OpenAI
In the fast-paced world of publishing, content teams are focused on producing fresh stories that capture readers' attention. And rightly so. New content brings tons of people to your site. But are you overlooking your existing archives?
A common practice for publishers is to guide readers to existing content through "further reading" sections. These can significantly boost engagement metrics, but can be annoying or outright difficult for writers to populate using traditional search systems. They either can't find content they know exists, or aren't knowledgeable enough about the library to begin with. So while these sections drive reader engagement, finding the right links remains a slow, imperfect process that often gets skipped under deadline pressure.
An Assistant That Knows Your Content
You don't need a new CMS or a moonshot AI project. You need a small chatbot that has read and watched your library, and can suggest, before you publish, relevant articles to include as further reading.
Imagine: You're drafting. In the sidebar, five suggestions appear—two articles, one evergreen explainer, and a short video moment. Each comes with a one-line motivation ("builds the timeline on housing permits," "you wrote this column about this topic 2021"). You are in control. Approve, tweak, or ignore the suggestion in seconds. Even more: next time, the suggestions feel a little closer to your intuition because the assistant learned from what you did.
A system like this is in place at the newsroom of 'Het Laatste Nieuws', one of the largest newspapers in Belgium. According to Wim Hellemans, Editor In Chief at Het Laatse Niews:
That one [the assistant] takes an article and looks for other relevant articles. It is being used a lot because it's very useful and has become really good. Doing this by hand is a bit of a tricky task. I don't have any data but I don't see why you'd ever not use it. For me it saves 2-5 minutes per article, but that could be even more for a less tech savvy journalist.
Why RAG Makes It Work (With Receipts)
Under the hood, this is retrieval-augmented generation (RAG): the system first searches your archives for relevant content, then generates contextual suggestions that cite the source. Recent advances in semantic search and the availability of large language models (LLMs) have made this practical for mid-sized publishers—you're no longer competing with Google's R&D budget.
Two technological developments make this work now: search systems finally understand meaning, not just keywords, so "that video of the Prime Minister arriving on a bicycle" becomes findable even if the title never used those words. And video and audio can be indexed by their content through speech transcripts and simple visual cues, so clips are discoverable by what happens in them, not only by how they were tagged.
What Changes You'll Notice
You'll have more internal references in your new content and those links will see more clicks because they are more relevant. If your staff has been complaining about the archival search, they will now be delighted. Writers rediscover more of the library because they're searching by meaning ("opinion article about zoning regulations in Amsterdam") rather than relying on metadata.
- More internal references in newly published content, with better click-through because the suggestions are more relevant.
- Less frustration from journalists and editors who already know useful archive content exists but cannot reliably find it.
- Faster rediscovery of evergreen archive material through meaning-based retrieval instead of fragile metadata search.
If you're curious, start small: point the assistant at a single desk and a bounded slice of the archive, run a test for a couple of weeks, and see if reporters stop tab-hopping and start clicking "approve." When the right suggestion shows up and you see your old content come to life again, you'll never go back!