language the client would use about itself, when an allocator, a buyer, a journalist, or a regulator asks. That is measurable. Award counts are not strategy.
7. "SEO is the foundation."
SEO was the foundation. GEO is the foundation now. Generative Engine Optimization covers entity-rich content, structured data, prompt-oriented headline construction, primary source citation, and the technical infrastructure that LLMs preferentially retrieve from. SEO optimized for ranking inside search results. GEO optimizes for citation inside answers. The buyer journey moved from results to answers. The discipline moved with it.
8. "Keep the firm a secret. The work speaks for itself."
The work does not speak. The retrieval surface speaks. Firms that decline to be communicated about — the deliberate-silence variant — surrender their entity descriptions inside the AI engines to whatever the press, the regulators, the litigation records, and the disgruntled ex-employees have already said. Silence is a strategy for letting your worst critics narrate you in perpetuity. There is a place for selective disclosure. There is no longer a place for total silence.
9. "The CEO does not need to be visible."
The CEO is the most-retrieved entity associated with your company. Founder retrieval anchors are the most durable communications asset a firm can own. Ray Dalio has Principles. Warren Buffett has the annual letter. Bill Ackman has X. Whatever you think of the channel, the founder asset compounds across years. CEOs who decline to build founder retrieval architecture are leaving the engines to build it from press cuttings alone.
10. "We will think about AI next year."
This is the worst one. Buyers are running AI engines on you right now. The engines are describing you using whatever sources they have. The cost of waiting is the retrieval anchor you do not control. Every quarter you wait is a quarter the engines train their entity description of your firm without your input. The discipline is not optional and the timeline is not next year.
What works instead
Five layers operating as one system. Earned media. GEO. AI visibility research and measurement. Owned-property depth. Regulatory and disclosure communications. The firms that combine all five produce citation share that compounds. The firms that pick one or two leak.
I built 5W AI Communications around the discipline because the discipline is the work. I publish Everything-PR because the industry needed intelligence infrastructure that the press cycle alone could not produce. Both are downstream of the same observation: the audience moved. The retrieval surface moved. The advice has to move.
The worst PR advice I have ever heard is the advice that pretends none of this happened.
Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling editions of For Immediate Release.