By Kurian Jacob
Source: Stock photo from Unsplash.
Muck Rack's 2026 State of AI in PR survey found that 76% of the PR pros it spoke to were using generative AI. Editing and refinement were among the most common applications. Given that it's a profession that's so very word-heavy, that makes perfect sense. It's useful for going through a tricky paragraph, cutting down on repetition or explaining something more clearly.
At the same time, AI's role in writing is coming under increasing scrutiny. Agencies will be expected to be able to explain how they use the tools and what work they actually do themselves.
Watermarking and journalists’ scrutiny
Anthropic has just introduced text watermarking for its supported Claude models, following the EU AI Act’s transparency obligations. It's an invisible pattern that can be used to indicate whether or not Claude was the one that produced the text. However, Anthropic points out that they can't tell the difference between text that Claude wrote and text that Claude has heavily edited, and detection is less effective on short samples.
From the conversations we've had with journalists, we know some of them do use AI detection tools to check out submissions. The tools can give very different assessments of the same piece of text, so their output shouldn't be taken as a definitive indicator of whether or not a piece is AI-generated.
In our work, we should be comfortable with explaining the use of AI and the reasoning behind the piece that we end up with.
How to incorporate AI into content development
Where we shouldn’t use AI:
Getting the AI to supply the view before we have the spokesperson’s input is just short-circuiting the process. It can generate a very good argument, but we're effectively assigning a view to someone who hasn't actually expressed it.
We shouldn't be outsourcing our thinking to AI or using it to plug the gaps that would have required a conversation with the client.
Before we write content, we need to gather the spokesperson’s views through briefing questions and conversations. What do they think people misunderstand? Where do they differ from the conventional wisdom? What advice would they give? Which experiences have shaped their opinion?
That's part of our job. We listen, we identify what's interesting, and we find the story from the client's point of view. Our own research and judgment as to what's worth pursuing and how to make it relevant to the reader is what drives it.
Where AI can help:
Once we have the ideas, AI can be used to help develop the approved material, suggest a better structure or improve the wording.
It can help us cut repetition or explain something more clearly. We'll still need to go through it to ensure that the changes have not altered the meaning or introduced any unsupported assertions.
We should also bear in mind that we're looking at the finished piece as something that a person will have to stand behind. Would the spokesperson have used those words? Does the piece reflect what they actually said? Could they have explained the argument if a journalist had then gone on to ask a further question?
Used well, AI can free up time for things like developing pitch angles, coming up with better questions, and thinking about which journalists would find a story of interest. Useful benefits for a busy PR team.
Being open about how we work
Transparency has to start internally and extend to the client. The team needs to understand which tools and uses are permitted and how AI has contributed to a draft. Reviewers should already have a solid understanding of the client’s messaging, but they also need to read the piece as the intended audience would. Claims and data should be properly referenced, and the argument should make sense without additional background material that the reader won't see.
Clients should also understand how AI may assist with their content and who reviews the output. That doesn't mean announcing every sentence we've shortened using AI. It means having a mutual understanding of our working practices and discussing any particular requirements an assignment or publication may have.
AI can help us put an idea across more clearly. However, that idea itself must come from the people we're talking to, from their experience, their views and the conversations we have with them.