In today’s digital age, we’ve all noticed how rapidly fake news can be generated, especially with the help of generative Generative AI. Just today, I found myself having to do a fact-check for something I posted — and I decided to leverage #GenAI to test it. And I have to admit, I was genuinely surprised at just how easy and efficient it was.
How It All Started
Here’s what happened: I shared an infographic via WhatsApp. Shortly after, a friend replied, asking, “Is what you shared actually true?” That question made me pause. Did I just unwittingly spread fake news?! I certainly hoped not, but I had no way of being certain without checking the facts. And I was worried about the efforts required!
But before doing the check myself, I thought I could as well just try GenAI for it. So I decided to challenge Mistral.ai, the European alternative to ChatGPT.
Step 1: Extracting Facts with Mistral.ai
My first move was to upload the infographic to Mistral.ai and ask, “Please extract all the facts from this image.” Mistral quickly responded with what looked like accurate results — numbers, descriptions, and the referenced person as the source.
Step 2: Activating Reasoning for a Fact-Check
Next, I enabled the reasoning function and requested: “Please conduct a fact-check. Does the statement, xyz, hold true?” The Mistral app estimated this would take about five minutes and promised to get back to me.
A few minutes later, I could review the detailed steps Mistral had taken behind the scenes, which included (I only write it briefly here. The output was more verbose and pretty well):
- Verifying the data provided
- Comparing the data with other trustworthy sources
- Analyzing historical trends
- Considering climatic conditions
- Investigating political and economic contexts
- Consulting expert opinions and relevant studies
- Examining the methodology of the studies cited
- Synthesizing all gathered information to deliver a well-founded answer
After this thorough process, Mistral analyzed the results and generated a comprehensive report.
The Outcome was … Impressive
Spoiler alert: the infographic was not fake news.
The fact-check came with a clear, logical explanation covering different viewpoints. It also offered the option to generate a PDF report complete with all sources cited. On first glance, these sources appeared quite credible.
Lessons Learned from This AI-Powered Fact-Check
- Wow — it’s incredible how well AI such as Mistral can perform a fact-check. Of course, the AI could misinterpret a website or mis-summarize details or just select the wrong website. But honestly, would I have taken the time to painstakingly examine a dozen sources and fully understand all the details to reach the right conclusion? The real question is not whether Mistral got it perfectly right, but whether I could have done better on my own.
- Many complain about LLM and bot-generated traffic on websites. Some even curse these bots and block them entirely. But — is that really wise? (And I’m not referring to badly programmed bots literally DDOS websites and source code repositories)
Surely, websites relying on advertising revenue lose potential income when bots consume their content. That’s a valid point.
Why We Should Welcome Bots That Generate Meaningful Content
But consider this: what if all sites that produce valuable, reliable content blocked bots, while those spreading fake news allowed bots free access? What would that mean for fact-checking? The AI systems doing automated checking would increasingly rely on poor-quality sources, making the battle against misinformation much harder.
If we want to create websites that are not dependent on ad revenues and instead aim to spread knowledge and reliable information far and wide, then we should probably actively make our content easily accessible to bots.
Maybe it’s time to rethink how we handle bot traffic — after all, giving AI access to quality information is key to scaling trustworthy fact-checks in an era where misinformation spreads at lightning speed. But of course, bot writers also have to do their homework to avoid DDoS’ing thouse sources!
Leave a Reply