Robo-Journalists — Ready to Go Solo?

Long before ChatGPT was part of our collective vocabulary, respected news publications like The Washington Post were using robo-journalists to report on the Olympics and even elections. For years, AI reporters have been relegated to reporting on statistics-driven topics but, as GenAI has become a part of the zeitgeist, it’s been given bigger tasks in the newsroom.

Early attempts to harness GenAI models to write news articles left much to be desired. Two years later, AI models have evolved into valuable assistants for journalists in newsrooms around the globe, but are they ready to go further?

The short answer seems to be, “No, they are not.” Many organizations have experimented with using GenAI to create original news stories from scratch, but even the most successful experiments were limited in terms of the depth and originality of content. So, what’s the best use for the increasingly important technology? Read on to find out.

Latest exploits in robo-journalism

Since GenAI burst onto the scene in a big way, more newsrooms have been tempted by the technology’s ability to create apparently convincing news content. Earlier this year, TechCrunch reported that under the byline “Quartz Intelligence Newsroom,” Quartz has been publishing AI-generated stories. Unsurprisingly, the quality of those stories has attracted valid criticism. From sourcing that only appears at the top of articles — which are cribbed from other outlets — to badly written headlines, the stories suffer from predictable issues even though Quartz claims each one is reviewed by a human before publishing.

Apple trips up

It’s not only media companies trying out AI for news. Tech giant Apple recently launched 'Apple Intelligence' - a news feature which used AI to create news content, but things quickly went awry when it began making up stories. The service was suspended following multiple complaints. It misinterpreted an article about a Brazilian tennis player so badly that it “... invented a story claiming tennis star Rafael Nadal had come out as gay…,” according to DailyAI, and did it using the BBC’s logo. It blew right past concerns about proper sourcing and poorly written content, and went straight into fake news territory.

Poor performance

The BBC itself recently conducted an experiment to assess the accuracy of a number of GenAI models. They tested OpenAI's ChatGPT, Microsoft's Copilot, Google's Gemini, and Perplexity AI by feeding them news content from their website and then asking questions based on those stories.

 The BBC reports, “... 51% of all AI answers to questions about the news were judged to have significant issues of some form. Additionally, 19% of AI answers which cited BBC content introduced factual errors, such as incorrect factual statements, numbers and dates."

The AI-generated newspaper

Issues like these, however, did not stop Italian daily Il Foglio from creating an entirely AI-generated supplement that was sold together with the main newspaper for an experimental period of a month. “The articles were structured, straightforward and clear, with no obvious grammatical errors,” reports The Guardian. It adds, “However, none of the articles published in the news pages directly quote any human beings.”

News commentator Claudia Giulia wasn’t impressed with the work. In her weekly "Tech and Privacy" Substack newsletter she wrote: “Today, analyzing Il Foglio AI after two weeks, initial observations show clear and logical articles, with some contextual gaps, missing details, and sporadic errors (names, dates). The style is often brilliant, but one feels the lack of human depth and verification.” She concludes "...these limitations, highlighted by the journalistic experiment, raise other questions: can an algorithm grasp the complexity of reality, irony, judgment?"

As well as the usual factual errors, these critiques also point to the lack of a 'voice', suggesting that, even if the tendency to 'hallucinate' is ironed out, GenAI authoring may be limited to more impersonal journalism where simple transmission of information is the aim.

The place of GenAI in the newsroom

Experiences like these indicate that, as originators of news content, GenAI models are not going to be replacing human journalists any time soon. Apart from the lack of a 'voice' the intrinsic unreliability of generative models (which developers have been making little headway on) means that their output must be thoroughly checked by a human author.

Where AI models can add real value is by taking care of routine time-consuming tasks, freeing authors and editors to concentrate on creating engaging content. AI assistants can 'package' a story in a few seconds to create headlines, summaries, subheadings, SEO elements and social posts.

Maximizing newsroom productivity

The arrival of the ProActions framework now allows users to create their own AI tools from multiple models, optimizing their integration into the workspace and maximizing productivity. One recent development generates the entire suite of story elements detailed above with a single click, significantly accelerating the creation and publication workflow.

Find out more about the ProActions AI framework.

For print publishers, the layout of print pages is still a major cost item. Eidosmedia customers in Germany are now using an integrated AI engine to cut the layout for a print edition from several hours to a few minutes.

Find out more about AI print pagination.