Using AI as a Fact Checker

As I was reviewing my continually dwindling LinkedIn Learning royalties, I noticed that my latest course is officially out: Generative AI for Podcasters.

I had a blast creating this course, and am hopeful it will be a boon to my recurring revenue.

Something I didn’t really discuss in the course is how I’m using AI to assist in research and fact checking.

That’s mainly because for a long time, I didn’t properly script my shows (I still don’t fully script most of them). But I’ve been leaning more into full, or fat, outlines, where most of what I want to say is in the outline.

I’ve also been integrating more stories and seemingly unrelated facts (that I tie back in) into my writing.

So I’ve been using AI to do two things:

1. Confirm what I think/remember is accurate
2. Iron out details I’m fuzzy on

This means I’ll have Ulysses, my writing app, up on the left, and Chat GPT-4 via Raycast up on the right:

It’s been a workflow I’m greatly enjoying, and my use of Chat GPT has gone way up.

I think there are a couple of reasons for this.

First, GPT-4 is considerably better than 3.5. The answers you get are more in-depth, better organized, and it seems, less susceptible to hallucinations. It routinely reminds me of its limitations, especially around current events.

My confidence is also greater because I’m not asking it to come up with something for me, or asking it for info on topics I know nothing about. I have a thought on something, and I’m using it to confirm my understanding and my own memory.

A great example of this when I wrote about using the Socratic Method. It has been a while since I learned about the Socratic Method and the life of Socrates — nearly 20 years — so I used Chat GPT-4 to make sure I had the facts right, as well as get some more context around the topic for me to use in a blog post.

I also used it to write a primer on generative AI. In that instance, I actually had it write 700ish words on what generative AI and LLMs are. Here’s the full prompt for that one:

Write a 700 word blog post answering the question, "What is generative AI?" You should include a basic definition, a definition of large language models, some common (and maybe uncommon) use cases, and some good tools. You should also mention how important prompts are in garnering good results.

What I got back was helpful and factually accurate. I ended up adding about 300 words and taking away 150ish, but I got to where I needed to be — especially since it wasn’t core content.

As a podcaster, this can help you with scripts and supplementary blog posts, or even to fact check this are said on your show.

This can help enhance your content without adding a ton more time to the research phase.

This is definitely one of my favorite uses for AI. I hope you find it helpful!

This was originally sent as a newsletter to my members back in October. If you want to get earlier access to these articles, you can sign up for The Podcast Foundry here.

For more writing like this, subscribe to my newsletter. You can also follow me on LinkedIn