AI is being touted as a way to revolutionise modern media from automation to personalisation, data analysis to news gathering. Some of the biggest wins are being recorded with the use of audience development AI, used to create, automate and deepen reader relationships.
- A suite of AI tools known as Sophi is helping Canadian newspaper The Globe and Mail increase return visits and engagement among its registered users. The system has also doubled email registration rates and boosted subscription conversion by over 50%.
- According to the Globe and Mail’s senior project manager Sonali Verma, deploying artificial intelligence has allowed the team to automate a range of important publishing decisions. These include content curation across the Globe’s website and social media accounts, accomplishing in seconds what it previously took two people two hours to do.
- The algorithm also manages a dynamic paywall to maximise conversion rates. It predicts which pieces of content should go behind the paywall and when to present readers with content likely to convert using factors like propensity to read, device and location. The result has been an additional 170,000 subscribers and millions of dollars in incremental revenues added to the publisher’s bottom line.
Filtering out online hate
- In another example of AI being used to automate audience interactions, Austrian publisher Kronen Zeitung has been using machine learning to moderate user comments. The robot moderator filters out spam, abuse, graphic content and online hate to protect journalists’ and editors’ mental health and speed the moderation process.
- Readers post about 500,000 comments to Kronen Zeitung’s website every month. About three quarters of these are moderated by AI, flagging up words or expressions that cross community guidelines. This is a huge time-saver for human moderators, who now just review flagged content to decide if it is to be allowed or deleted.
- Peter Zeilinger, head of community at the paper says, “There has to be a human who understands the different meanings of words.” But the time saving means comments can now be published in real-time. Previously, publication could take hours, delayed by the availability of human moderators. Speeding the publication of comments has increased engagement by around 25%.
Despite the promise of AI, smaller news organisations may lack the resources to engage meaningfully, says Professor Charlie Beckett in an article for InPublishing. Becket, the leader of the LSE JournalismAI project says publishers risk struggling to reap the potential benefits of AI unless they learn to work together.
Becket says AI is a relatively new set of technologies for journalism and, although publishing is traditionally competitive, even cutthroat, collaboration might be a better strategy. “We are living in an increasingly algorithmically-powered and data-driven world. We need to keep up,” he says.
The LSE JournalismAI project has been running collaborative innovation experiments with journalists from news organisations around the world for two years. In its 2020 Collab, teams came from different roles in a range of publishers across a mix of countries.
The challenges they tackled were also varied, from subscriber churn to automated summarisation of archive content. From the practical work the teams did they produced a series of reports designed to help any publisher looking at similar challenges. Their findings were all made available publicly.