Status report: artificial intelligence (AI) in public relations

The impact of AI on public relations is set to be transformative for public relations practice,­­­­ but it will demand updated practitioner capabilities. 

We have almost certainly underestimated the medium to long-term impact of AI on public relations. The pessimistic argument is that it’s a threat to all professional disciplines. The optimistic viewpoint sees it as a tool to help public relations practitioners work more efficiently and effectively.

The technology has developed slowly over the past ten years following innovation in data processing and natural language development, but then incredibly quickly in the past 12 months. The launch of ChatGPT by OpenAI put AI in the hands of anyone with a web browser. It was rapidly followed by Anthropic Claude 2 and Google Bard.

AI tinkering and experimentation

The approach to AI by public relations agencies and communication teams is characterised by tinkering and experimentation. Transparent data management and security policies are required before the technology can be deployed in commercial applications.

Please do not cut and paste or upload documents to any web-based tool without checking its data management policies and asking yourself whether you trust the vendor.

The AI industry recognises the need to develop enterprise-level solutions. Large organisations have sought a competitive advantage by licensing a large language model API and building in-house tools. An example is an in-house AI application developed by the marketing services group Publicis called Marcel and rolled out to agencies within the group.

Making sense of AI tools

There is an ongoing tension between large language models and third-party developers. Already, we have seen solutions built on top of the AI models such as chatbots, editing, text-generation and summarisation tools, be cannibalised by subsequent generations of AI large language models. The market needs to settle down before serious investments can be made in workflow.

Large language models aren’t the only way AI can be used in public relations. Another growing application sees machine learning applied to datasets to support human decision-making. This is an area of focus by measurement and monitoring vendors such as Cision, NewsWhip and Signal AI.

Watch out for a new version of #PRstack set to be published before the end of November by the #FuturePRoof community developed by my partner Sarah Waddington. We’ve worked with more than 20 practitioners and vendors to review each of the large language models and different generic and industry-specific tools.

The first draft of anything is shit

Attention has been focused on generative applications where tools built on large language models can create the first draft of anything based on a simple prompt. The results are typically bland, generic copy. It’s often a bit shit, but then it reflects the sum of the data, typically content from Google Books, Wikipedia and other websites, used to train models.

Whatever your current view of computer-generated copy, it will get better as models improve. I suggested to colleagues 12 months ago that I could spot synthetic copy. It had a recognisable pattern and syntax. No longer. Improved prompting has already led to tools producing undetectable facsimiles of notable writers.

Copyright, plagiarism and the threat to reputation

I was recently interviewed about AI by Pattern’s Arlen Pettitt and Christopher Owens. They produced two versions of the article: an original with words by Arlen and pictures by Chris, and a second using ChatGPT to write in the style of Christopher Hitchens, Marina Hyde and George Orwell. The results are equally surprising and terrifying. It’s an insight both in terms of the reputational impact to individuals and organisations, and the threat to creative and professional work.

Copyright and plagiarism are among the regulatory issues that need to be addressed. The Authors Guild, backed by John Grisham, Jonathan Franzen and Elin Hilderbrand, is suing OpenAI in the US for breach of copyright. It claims that their books have been used to train ChatGPT.

Entry-level and junior roles

The AI Generation published by Demos suggests that AI productivity gains may be higher for lower-skilled roles and threaten graduate jobs traditionally done by trainees. The functional public relations role of producing words and pictures such as press releases, articles, quotations, media pitches, and social media posts can already be automated.

This raises important questions about how we train the next generation of practitioners. Demos suggests that traditional linear career ladders could disappear. Graduates may face more fluid careers and need adaptability, self-motivation and entrepreneurial skills.

Andrew Bruce Smith, chair of the CIPR AI in PR panel, is optimistic. “AI is a rising tide that lifts all boats, demanding junior public relations practitioners tune their skills beyond the rote and routine to the analytical and creative,” he said.

Fact-checking remains an issue. A large language model will generate a convincing output even when it doesn’t have the answer. It’s important to remember that it is artificial and not intelligent.

Large language model vendors seek to address this issue by reducing the lag on training data and combining results with live web data. The OpenAI GPT-4 Turbo model is trained on data up to April 2023 and both OpenAI and Google Bard offer options to incorporate live web data in results.

Senior roles and management

AI's arguably more powerful, reductive application in public relations has largely been overlooked. Here, large language models are trained on spreadsheets, reports or research papers and instructed to create summaries and review data from different perspectives. This is powerful in planning, measurement and decision-making. The ability to spot patterns in data offers the potential to create new knowledge.

A report by the CIPR, Humans needed more than ever, suggested that public relations practitioners will be needed to support organisations with relationship-building, emotional intelligence and ethical decision-making outside the reach of algorithms.

“In hiring and nurturing talent, public relations leaders must look past resumes to recognise the inherent value in flexibility and innovation, qualities not just tested by AI but essential for harnessing its full potential,” said Smith.

This is the relationship management perspective of public relations. The outcome is trust and reputation. If practitioners are to achieve this level of expertise, they will need to improve both their domain and management expertise, crucially in terms that are recognisable to management.

Thanks to Antony Cousins, Andrew Bruce Smith and Michael White for comments and feedback on an initial draft of this article via a LinkedIn discussion. NewsWhip is a Wadds Inc. client.

Image by Christopher Owens.