AI tools offer tantalizing bite-sized summaries of current affairs, but can we trust the information we are being fed?
This technology is becoming an everyday source of news for millions of people, who now use tools like ChatGPT and Google’s AI Overviews as shortcuts to accessing it. But according to a new report from the Institute for Public Policy Research (IPPR), the convenience of AI-generated news short-cuts come with risks.
As reported by the Guardian, the left-of-centre thinktank argues that AI-generated news should carry “nutrition labels” with clear disclosures explaining where information comes from, what sources were used, and how reliable they are. The idea is to make AI news more transparent about its contents, enabling people to make informed decisions about the news they consume.
The ‘gatekeepers’
AI tools typically work by providing summaries of news events rather than directing users to original journalism. Research from the Reuters Institute for the Study of Journalism suggests that around a quarter of people already use AI tools to get information, while Google’s AI Overviews alone reach roughly two billion users each month.
The impact can be significant for publishers. Many users read the AI summary but never actually click through to the original reporting, which cuts traffic and advertising revenue. Furthermore, users often have little sense of how answers are generated or who is behind the journalism that supports them.
AI companies, the thinktank argues, are becoming the new “gatekeepers” of the internet, warning that without intervention, this development could severely damage the future of news as we know it.
What would a ‘nutrition label’ include?
At the core of the IPPR’s proposal is transparency surrounding the AI answer, by giving users valuable context about its makeup. These disclosures would show the sources used to generate an AI answer, such as peer-reviewed research, professional journalism, or other online content.
The goal is to provide readers with enough context to judge the reliability and quality of the information they consume.
“If AI companies are going to profit from journalism and shape what the public sees, they must be required to pay fairly for the news they use and operate under clear rules that protect plurality, trust, and the long-term future of independent journalism,” said Roa Powell, senior research fellow at IPPR and co-author of the report.
Who really pays for AI journalism?
The IPPR is also calling for a formal licensing regime in the UK that would allow publishers to negotiate collectively with tech companies over the use of their content in AI-generated news.
Rather than weakening copyright law, the report says it should remain unchanged to allow a licensing market to grow. It also highlights the potential role of the Competition and Markets Authority (CMA), using its new enforcement powers. The CMA has proposed giving publishers the ability to block Google from scraping their content for AI Overviews.
Collective licensing could be beneficial for avoiding situations where only the largest publishers benefit, and smaller outlets are left out.
Putting AI to the test
To understand how AI tools already rely on journalism, the IPPR tested four major platforms using 100 news-related queries and analysed more than 2,500 links in AI-generated answers.
ChatGPT and Google Gemini did not cite any BBC journalism at all, as the BBC has blocked the bots used to generate AI responses. Yet Google’s own AI Overviews and Perplexity continued to use BBC content despite the broadcaster’s objections. Meanwhile, outlets such as the Telegraph, GB News, the Sun, and the Daily Mail appeared in fewer than 4% of ChatGPT responses.
Meanwhile, the Guardian, which has a licensing deal with OpenAI, featured in nearly six out of ten answers and about half of Gemini’s responses. The Financial Times, which also has a licensing deal, appeared frequently. This calls to light how financial relationships shape what an AI system outputs as authoritative information.
Risks and moving forward
If licensed publications appear more prominently in AI news output, smaller and local news organisations that are less likely to get AI deals may struggle for visibility.
The IPPR warns this could further weaken an already fragile news ecosystem. Additionally, over-reliance on tech giants could make news organisations vulnerable if deals are withdrawn or copyright rules change.
To help combat this issue, the thinktank is calling for public funding to support new business models for investigative and local journalism, and for the BBC to be supported in “innovating with AI.”
An OpenAI spokesperson told the Guardian that the company is “committed to supporting quality journalism,” adding that ChatGPT draws from a broad range of publicly available sources and provides citations and links to help users verify information and drive traffic to publishers.
Whether the AI nutrition label idea ever comes to fruition remains to be seen. But as AI increasingly mediates how people understand the world, the world deserves to know what it’s consuming.
Also read: A class-action case argues AI training data practices could force major vendors to rethink how they source copyrighted content.