The growing role of LLMs in transforming content has sparked industry discussion about the need for publishers to enable “liquid content.” This post will review this new term and discuss its implications.
Liquid content was introduced in a new report from the Reuters Institute at Oxford University. The report predicts that publishers will be “looking beyond the article, investing more in multiple formats especially video and adjusting their content to make it more ‘liquid’ and therefore easier to reformat and personalise.”
The Reuters Institute cites new genAI tools that allow users to consume content tailored to their preferences. It notes:
These developments mean that content is becoming increasing ‘liquid’, in that the format can be changed – actively or passively – based on the viewer’s context, interaction, time, or location. This means that it will be harder for publishers to control how news stories look in the future. It will also be harder to know how content is being used. If an AI browser automatically summarises content on behalf of a user, does this count as a human visit? With more agentic bots reading content how will measurement and therefore monetisation be affected?
Digiday picks up on the theme, asking WTF is liquid content? It notes that many publishers are experimenting with generative AI tools to offer readers alternative formats and editions. Developments in AI technologies have made it easy to do this. However, they note, two major risks:
- The accuracy of AI-transformed content can be uncertain
- The reader demand for such alternatives isn’t yet demonstrated
Does content want to be liquid?
Interest in breaking free from the past constraints of articles seems to be building.
Liquid content is similar to another recent term, kinetic content, that’s been coined by a new group I participate in, the Kinetic Council. Kinetic content refers to content that can be “combined and recombined, have real-time data integrations, be distributed in multiple ways to multiple audiences, and with as little human intervention as possible.” Kinetic content is technology agnostic, unlike liquid content, which seems focused mostly on generative AI technologies such as LLMs and agentic AI.
The question is: how far can generative AI technologies go in realizing liquid content?
I’d suggest this will be a learning process for both readers and publishers.
First, readers are starting to use third-party tools to transform publisher content. They are learning what capabilities these tools offer, and importantly, taking responsibility for the choices they make when electing to transform existing content. If the output isn’t good enough for them, they can’t blame the publisher.
Publishers, meanwhile, will see how much and what kinds of interest readers have in alternative formats and editions. They will face more pressure than third-party tools to offer accurate outputs. But they will also have far greater control over these outputs. They can curate the information sources, the editorial style, and the agentic procedures that shape the outputs. Generative AI alone may not be sufficient for some topics and publishers, but it can deliver credible outputs across a growing range of use cases.
Liquid content is not a new aspiration among content professionals. Now, the barriers to realizing this goal are falling.
— Michael Andrews