Q: How are publishers using AI in product development?
A: Publishers are taking a measured, iterative approach to AI rather than chasing big, standalone launches – embracing measurement and considered iteration over absolute innovation. AI is enriching their existing products, growing their intelligence to learn more about what the product could do and what their audience wants and needs. Rather than shipping AI-only or AI-generated products to their audiences, they're using AI to deepen product intelligence, better understand what audiences want, and work out where AI fits into decision-making – so it can fine-tune experiences and automate parts of the workflow, such as how and when content surfaces to specific segments. Standalone AI launches often struggle precisely because they ask audiences to change behaviour overnight, whereas iteration lets improvements arrive quietly, inside habits readers already have.
This shows up in a few ways: automatic grouping of customers, personalised story / watch / listen lists, and more analytical product development based on usage patterns. Of course, it’s also helping in areas like live sports scores, where – if we could transport ourselves back to the golden 80s like my AI-generated headshot accompanying this article – an automated feed telling you that Wolverhampton Wanderers legend Steve Bull has scored in the 75th minute essentially powers a live news page with thousands of viewers – and will embellish it, too, stating just how much that screamer of a volley sealed the deal against West Bromwich Albion, their local rivals.
AI is also helping to create and nurture audience cohorts, building them from usage patterns that map content to demographics, which can grow usage and stickiness. Analytics drive content choices and automatically build experiences for these cohorts. The analytics and storage themselves aren’t AI, but the intelligence layered over that knowledge is what empowers editorial teams and audiences alike – helping publishers get the right content in front of the right people across their audience, and reducing their reliance on third-party platforms and algorithms to do that matching on their behalf.
Q: What is a particularly illustrative use-case?
A: A publisher combined first-party usage data with a content taxonomy to automatically group readers into cohorts and serve tailored story lists on the homepage, rather than relying on a single static front page for every visitor. The key learning wasn’t the technology itself, but the groundwork underpinning it: the model was only ever as good as the data hygiene and content tagging that fed it, and the real unlock came from pairing rich audience data with equally rich content data – without both sides in place, the AI had nothing meaningful to work with.
Success came from starting narrow – one section, one cohort type – rather than attempting a full homepage overhaul from day one. This let the team validate the logic, catch tagging errors, and build trust in the outputs before scaling further. Just as important was treating the AI as a decision-support layer rather than an autonomous editor: humans still set the boundaries of what content could be surfaced to whom, with the model doing the matching within those rules rather than making unsupervised editorial calls.
The resulting gains were real but incremental rather than transformational overnight – words read up over 15% and minutes consumed up 12% in the areas where the approach was applied, building steadily as the underlying data and rules matured, rather than arriving in one dramatic leap. The clearest takeaway: engagement gains tracked with data discipline and process, not model sophistication. Most AI product wins in publishing are, at heart, data and workflow wins wearing an AI label – and the advantage compounds with the quality of the data feeding the system, not the cleverness of the model sitting on top of it.
Q: What are your three top tips?
- Think beyond text and content creation, and into process-based workflows. Embrace AI for audience intelligence and the automatic grouping of users who should receive content selected specifically for them – essentially, the Facebook or Instagram feed model, but for your own audience. Once usage data and content data sit in the same intelligent system, this becomes entirely possible and should grow usage. It can be powerful in newsletters, customised homepages, channel pages and curated experiences.
- Don’t necessarily trust the AI built into vendor products – it’s almost always going to be weaker than what you’d get by publishing integrations that tools like ChatGPT, Gemini and Claude can use directly. Remember that most vendor AI features are just wrappers around those same tools, letting the vendor add metered revenue to your account with them. Favour products and services that happily open up and integrate with the tools your staff already use.
- Collect data everywhere you can – ideally in one system, though don’t stress about whether it’s called a “data lake” or some other awful term – and have AI act on that data to help make product decisions. Be wary of AI-only products: if they were really going to change your world, why would OpenAI and Anthropic be selling AI to other companies at all, rather than simply owning that world themselves?
About us
Full Fat Things makes publishers lives better by helping to make money and worry less about product. We build your websites and apps and sell Full Stack Things, a full stack subscription, CMS, B2B ready, intelligent and frankly fabulous way to engage with your audience on the internet.
Stewart and the other contributors to our AI Special will be taking part in an ‘AI Special – Q&A’ webinar on 24th November. Click here for more information and to register.
This article was included in the AI Special, published by InPublishing in October 2026. Click here to see the other articles in this special feature.
