From cribbage board to battlefield, AI-generated content is not to be trusted blindly

Once again today, I find myself defending the work of my editorial team because an armchair quarterback has had an AI chatbot provide analysis of one of our articles.

In that person’s mind, he was providing definitive proof that our article was deeply flawed and driven by agenda.

Far too many people are treating AI chatbots as if they are the ultimate answer key. That is a big problem.

I will take the analysis of my editorial team over that of an AI chatbot every day of the week.

A personal story helps paint the picture …

The other day, I was playing cribbage with a buddy, and he did not trust my counting on what I said was a hand of 23 points – three fives, one 4, and one 6.

He typed in the cards on a popular AI platform, and with confidence, he said, “Nope – it’s 24.”

I said I would bet him $100 that I was right. He said, “You think you are smarter than AI?” I said, “No – but I know that these platforms are often wrong because they scrape information from the web and other places that is often inaccurate.”
He punched the hand into a well-known cribbage site, and was surprised to see my counting confirmed. Gemini is still getting it wrong at this moment as you will see in the attached image … because it arrived at its answer using the yammering of a bunch of amateur crib players on Reddit who clearly do not excel in simple math.

Too much trust in authority has never been a good idea, and AI is an authority with inherent flaws.

Large language models work with statistical averages. And the unfortunate truth is that the value of much of the content being used to fuel chatbots is far from accurate.

And it is not just the lightly educated but heavily opinionated throngs on Reddit and other popular platforms.

You will find inaccuracy in many critical matters. For example, it is no secret that a lot of BS “science” has been produced to support false narratives in the medical and environmental science communities.

For a quick tutorial on how “peer-reviewed” studies are presented as factual but are actually far from it, please refer to Exhibit A (Higher rates of adverse events after mRNA vaccines interpreted poorly by study authors) and Exhibit B (Real science requires real peer review ). The Internet is full of garbage science that has been accepted in the scientific community because of ulterior agendas – and that is then factored into the confident answers of AI chatbots.

Some armchair QBs are better than others at asking the right questions and getting down to more reliable AI answers, but, in general, we should all be reluctant to put too much faith in AI-generated content.

Of course, the stakes are higher when military commanders rely on AI (Unready for war, AI may already be causing deadly mistakes).

Indeed, be it for a crib game or war games, it is critical that we treat AI-generated information with due caution.

(Rob Driscoll – BIG Media Ltd., 2026)