Whoops. I just tried to use ChatGPT to help me with a Nintendo article and it got a key detail wrong.
It also had trouble counting.
Here a Game File, I am skeptical that ChatGPT can make me a better writer or reporter. But I am an open-minded sort, and I’ve had people recommend I try the tech to enhance my journalism.
So, perhaps this quick example might be instructive for everyone.
Less than an hour ago, I was marveling at the length of the title for Nintendo’s Switch 2 release tomorrow: Super Mario Bros. Wonder – Nintendo Switch 2 Edition + Meetup in Bellabel Park
I wondered: Is this the longest title Nintendo has ever used for one of its game?
I Googled for a list of Nintendo game titles. I found one, dating back to the 1980s.
As I started scrolling past the likes of “Sin and Punishment: Successor of the Earth” and “Super Mario Advance 4: Super Mario Bros. 3-e Series 1”, I realized this was an idea I did not have the bandwidth to turn into reality.
A thought flashed. Could AI help me?
I loaded ChatGPT and asked for an assist:
“list the 10 longest video game titles on this page, from longest to shortest https://nintendo.fandom.com/wiki/List_of_Nintendo_games”
Quickly, ChatGPT produced an answer:
Great, right? 10 titles by length, cool trophy emoji, some methodology notes that assured me that “I considered full visible titles exactly as written.”
There was just the wee little problem that ChatGPT’s list did not include “Super Mario Bros. Wonder – Nintendo Switch 2 Edition + Meetup in Bellabel Park”
I asked:
why does your list not include "Super Mario Bros. Wonder – Nintendo Switch 2 Edition + Meetup at Bellabel Park"?
It answered, providing three possible excuses and three actual excuses:
“The list I gave was incomplete, not correctly exhaustive”
Really!
I asked ChatGPT to re-type its list with character counts at the end. It could confidently miss game titles, but surely it could count numbers good.
It replied with this list, complete with apologetic footnote for excluding the Mario title.
Notice how ChatGPT was now offering to create a “fully accurate” list? I mean, sure, that’s what I wanted in the first place. I get more helpful first-time answers when I’m asking my kids what they want for lunch.
I asked the AI for an accurate list using all the game titles on the page.
The result:
See that game up top?
“Super Mario Bros. Wonder – Nintendo Switch 2 Edition + Meetup at Bellabel Park — 96 characters”
“Your example is now correctly #1 by a wide margin,” ChatGPT told me.
It even offered to “export the entire page parsed + sorted programmatically (so it’s 100% reproducible).” Very nice of it to do so.
There I had it. A full list. My pre-publication instincts proven correct. What more could a hard-working journalist ask for?
Just one other problem.
While ChatGPT was thinking (or whatever passes for thinking), I had counted the characters for “Super Mario Bros. Wonder – Nintendo Switch 2 Edition + Meetup at Bellabel Park“ by hand.
I counted 78, not ChatGPT’s 96.
I pasted the Mario title into Google Docs and hit the word count button: 78 characters there, too.
I returned to ChatGPT once more, though, full disclosure, I typed 79, not 78. But who cares, right? Just numbers…
“The earlier numbers (including 96) are not reliable.”
Ha ha. Oh, silly ChatGPT.
“If you want a truly correct result….”
I did double-check with ChatGPT about the 78 vs. 79 thing. You see how it had told me that its “key takeaway” was that my count of 79 was “correct”?
Well, after I double-checked, it gave me a “verified count” of 78.
It offered to provide me with a “character-by-character index breakdown so you can verify it independently.”
No, ChatGPT, I think you produced enough results for this one.










My issue with ChatGPT is when you have to double check the AI’s work to make sure everything is accurate. At that point, why waste time using the generative AI in the first place? I’m glad that we have real human reporters such as yourself, Stephen! Unless you’re a robot in disguise. Surely you’d be upfront about that, of course.
That said, ChatGPT correctly pointed out that Fandom pages are bad ;)