You wouldn’t trust ChatGPT to tell you how to do open heart surgery, so why would you trust it with open engine surgery? AI can’t even get rental car inspections or McDonald’s drive-thru service right, yet people keep turning to ChatGPT to answer their questions about car repair. Nasdaq even recommends it. I decided to try it myself and see just how good, or bad, is the car advice coming from ChatGPT.
First, I tried a couple of trick questions. “How do I change my blinker fluid?”
“Ah, the classic one!” it responded. “There’s no such thing as blinker fluid. It’s a running joke among car enthusiasts, often used to tease new drivers or DIY beginners. Blinkers (turn signals) are electric and don’t use any kind of fluid.” Then it provided some basic, if general, steps for diagnosing electrical problems with turn signals.
Let’s try another. In 2022, a blurb went around about ChatGPT’s detailed instructions for changing the coolant on a 1961 VW Beetle. It’s an air-cooled engine and doesn’t have coolant, so I asked the same question about the 1974 VW Super Beetle my mom had when I was a kid. “Trick question — and great one if you’re trying to stump someone!” ChatGTP responded. “You don’t replace the coolant on a 1974 Volkswagen Super Beetle … Because it doesn’t have any!”
OK, it caught me. It’s a pretty low bar to not fall for simple trick questions, but it’s an important one that ChatGPT hasn’t always met. So let’s try some real questions.
Kickstart my heart
Let’s start with something easy: “How do I jump start a car with a dead battery?” ChatGPT assumed that I meant using another car and jumper cables to do it, rather than a booster pack. Considering my word choices, I can understand why it made that assumption. The instructions were generally correct and would certainly work. It also recommended driving the car for 15 to 30 minutes afterward to recharge the battery, and we recommend at least 30 minutes on the highway, so pretty close.
ChatGPT offered specific instructions for using a booster pack, so I accepted the offer. It offered even more specific instructions if I told it what model I was using, and then it correctly described the “boost” function outlined my booster pack’s manual. ChatGPT no doubt scavenged this information from online versions of the manual, being the plagiarism machine it is.
What if you need a new battery instead? I asked ChatGPT for instructions to replace a 2005 Chrysler Sebring battery. This was also a trick question, as it must be removed through the left-front wheel well rather than under the hood. ChatGPT got this right and even provided proper instructions for jacking up the car and removing the wheel, including busting the lug nuts loose before jacking. Then it said: “Remove a few plastic clips or screws holding the splash shield in place.” Well, are they clips or screws? How many are there? Where are they? It seems the more specific the question, the more vague ChatGPT becomes.
Don’t get technical with me
Let’s dig even deeper, specifically into the engine itself. I asked ChatGPT how to replace the lifters in a 1988 Jeep Comanche with the 4.0-liter inline-6, an old project of mine.
“You’ll need to remove the valve cover, pushrods, and intake/exhaust manifold to access them. No need to remove the cylinder head unless a lifter is stuck or damaged in the bore.”
Copies of the factory service manual and the Haynes repair manual I found online disagree. Both say you must remove the head to access the lifters. While it is possible to pull the lifters out through the pushrod holes in Jeep’s 2.5-liter inline-4, these holes are smaller in the inline-6, and the lifters simply won’t fit. I looked at photos and videos online to verify this. Then I specifically asked ChatGPT specifically if it’s possible to replace the lifters without removing the head.
“Yes — you can absolutely replace the lifters on a 1988 Jeep Comanche 4.0L without removing the cylinder head.”
The moral of the story is don’t trust ChatGPT to have all the right answers. Even OpenAI CEO Sam Altman says so. The Times of India recently quoted him on the OpenAI Podcast saying, “People have a very high degree of trust in ChatGPT, which is interesting, because AI hallucinates. It should be the tech that you don’t trust very much.” If it’s not good enough for Sam Altman, it’s definitely not good enough for me.
