Google NotebookLM and the Era of Ultra-Personalized Content

If you can create a podcast based on a topic, why not a TikTok? Or a documentary? Or a longform article? Media is about to get weird.

Tailoring content to a user is about to get trivially easy. The question is: should media companies do it? Or should platforms? (Credit: Midjourney)

I'm not sure when it happened exactly, but the Wall Street Journal recently began offering AI-generated summaries at the beginning of some articles. The "key points" aren't unusual for a news site, but the fact that they're AI generated (and edited by a human) is somewhat novel. To my knowledge, there was no announcement about the practice, and even if there was, it likely would have been drowned out by bigger AI headlines about funding or rumored new products.

The Wall Street Journal’s AI-generated bullet points for articles.

But this small step is representative of the bigger leap that AI represents: removing barriers between readers and information. Summarization is usually the first phase when building an AI-powered experience, but it's by no means the last. Once an LLM has processed an article, converting it into another format becomes more about engineering than storytelling.

Google NotebookLM epitomizes this idea. Part of Google's suite of tools built around Gemini, NotebookLM is a very simple idea: put a bunch of files in a folder, then use Google's AI to interpret them. While that idea isn't groundbreaking — there are plenty of "chat with PDF" products — its Audio Overview feature is: Hit a button, and in a few minutes NotebookLM creates a podcast all about your files, hosted by a pair of AI voices.

The feature is proving so popular that it's easy to predict that this idea — an "instant podcast" button for content — will soon spread to other products. Also, granular controls are clearly needed. Google has rapidly iterated on the feature, but depending on the complexity of the topic and your personal preferences, you might want a longer podcast, or something more like a debate panel. Why not celebrity voices while we're at it?

Then you start to wonder: why stop with podcasts? Why not give the reader the option to get the information via a longform feature article, or a YouTube documentary, or a series of TikTok videos? Obviously, creating visuals to accompany stories is much more complicated, with a whole different set of variables around sourcing the material, but it's technically possible. This appears to be a big part of the operations of Channel 1, the AI-driven news service that's set to launch before the end of the year.

As AI-powered content transformation becomes an option for media companies, it presents a dilemma: Should you build those experiences in-house, or should you leave them to the major tech platforms?

Let's think it through.