Crawford Media

Audio remains the hottest medium in news

Welcome back to Crawford Media, coming to you from a soggy Sydney summer where for the past month is has been more likely to rain than not. A great season for leeches, particularly here on the outskirts of the Royal National Park.

While I wait to record my first podcast of the year, I thought I’d delve into an interesting document: Reuters Institute’s “Journalism, Media, and Technology Trends and Predictions 2022”. As you will remember, RI puts out the Digital News Report every year, and these predictions are a kind of counterweight to that document. RI surveys more than 200 global news leaders on what they expect in the coming year, and author Nic Newman has unearthed some good stuff. I am going to ignore his overall summary – you can read the report yourself, it’s not enormous – and draw out the important points from my point of view:

  • Publishers want to own their own audiences (mostly through subscriptions) and are moving away from the biggest platforms (Facebook, Google)
  • While subscriptions are the single biggest area of focus, display advertising remains important
  • With the shift to subs comes an increased focus on wealthy, highly educated audiences (who have high content demands and deep pockets)
  • Businesses are aware of subscription fatigue and are looking at ways of countering it
  • Audio remains very hot, with continuing demand for podcasts (and podcast advertising) and the development of new audio formats and products
  • The power shift to individual content creators (many previously known as “journalists”) continues, but the report expects to see big publishers bringing more content talent back in-house
  • There may also be a move to collectives, representing a hybrid between traditional publishers and solo content creators, as writers and reporters find that working alone is tough
  • AI continues to make progress

Here are my takes on the predictions above.

On the move away from platforms

Search and social remain vital sources of traffic for anyone seeking an audience. Publishers have learned to be wary of pinning entire business models on someone else’s algorithm, but you can’t beat search and social for reach. No point making wonderful content if no one knows about it. Content businesses should see these platforms as marketing tools, rather than long-term distribution channels. Traffic from social and search is relatively low quality, and audiences don’t really seem to know or care who is providing the content. 

“[Platforms] did allow for a lot of companies to acquire large audiences at various points, but they found that they were renting them.”

Brad Bosserman, CRO at Grid (speaking to Brian Morrisey)

On the problems with subscriptions

When I investigated paid search engine Neeva last year, CEO Sridhar Ramaswamy made the point that under an advertising model, the incentives of the business are not always aligned with the interests of the audience. I can see this, in fact I’ve experienced it over a career in display ad-supported businesses. The realisation that ad funding has problems can lead to the conclusion that subscriptions – getting your money directly from your audience – is perfect. Unfortunately, that’s not true. 

In a world where every digital business – which is fast becoming every business – has discovered that subs are the answer, customers are getting pretty sick of signing up. This is “subscription fatigue”. There is also the problem of “subscriber capture”, where a content creator continues to feed his or her audience exactly what they demand in order to reduce churn and build numbers. A related phenomenon, “mission fever”, is where publishers are pushed to more extreme political and ideological positions in order to better satisfy the worldview of those more likely to subscribe or contribute. Ethically minded organisations dependant on contributors seem particularly susceptible to mission fever.

The final problem with subscriptions is the focus on rich, well-educated people. These are the people who pay. If you are scraping together a business plan based on content at the moment, what kind of lunatic is going to go for an undifferentiated scale audience? That’s a battle the platforms have already won.

The result is a plethora of established players and content startups catering to rich people. BusinessDesk in New Zealand is a good example: people who want to read it have money and are in the habit of paying for information. Globally, there’s a rash of new players catering to the political and cultural elite (eg GridPuck, whatever Ben Smith’s new thing is called). They call this “super-serving”.

Audio keeps getting bigger

In 2005 I understood for the first time the power of voice. I’d been working at Radio Netherlands – in digital – and I got the chance to make an audio documentary. It was a revelation. A lot of the power in television and film is wrongly ascribed to video. Of course the pictures are important, but audio is often where the true narrative and emotional muscle resides. People who work in radio have always known this. So it is that podcasts continue to grow in importance, even without a unified and simple means of measuring and monetising engagement.

An overwhelming majority of news publishers (80%) intend to increase investment in podcasts and audio products this year.

As voice assistants gradually improve and audiences adapt to their use, the potential for audio news products grows. Audio is cheaper and easier to professionally produce than video, and can be consumed in more settings (for example, while exercising or driving). News providers are increasingly providing narrated versions of their written content: AI-generated audio from text remains a pretty poor experience. One thing many are learning is that writing for reading is very different from writing for listening.

Where will AI lead us?

“A pile of bats is sitting on the table” (OpenAI)

Last year I spent a lot of time thinking about a turning point in AI development as it relates to content generation: the ability for machines to invent compelling stories. We’re not there yet as far as I know, but there’s no reason to think that this cognitively difficult act will remain outside the powers of AI forever. When “story AI” arrives, expect the current tide of misinformation to swell to an absolute flood. Fake news is already much cheaper to produce than real news, and when compelling fiction can be churned out with no human intervention, we’re heading into new territory. Perhaps one result will be an increased emphasis on provenance and the identity and integrity of the source (whether individual/collective/publisher). Crypto will have a role to play here, no doubt.

For now, what we have is AI that can create somewhat meaningful paragraphs and convincing fake images and video. In positive terms, the ability of AI systems like OpenAI’s DALL-E to create stock imagery and illustrations from a text description is already with us (albeit not yet commercialised). This newsletter’s lead artwork was generated by DALL-E from the text “painting of a capybara sitting on a mountain at sunrise”. The example at the beginning of this section points to greater utility: “a pile of bats sitting on the table”. Editors have been reading about these kinds of abilities for a couple of years, and now want to use them in the field.

One reason for the delay is cultural rather than technical. Like children, image generation AIs have an embarrassing tendency to know nothing of propriety. They ride roughshod over correctness, making use of forbidden imagery (for example swastikas and Confederate flags), stereotypes (guns as toys for boys, dolls for girls) and cultural bias (churches represent “religion”). I find this interesting, but note the creators of GLIDE (a DALL-E alternative) characterise their AI’s failure to meet PC norms as a “safety” issue.

That’s my wrap for the moment. I’ll be back next week with the first CM podcast of 2022. I too believe in Audio.