The biggest AI ethics risk in public relations is not what practitioners know but what they do under pressure. Doug Downs, host of the Stories and Strategies podcast, argues the fix is rehearsal, not another training module. His case draws on research by Baylor University professor Marlene Neill for PRSA and the Arthur W. Page Center.
What did the WHO surgical checklist reveal about the gap?
The World Health Organization checklist showed that doing what you already know beats knowing it. In 2008, the WHO asked surgeons in eight hospitals around the world to do something they found insulting. Read down a 19-item checklist before operating. Confirm the patient's identity. Give the antibiotics. Count the sponges. The best hands in medicine, reading a list like student drivers. Surgeons hated it.
Then the results came in. Complications fell by more than a third. Deaths fell by nearly half. Nothing about what those surgeons knew had changed. Only whether they did it.
Public relations has the same gap, and AI just made it wider. We know the ethics. We have the codes, the competencies, the annual training module. What we don't have is the habit of practicing them when the pressure says skip them. That's the whole problem, and no tool is going to solve it for us. I wish one would. It would be easier.
I talked about this with Marlene Neill at PRSA ICON in Orlando. Marlene is a professor and graduate program director at Baylor University, a Fellow of PRSA, and she has spent years researching ethics in our profession. Her latest work, done for PRSA's Board of Ethics and Professional Standards, maps 10 essential AI ethics competencies, from critical thinking and ethical judgment to privacy, accountability, and governance. The finding that stings most: critical thinking is both the most essential competency and one of the profession's biggest gaps. We're weakest at the very thing our job is.
Why does an AI draft that sounds sure still need checking?
An AI draft needs checking because the machine sounds sure even when it is wrong. Ask a room of communicators whether AI can do ethics and most will give you the right answer. It can't. It can codify ethics, but it can never determine them. Empathy and ethical judgment are where the practitioner earns their keep. Ask the same room what they actually do at 4 p.m. on a Friday, with a client waiting and a draft the machine produced in nine seconds, and you get a different answer.
The pressure for speed is real, and the training and the guardrails have not kept up. Marlene calls it a perfect storm, and she's right. And the storm hits hardest where the guardrails are thinnest: the solo practitioner, the small team, the person with no one to check their work.
The outputs sound quite convincing. That's the danger. Marlene told me about a practitioner who trained a Claude agent to rewrite bios. It did good work, she started to trust it, and then on a simple rewrite it added a position she never held. She caught it because she checked. Marlene tried the tools herself and got an infographic back with a culturally offensive image and typos in it. These aren't edge cases. They're what the tools do when nobody is watching. As Marlene puts it, ultimately we're responsible for the output we share with the public. If we don't check it well, that mistake is on us.
Senior practitioners are already hearing it from their own CEOs. You advise going left for ethical reasons, and the CEO says, well, that's not what ChatGPT says. Your counsel has to be grounded in facts and case studies to carry credibility now, because the machine has an answer for everything and it delivers that answer with total confidence. Facts and case studies are the only thing that beats a confident machine. That's what critical thinking actually looks like in practice, and it's the competency we're worst at. Marlene's research says so plainly. The competency we need most is the one we have least.
What is left for the practitioner when AI writes the draft?
What is left for the practitioner is the judgment, not the draft. If AI can draft the release, write the statement, and build the campaign plan in seconds, the question is fair. Marlene has been researching this for the Page Society, and the answer from chief communications officers is blunt. They won't pay an agency to do something a junior staffer could do, or an AI tool could do. They're looking to cut their agency budgets, in some cases maybe in half, and spend instead on the things the tools can't do: strategic insight, thought leadership, crisis counsel. The higher level thinking. That's what they're buying now.
Agencies hate this. They like rolling out the senior people in the first meeting, before the contract is signed, and then you get the mid-level and junior staff the rest of the way. I've seen it for years. You don't see the senior person again until the Christmas party. Clients have figured it out, and now they have a machine that does the junior work for free. The big firms will be fine. The small specialists will be fine. It's the middle I worry about, the firms selling hours instead of judgment.
There's an uncomfortable truth inside that shift. Your value was never the draft. It was the judgment about whether the draft should go out at all, and what happens if it does. When a company uses AI to create diversity in a campaign instead of hiring diverse models, that's a crisis waiting to be revealed, and it's been revealed. When AI writes the apology statement after a crisis, people can tell it didn't sound authentic or sincere. The machine can produce the words. It can't take responsibility for them. Only you can, and that's the whole job.
Does ethics training change behavior, or does rehearsal?
Rehearsal changes behavior, and annual ethics training does not. So how do you close the gap? Not with another training module. Marlene's research for the Arthur W. Page Center looked at the annual ethics training most companies run, the online module everyone completes once a year, check the box, done. It doesn't change behavior. It was never designed to. And the cost of the gap isn't theoretical. In a national sample of US agency employees, about a third said they were facing retaliation and staying silent when they saw ethical issues in their organizations. A third. These are people who know what's right and say nothing anyway.
The fix Marlene points to is a program called Giving Voice to Values, developed by Mary Gentile. It doesn't start with deciding what's right. It starts with rehearsing what you will actually say when the moment comes. What will you do, not what should you do. The rehearsal covers four questions:
- What will you actually do when the moment comes?
- What rationalizations will the people around you use to justify the unethical choice, and what will you say back?
- Who is your ally, the person you bring with you so you aren't standing alone?
- How do you tie your concern to the organization's core values, the ones it actually lives, not the ones it frames on the wall?
Because if those values are aspirational and the organization isn't living them, you've no credibility with your employees. That's Marlene's line, and it's the sharpest one in the whole conversation.
This is the part of the work nobody can automate. You can't rehearse courage on a Friday afternoon with a chatbot. You do it before you need it, the way a surgeon runs the checklist before the first incision, so that when the pressure hits, the doing is already decided. Decide what's right and you've done the easy half. Plan what you will say, expect the pushback, bring an ally, and tie it to values the organization actually lives, and you've done the half that matters.
The knowing-doing gap is about courage, and better AI tools will not close it. The counterargument is fair, and I want to answer it straight. The tools keep getting better. Verification gets easier. The models get guardrails built in, and maybe one day the machine catches its own mistakes before you do. Fine. None of that closes this gap, because the gap was never about the tool. It was about what a person does under pressure, with a deadline, a client, and a boss who has already decided. A better machine doesn't make a braver practitioner. It makes a faster one. Faster isn't the same as braver.
The other objection is harder, and it deserves honesty. Speaking up costs you. It can cost you the client, the account, the job. Marlene's research says a third of agency employees already know that price firsthand. But silence has a price too, and it's higher. You're the voice and the face of the brand. Your credibility's on the line with every statement that goes out under your name. The machine won't stand up in the meeting and take the blame. You will. And when the apology the machine wrote sounds hollow, when the campaign the machine built turns into a crisis, nobody asks the model what it was thinking. They ask you.
The gap isn't knowledge. It's courage. The checklist is 19 items long and any surgeon could have recited it from memory. What changed everything was doing it, every time, out loud, in front of the room. Our checklist is shorter. Verify everything the machine gives you. Rehearse what you will say before you need to say it. And when the moment comes, say it. Don't be the silence.
We cover stories like this every Tuesday, on the Stories and Strategies podcast.