Monday 22nd January: A middle ground for audio articles

Good morning! Today’s newsletter is brought to you by Esther.

On Wednesday we’ll be releasing the first of our four-part special podcast series looking at futureproofing local news, supported by the Google News Initiative. Search and follow ‘Media Voices’ wherever you listen to podcasts to get the alert, or of course we’ll link to it from here when it’s live.

We’re about to increase the daily rate for this ad slot in the newsletter as subscriber numbers have grown, but before we do so, you have until the end of January to book slots (for any weekday this year) at the current rate!

The popularity of audio articles is well-documented now, with most publishers who have added a ‘Listen to this article’ function reporting success with it. But with audio articles comes a choice: spend money on getting a person to read it out, or plug in a robot. Both have their pros and cons, which I won’t get into here. Now, however, Schibsted is taking the best of both worlds, and has trained a custom AI voice based on Schibsted podcast host Anne Lindholm.

I’m hugely sceptical of the ability for AI voices to ever replace human podcast hosts. But here, Schibsted has taken a brand-familiar voice and developed something that makes offering audio articles across the brand viable.

This is a vendor post on INMA from project partners BeyondWords. The results speak for themselves though; after just a few months live on Aftenposten, audio article usage is on par with podcasts, and Schibsted has commissioned several more custom AI voices across its brands in Norway and Sweden.

Cyberattacks on the news industry appear to be getting more frequent: 70% of respondents to a survey from Sophos said they had been hit by a ransom attack in the previous year. “Even though budgets are always a consideration, imagine what it would cost to recover from an attack as opposed to preventing one in the first place,” one expert advised.

There’s some excellent points coming up about the church and state divide in news organisations in our upcoming episode with Google about futureproofing local news (coming this Wed), so the timing of this piece from FT Strategies’ Lisa MacLeod is on point. “In this modern world the notion that newsrooms should have little or no role in commercial endeavours - including subscription growth - is becoming increasingly obsolete, if not completely counterproductive to sustainability and growth,” she writes.

Propensity models use lots of data points to decide how likely it is that customers will take a particular action. We come across them a lot with paywalls, where publishers use these models to decide whether to gate an article or show a certain offer based on behaviour. Here, Greg Krehbiel outlines four ways publishers can use propensity models, as well as the downsides of personalisation.

More from Media Voices