Fast, Personalized Fact Checking
It can feel impossible to stay informed with any hope of accuracy. Fortunately, AI tools are making it easy to create summaries with citations, even on complex topics.
We all know the incentives on social media are misaligned.
The algorithm rewards fast, hot takes. By the time you research and create a detailed reply, the conversation may have already moved on.
This is a huge disservice to the quality of online discourse. It also exposes you to reputation risk if you make bold claims on controversial topics only to get easily fact-checked.
I want to say I’m immune to this behavior, but I am not. Suppose I have a minute to reply to something before my next meeting. In that case, there’s no way I will take 5-15 minutes to Google around and compare notes across 5-15 sources. Maybe when I’m retired!
Luckily, I have a new path forward.
Consider this exchange I had on X today. It’s a great question and one I’ve thought about for a while. Apparently, 10% of Americans (about 25 million) are ex-vegan. But why? I only had a minute to reply, so I responded with this. My first belief is that it’s largely an issue of practicality, followed by some potential issues around it being long-term sustainable for one’s health. This is not a bad impulse answer, but I knew it was incomplete and probably semi-inaccurate.
Enter AI.
Two powerful products have been released in the last 2 to 3 weeks. DeepSeek and OpenAI’s Deep Research. While there is still a lot of concern over LLM‘s hallucinating answers, these models try to reduce this through interation. What does that mean? It means it passes the question and answers back through its system multiple times to keep reasoning through its answer until it feels confident it’s converged. In short, it thinks (a lot) before it just blurts out an answer. Note this is the same solution to a lot of social media fighting. If people took an extra 30 seconds to think before responding, I bet their responses would be more coherent and accurate.
So I gave it a whirl. I took a simple question and posted it to both products. You can see the results of each on the links below.
The biggest takeaway for me was that I was partially right. There are approximately five overarching categories, but I only hit on two. I was also off on the percentages.
Now, this isn’t a huge deal. Still, the fact that I had information from 25 sources (including the main one in question), meant I could have some degree of confidence this was thorough and accurate. Given that the sources were listed inline, I could easily dig them to confirm specific parts or passages.
The big thing for me is this. Instead of taking 5 to 15 minutes to do this research over many Google searches, I could fact-check this in about 15-30 seconds. This is a game changer, meaning I could easily use it in the heat of a social media debate without slowing down and trudging through the research.
You can think of this as your new, personalized fact checker. And it’s fast.
I like this more than fact-checking websites like Snopes for a few reasons. One is to get a better, unbiased opinion. Two, you can fact-check for situations that are updating in real-time, which is typically very difficult to do.
Anyway. Give both a try and see for yourself.