Custom CSS

Breaking news: How AI technology is helping science journalists stay on top of the latest research

Photo of the author who wrote this blog post
Emma Warren-Jones
7 min read

Science journalists are a vital bridge between researchers and the public. They “shine a light on discoveries” [1] and translate complex studies into accessible articles that help to educate and inform the wider community. This isn’t always an easy task — science journalists face an uphill battle to secure an exclusive story about significant scientific breakthroughs and developments. Not only are journalists overwhelmed with the increasing volume of scientific papers, they’re also under pressure to transform important discoveries and findings into engaging articles as quickly as possible. The challenge is sifting through lots of often dense and complex papers to nab that unique story that will engage and grow their readership. Those papers could be wide-ranging in their coverage: from climate change to quantum mechanics, so a balance needs to be struck between being well-informed across a range of subjects without necessarily having formal training in these fields.

AI technology is helping science journalists

Four hurdles faced by science journalists

Delivering a news story that captures public imagination involves a lot more than simply re-writing a press release. Here are just a few of the most common problems science journalists face daily:

1. The volume of peer-reviewed research published each year is increasing

It’s estimated that around two million articles are published in scientific journals each year [2] — which translates into an overwhelming amount of research for science journalists to sift through to find that golden news story.This challenge has only grown since the start of the Covid-19 pandemic. Sabine Righetti, a science and innovation writer for the Brazilian daily Folha de. S.Paulo said that, towards the end of 2020, more than ten scientific papers in journals indexed in the Web of Science were being released every hour [3]. With so much source material to work with, journalists are continually faced with the difficulty of screening enough new literature, but quickly, to identify just the key findings and takeaways. Abstracts may be helpful here but relying on them may result in something significant being overlooked.

2. Preprints are also on the rise

A preprint is a full draft research paper that’s shared publicly before it’s had a chance to be peer reviewed. Apoorva Mandavilli, a science and global health reporter for The New York Times has said that the Covid-19 pandemic has “really accelerated the pace and visibility of preprints” [4]. There are some obvious advantages to researchers publishing preprints, not least because it means they get to release their findings to the world much faster. But they’re often approached cautiously by journalists who will need to read the whole paper to tease out any unanswered questions. It’s possible for some preprints to include work that doesn’t meet certain standards, such as “randomly selected patients or control groups enabling more definitive conclusions” [5]. It’s therefore more important than ever for reporters to analyse the full preprint in detail.

3. Press releases can’t always be trusted

Designed to pique a journalist’s interest, a press release will often draw out the key takeaways of a research paper. By their very nature, they can often be overhyped or sensationalised — it’s very often the case that journalists will want to read the full research paper to fact check the claims found within a press release. Research conducted by James Randerson for The Guardian showed that 100% of the science journalists he surveyed ‘always’ or ‘mostly’ choose to read the full scientific paper when writing their stories [6]. Expanding on this, Mark Henderson, former science editor at The Times, said: ‘press releases can be misleading — they may be hyped, they may miss the real story, or they may simply contain insufficient detail. You have to check that the press release is an accurate reflection of the science’[6].

4. Journalists don’t have a lot of time

We’ve seen that science journalists can’t always trust press releases to properly convey the paper’s key points, which means they’ll often favour reading a scientific paper in full.The problem is that journalists only have 24 hours in the day, just like the rest of us. When they’re inundated with papers, they must be selective about the ones they choose to read.In addition, research papers cover a wide range of specialist topics, and they may not have the specialist knowledge to decipher jargon or complex terminology. This can make reading a long and chunky scientific paper even more time-consuming.

How emerging technology can help science journalists

With so much impetus on finding an original story and reliably reporting scientific discoveries, and their implications, many news and information organisations have been exploring the use of AI technology -  not only to identify source material but to make this content easier for editorial departments to digest and translate for their readers.There are currently AI platforms helping journalists perform time-consuming tasks more easily, including fact-checking in real time, analysing data to determine the potential for a story, and generating automated news coverage.In 2018, Reuters went one step further by building a tool designed to help journalists analyse data and suggest story ideas. It sorts through data to spot patterns and allows staff to ask questions. Fast forward a few years and that tool, Lynx Insight, is now a feature of newsrooms around the world. Recently, machine learning has taken a big step further in making complex academic research easier to screen for science writers and journalists. Scholarcy’s new deep learning engine – Smart Synopses – creates journalistic style reviews of scientific research that are both accessible, but also isolate and highlight key information needed for accurate and reliable news stories. These reviews include details about study participants, sample sizes, limitations and how the research builds on or deviates from the findings of earlier studies. Sample output can be viewed here: For science journalism to prosper, the industry will need to adopt new tools and technologies to help translate the right information and deliver this to readers in an accessible way. By embracing AI-powered tech, capable of automating informative snapshots of large volumes of research, journalists can keep on top of new papers, more easily pull out the crucial stories to report on, and feel more confident about accurately reporting on complex but significant studies.


(1) 2018. No Enemy. [online] Available at: <> [Accessed 25 July 2021].(2) 2018. Too much academic research is being published. [online] Available at: <> [Accessed 25 July 2021].(3) ​​Makri, A. What do journalists say about covering science during the COVID-19 pandemic?. Nat Med 27, 17–20 (2021). <> [Accessed 25 July 2021](4) Makri, A. What do journalists say about covering science during the COVID-19 pandemic?. Nat Med 27, 17–20 (2021). <> [Accessed 25 July 2021](5) 2020. How Times Reporters Handle Scientific Studies. [online]. Available at: <> [Accessed 25 July 2021](6) 2012. Should science journalists read the papers on which their stories are based? [Online]. Available at: <> [Accessed 25 July 2021][/vc_column_text][/vc_column][/vc_row]