Almost Timely News: 🗞️ How to Update Old Content With AI (2025-12-07)

Almost Timely News: 🗞️ How to Update Old Content With AI (2025-12-07) :: View in Browser

The Big Plug

🚨 Watch my latest keynote, How to Successfully Apply AI in Financial Aid, from MASFAA 2025.

Content Authenticity Statement

95% of this week’s newsletter was generated by me, the human. You will see Google Gemini outputs in the text and significant Gemini outputs in the video version. Learn why this kind of disclosure is a good idea and might be required for anyone doing business in any capacity with the EU in the near future.

Watch This Newsletter On YouTube 📺

Click here for the video 📺 version of this newsletter on YouTube »

Click here for an MP3 audio 🎧 only version »

What’s On My Mind: How to Update Old Content With AI

This week, let’s talk about revising old content with generative AI so as to make the content more valuable. There are a lot of ways to do this; at recent marketing conferences, I saw no fewer than a dozen sessions all showing how people could tune up old content using ChatGPT and similar tools.

Now, without disparaging other people, I disagree with a lot of what was shown. Here’s why: almost all of the approaches rely on the latent knowledge of AI models. Most of the approaches rely on asking AI to improve old content natively, meaning not bringing in any outside information. For example, paraphrasing one speaker, their prompt went basically like this:

Here’s a blog post about account-based marketing. Make this blog post more appealing and more interesting to our audience, and make it longer. Include a stronger call to action in it to (website URL), and make it more likely to convince our audience to convert as a lead.

Yeah, that’s going to go well. What’s happening here is that they’ve oversimplified the process to the point where it’s going to generate bad results. If you’re familiar with the term AI slop, this is going to generate AI slop.

So let’s try to fix this and do better, both for our businesses and our audiences.

Part 0: Fundamentals

Before we start, some fundamentals. All AI models today are trained on massive amounts of public (and not public) data. The amount of information in today’s models equates to a bookshelf of books that goes around the equator of the planet 8-12 times. It’s massive.

It’s also not necessarily current or correct. For every peer-reviewed paper with randomized controlled trials and bespoke research in AI’s knowledge, there’s equally complete garbage from Aunt Esther’s Healing Crystal and Flat Earth Blogspot blog. In fact, there’s significantly more garbage than high quality content because garbage is easy to make and doesn’t rely on inconveniences like facts, as capably demonstrated by many politicians.

That means that we cannot and should not rely on AI to be an arbiter of truth or facts. If we’re going to upgrade our older content, we want it to be better, not worse.

So the first principle of revising older content: make sure it’s genuinely an upgrade. And we make sure it’s an upgrade by using the highest quality data, the highest quality ingredients.

Next, we cannot lose sight of the basics of content marketing, namely that our content has to meet basics like the 3E rule: your content should entertain, educate, or engage (ideally all three). The corollary to that is the 3L rule - if, while you are creating content, you did not laugh, love it so much that you told a close friend about it unsolicited, or learn something in the process, the content is probably no good.

Why? Because we are ostensibly the subject matter experts in the content we’re making. If we, as leaders in our spaces, aren’t getting anything out of our content, then our audience isn’t either.

Here’s the bench test, the acid test of great content, borrowed from my friend Jay Baer. Would anyone - including you, if it wasn’t your content - pay you even a Euro/dollar/peso/pound/kroner for your content? I said in a recent YouTube video to try putting a tip jar on your content. If you don’t collect even the smallest unit of money in your local currency once during the year, if no one sees enough value in your content to tip you at least a pittance, then your content isn’t any good.

AI won’t fix that - and that’s a significant risk of updating content using AI. If the starting content isn’t any good, AI isn’t likely to make it any better.

As AI strategy expert Katie Robbert often says, “new technology doesn’t solve old problems.” Content that sucks, when processed by AI, will suck faster, at greater length.

With those warnings in place, let’s get our recipe together.

Part 1: Mise En Place

To do great content upgrades, you’ll need a variety of tools:

• Web analytics: Assuming we’re working on public content, you’ll want some way of figuring out which content to work on. Web analytics tools of any stripe, be they tools like Adobe Analytics or Google Analytics or native measurement systems like the stats provided by Substack or Wix are where you want to start.

• Large language model: ChatGPT, Google Gemini, Anthropic Claude, DeepSeek, Alibaba Qwen - pick your poison. All of today’s large models and platforms are smarter than a PhD in their field, so pick the best one you have access to. Because we’re working on public content, privacy risks are low.

• NotebookLM: An AI-powered research tool like NotebookLM is essential for today’s content upgrades because we’re going to be gathering and aggregating a lot of information together, and NotebookLM does this better than most. It’s also very capable even in its free version.

In addition to the tools, you also need great ingredients. I coined the term knowledge blocks, like digital building blocks, for this process. Some of the things you’ll want to have on hand:

• Ideal Customer Profile: Who is the audience you want? Who is the content for? We’re not making content for everyone. Presumably, if we’re doing this for any kind of business reason, we’re making content for a very specific kind of person.

• Who you are: Who are you? What should AI know about you? For content creation, at a minimum you should have things like a detailed writing/speaking style, what you’re expert in, what jargon you know, etc. I’d recommend having both - a style guide plus a knowledge block about who you are.

• What constitutes great content: What is great content, and specifically what does great content mean to you and your audience? I’ve given you two of my frameworks, but you should probably have your own, tailored to your business and your needs. Critically important, your stated goals for what your content is supposed to do for you must be in here! If your content is to generate leads, make sure that is stated explicitly. If your content is to generate newsletter subscribers, make sure that is stated explicitly. Put your goals in the great content document.