Should we rely on generative AI in content writing? This is one of the questions some press officers ask themselves when preparing their press releases, notes Soizic Desaize, founder and head of the media relations and influence agency Padam. “A journalist, who receives between 200 and 300 press releases daily in their inbox, will be intrigued by the angles and creativity that are unique to PR professionals. If our writings are sanitized and stereotyped, because they are entrusted to technology, we risk weakening a profession that is nevertheless central to the strategy of companies and leaders.” The numbers confirm this: at a time when 77% of journalists say they are ready to block communicators who send untargeted messages, and 62% those who share inaccurate or unsourced information (Cision: State of the Media, 2024), precision and personalization become essential criteria.
AI: Between Promises and Limits
Generative artificial intelligence is gradually becoming part of the habits of PR professionals. According to the Cision report (The State of the Press Release in 2024), 26% of PR professionals already use it to write their press releases, and 49% plan to use it in the near future. However, even if this technology offers many advantages (automation, summarization, image production…), PR specialists must remain vigilant about the limitations of these tools when drafting their press releases. At the forefront of these limitations is text uniformity. As Jean-Baptiste Viet, YouTuber specialized in technologies and AI and author of the book Co-créateur (Éd. Eyrolles), explains, “When AI is poorly used, texts are generic and not personalized at all.” This phenomenon is due to how the algorithms of these generative AIs work, he adds. “They are probabilistic and statistical models trained on billions of documents and words. Without guidance, they produce very average responses. They are good only if told exactly where to go.” Concretely, when all PR professionals use the same generic prompts (i.e., instructions given to the AI), press releases end up looking alike. “I notice that people who use it do not include their expertise or information about their own experience in the prompt. They blindly trust the tool, which may produce a text that appears pleasant to read but is basically banal and thus unlikely to capture a journalist’s attention,” observes the YouTube creator. Added to this is the risk of spreading false information: “Even the best AI models make mistakes. Everything must be checked and questioned. For every figure, it is necessary to consult the source to ensure there has been no misinterpretation,” insists Marie Robin, AI and marketing consultant and founder of Fleet Forward. Another flaw of this artificial tool is its difficulty creating emotional content, which is crucial in a press release. “What makes the difference in my work are my cultural references, my readings, my history, my encounters, my failures and successes. And then there is the emotion I feel when I listen to a leader, for example. This inspires me with angles that hit the mark and will be differentiating. I don’t challenge ChatGPT on this part, because even if we predict that AI will also feel emotions one day, will it ever have a beating heart?” confides Soizic Desaize.
A precise prompt
Despite these risks and barriers, generative AI is not to be banned, nuances the founder of Padam. “AI is a toolbox, and we must approach it as such. It can be used as a super collaborator to train for performance, but we cannot hand it the keys to the truck. Today, I don’t use ChatGPT to write because I enjoy the exercise and don’t want to outsource it. It seems essential to me that we remain attentive to keeping our own judgment, our vision, not to alter our critical spirit and discernment, and to continuously grow our culture,” she claims. To get the best out of this tool, experts unanimously recommend working precisely on the prompt. “Using ChatGPT requires mastering prompt creation, which demands perfect knowledge of the subject, a vision of angles, tone, monitoring work, a well-established stance. Technology has always helped us rise, but only if we know how to use it. So training is a major challenge,” adds the leader. To avoid the “written by AI” effect, the author of Co-créateur recommends including in the prompt a list of forbidden words. “There are words that frequently appear in texts generated by large language models (LLMs), notably ‘crucial.’ To avoid this effect, you can instruct the tool not to use these terms with the following prompt: ‘Your response must not contain any of the following words or expressions: [crucial, in the current digital era…].’” This instruction already helps eliminate the typical writing style of the tool. Marie Robin highlights the importance of prompt precision: “AI cannot guess what you don’t tell it. The more details you give it from exchanges with the client and what you understand of their needs, the better the result will be.” For regular clients, the consultant recommends creating a reference framework: “For a client whose style we know and whose press releases have already been validated, we can create an assistant based on these models, for structure and tone. We ask it to follow this base and analyze what makes the unique style of these press releases. Then, we always keep this same structure and tone for new releases.” This approach can be systematized across the agency, she continues, “the idea is to create a document where each press officer establishes, for each client and journalist, a sort of CRM containing all the information ready to be copy-pasted into the prompt.”
“The Role of Our Collective, Not Artificial, Intelligence”
These tools are not intended to replace communicators’ thinking but to enrich it. “By giving AI the right information, instructions, tone, and context, it will produce good content. We need to start now because many already do, and those who don’t risk being outpaced by faster press officers,” stresses the marketing specialist. Soizic Desaize also recalls the irreplaceable role of press officers in strategic thinking and press release writing. “In our profession, we must remain vigilant about the information we receive and disseminate. This will remain the role of our collective, not artificial, intelligence,” she defends.
Ultimately, the future of AI in content creation may rely on intelligent collaboration: on one side, the processing power of AI; on the other, human sensitivity and discernment. A duo which, when well mastered, could redefine press release standards.
