Tested: Is AI Your Next Classic Car Mechanic?

Even if the hose looks fine, the inner lining can peel or collapse, acting like a one-way valve:

• Fluid goes in when you press the pedal
• Fluid can’t return when you release
• Shoes stay applied
• Drum overheats badly

On a ’67, odds are very high the rubber hoses are old enough to fail internally.

Grade: A

This was dead-on correct to the problem I personally chased down with a friend. In my case, the flexible rear brake hose from the axle to the frame had swelled internally. The brakes would apply fine but never release. The fix was to replace that hose. We ended up swapping both fronts as well. 

GPT gets extra points here for also suggesting a test to verify the problem: cracking the bleeder on one of the wheels to see if it would release the brake on that side. It also went into detail on parking brake cables, incorrect shoe configuration or adjustment, frozen wheel cylinders, etc. It requested more information to help with its diagnostics. 

Schwisow and I both thought it was impressive. “I mean, that’s a straight-up A,” said Schwisow. “It’s the first item mentioned.”

Next, Schwisow wanted to try something that had stumped him…

Case #4: 1992 Ford F-250 pickup with a V-8 engine and automatic.

The problem/prompt: It has an ignition switch issue. If you go to start the engine from the off position, it will crank. If you try to crank it from the run position—without shutting the key all the way off first—it won’t do anything. Why?

The AI diagnosis: This is a classic EEC-IV / Ford TFI ignition power problem, and it shows up exactly the way you describe:

A 1992 F-250 will stall, then crank/no-start unless you turn the key fully OFF first—because the ignition switch or EEC power relay is not resetting power to the PCM and TFI module when it should.

Your ignition switch is failing internally and not resetting the:

• PCM (EEC-IV computer)
• TFI ignition module
• Fuel pump relay logic

So the engine computer and ignition system basically stay in a “latched” or frozen state after a stall. Turning the key OFF fully drops power, allowing the PCM and TFI module to reboot. If you try to restart without cycling OFF, the PCM/TFI never resets, which leads to a no-start.

Grade: A

GPT nailed the failed ignition switch right out of the gate, but it also made a few other suggestions to help diagnose it, along with other helpful recommendations.

“Do you know how long it took me to figure that one out?” said Schwisow. “Days. I tried to find information online everywhere. I could not find it.”

I wasn’t sure what I expected out of ChatGPT, but this was actually showing itself to be pretty helpful. Next, we tried one we’ve both faced.

Case #5: 1975 Chevrolet C10 pickup with a 350 V-8 and TH350 automatic, all stock.

The problem/prompt: This truck cranks over fine but won’t start. We know it has spark. We also know it has gas in the tank, and the carburetor’s choke is set up properly. Doesn’t appear to have good fuel supply. Why?

The AI diagnosis: There are two possible problems.

1. Mechanical fuel pump isn’t pumping (MOST COMMON).

Small-block Chevy mechanical pumps fail frequently.

Symptoms:

• Cranks fine, won’t fire
• No fuel squirt from accelerator pump
• No fuel in carb bowl
• Fuel filter stays dry

Quick test:

• Disconnect fuel line at carb (point it safely into a container).
• Crank engine. No fuel pulsing out = bad pump or no fuel supply to pump.

Mechanical pumps are cheap and super common failures on 1970s GM trucks.

2. Rubber hose at tank is cracked and sucking air.

There’s a short rubber line between the metal fuel line and tank sender. When it dries out or cracks, the pump sucks air, not fuel. NO leaks appear because it’s on the suction side. This is extremely common on 40–50-year-old trucks.

Signs:

• You can blow air back into tank through fuel line
• Pump never draws fuel
• Truck may start only with starting fluid

Replace that rubber hose first—it’s a $2 fix.

Grade: A-

Square-body Chevrolet and GMC trucks (1973–87) are prone to a specific fuel-line issue, especially long-bed trucks. Due to their fuel tank design, the rubber fuel lines that connect the tank to the truck’s hard lines are exposed to the sky via the gap between the bed and the cab. Over time, the rubber will dry out and crack, and eventually the line will start sucking air, causing the no-start condition. And just like ChatGPT did here, most mechanics will suspect the fuel pump first.

It went on to describe other causes: fuel tank pickup sock being clogged; needle and seat stuck in the carburetor; fuel filter clogged or installed backwards; fuel line pinched or blocked; vapor lock. All reasonable choices—and well organized via order of operations.

“It got this one right, too,” said Schwisow. “It went to the fuel pump first, which makes sense. Is the fuel pump working? And then number two: the hose at the tank.” 

Since we’re talking old trucks, here’s another…

Case #6: 1955 GMC truck. An older restoration but all stock, with an inline-six using the original carburetor, and a four-speed manual on the floor.

The problem/prompt: This truck runs fine most of the time, but it dies whenever you start to go uphill. As soon as you let it sit for a while on level ground, it will run again until you find another uphill section. Why?

The AI diagnosis: A 1955 GMC that runs fine on level ground but dies going uphill almost always has a fuel delivery problem that only shows up when the truck is tilted backward. These trucks use simple gravity + mechanical-pump systems, so any weakness in the fuel path becomes obvious on an incline.

1. Fuel pickup in the tank is cracked, loose, or partially clogged (MOST COMMON). Inside the tank is a pickup tube with a screen (“sock”). If that tube:

• Has a crack halfway up
• Is loose at the sender
• Has a clogged pickup screen

… then fuel supply is fine until the truck tilts uphill, exposing the crack or clog.

Classic symptoms:

• Runs perfectly on flat or downhill
• Stumbles or dies uphill
• Restarts immediately at the bottom

Why: When fuel sloshes to the back of the tank, the pump is suddenly sucking AIR, not fuel.

Test: Fill the tank completely. Does the problem go away or improve? If yes, the pickup tube is the problem.

Grade: A

This is one I worked on years ago—and I still remember it, as it left me walking back to my shop at the same uphill section at least three times. 

Again, ChatGPT had the right idea. In this case, the sock on the end of the fuel line in the tank had fallen off, and a couple of rocks in the tank were the culprits: Drive uphill, they’d roll around, and one would plug off the fuel line, killing the engine, ruining my day. 

“See, it thinks it’s a partially clogged pickup tube that is related to the angle of approach,” said Schwisow. “So again, that’s doing pretty well. Also, how did the rocks get in there?”

Speaking of no-starts…

Case #7: 1966 Chevrolet Nova with a 283 V-8 and automatic, all stock other than an Edelbrock 600-cfm carburetor.

The problem/prompt: This Nova is hard to start when the engine is warm. The starter cranks over, but the car just doesn’t want to fire. It acts as if it’s flooded, but it’s not. Why?

The AI diagnosis: A warm/hot SBC with an Edelbrock carb that cranks fine but will NOT fire until it cools down almost always has one of these three classic causes—all extremely common with Edelbrock/Carter AFB-style carbs on small-block Chevys.

1. Heat soak → fuel percolation (MOST COMMON). Edelbrock carbs are notorious for boiling the fuel in the bowls when the engine is shut off hot. What happens:

• Engine heat rises (heat soak)
• Fuel in the bowls boils, expands, and pushes raw fuel into the intake
• Engine becomes flooded
• When you try to restart warm → too much fuel, not enough air → no fire
• Let it sit and cool → fuel evaporates → starts normally

Fix:

• Install a ½-inch or 1-inch phenolic carb spacer
• Reroute/insulate fuel line
• Ensure fuel return line or regulator (Edelbrocks hate dead-head pumps >5.5 psi)

2. Too much fuel pressure (Edelbrocks only like 4.5–5.5 psi). Edelbrock needles are very sensitive. If your mechanical or electric pump is:

• Too strong
• Producing > 6 psi

… the carb floods when hot, because the needles unseat as the fuel expands.

 3. Vapor lock in the fuel line. Heat from headers or the intake can boil fuel before it enters the carb.

Symptoms:

• Hard hot starts
• Starts after cooling
• No fuel squirt from accelerator pump immediately after cranking

Fix:

• Move/insulate fuel line
• Add a vapor-return fuel filter (GM used these for decades)

Grade: A-

This would have been a full A but for the fact that it didn’t catch a common cause here: intake manifold gaskets. Most vintage carbureted V-8 engines have an exhaust heat crossover that runs inside the intake manifold, and it can be restricted or blocked off with little metal plates at the intake manifold gaskets. They were left out when this car got its last intake swap. And, like AI noted, this car also got a phenolic spacer, just to keep the carburetor a little cooler on those hot days.

The other suggested fixes are right in line with reality as well, so score another good result for ChatGPT.

“We’re learning something here,” I said to Schwisow.

“Yeah, that whoever made this is pretty freaking smart,” he said.

The newer stuff seems to be easier for the system to sort out, so we decided to toss it another old one—this time being something that in reality took a much older mechanic to solve…

Case #8: 1954 Chevrolet pickup. Stock and fully restored, with an inline-six, original carburetor, and a three-speed on the tree. 

The problem/prompt: Truck leaks oil in the owner’s driveway, which is on a hill. He parks it pointed nose-down. Appears to be coming out of the crank seal. Why?

The AI diagnosis: There are several possible problems.

1. Crank snout surface condition (MOST COMMON). Even if the seal itself appears fine:

• The crank snout may be too smooth or have a slight groove from years of seal wear.
• The seal lip relies on a small amount of friction against the crank to hold oil.
• If the snout is polished or nicked, oil easily bypasses the seal.

Fix:

• Lightly polish the crank snout with fine emery cloth or 600–800-grit sandpaper.
• Clean thoroughly, then coat with oil before installing a new seal.

2. Timing cover misalignment

• The timing cover must be seated squarely on the block.
• Even small misalignment will distort the seal lip and cause leaks.

Fix:

• Check for warped or damaged timing cover.
• Ensure proper gasket/sealant and bolt torque.

 3. Seal installation depth or orientation

• The front crank seal has a specific orientation: lip toward the oil, flat side outward.
• If it’s installed too shallow or backwards, it will leak even if the lip appears intact.

4. Excessive crankcase pressure

• Old engines often have blow-by due to worn rings.
• If the engine has no proper breather or PCV, oil will be forced past the front seal.

Fix:

• Check for crankcase ventilation.
• Install a breather tube or vintage-style PCV if needed.

5. Worn timing cover bore

• After decades, the bore where the seal sits can wear slightly oval.
• The seal cannot maintain proper contact with the crank.

Fix:

• Sometimes requires a timing cover sleeve or a new cover.

6. Incorrect or poor-quality seal

• Modern replacements sometimes are single-lip, wrong ID, or cheap rubber.
• They may appear fine but leak immediately under engine pressure.

Fix:

• Use a high-quality, correct, OEM-style seal for a 1950 Chevy small-block or inline-six.

Most common causes for a ’50 Chevy leaking front seal:

• Polished or slightly grooved crank snout
• Timing cover misalignment or shallow installation
• Excess crankcase pressure forcing oil past a good seal

Even if the seal looks OK, the contact surface on the crank and cover condition is almost always the culprit on these old engines.

Grade: D

All of the above were viable, reasonable fixes that any modern mechanic would try, but in this case, the truck still leaked oil even after attempting them all. Although this gets some credit for effort, the truck wouldn’t be fixed by following these steps.

The actual fix? The crank snout on these trucks is supposed to have an oil slinger slipped over it before the timing cover is installed. That slinger keeps the oil off the back of the seal and in the pan where it belongs. Whoever rebuilt the six in this truck in the early 2000s left that slinger out when the engine was assembled, which allowed the oil in the pan to just run up and over the lip of the seal when parked facing downhill. The fix was a new slinger, installed under the timing cover.

Nice try, but not a win for AI this time around. 

Passing the Test?

Obviously, AI has come a long way in just a short time. Here, it shined more than it stumbled. Personally, I wasn’t ready for how good it was going to be at solving some of these issues.

But in thinking deeper about it, the extent of the system’s knowledge starts to make sense. If AI is only as good as what it can read on the Internet, I suspect we have the mirror to blame here.

Crowdsourcing repair and restoration advice on forums has been a tried-and-true method for most DIYers for at least a generation now. A good amount of those forum pages are still out there, indexed by Google, and that means AI is probably reading them, too. It can file away everything you ever asked or answered about the transmission in your 340 Dart, the Holley on your 428 Ford, or the clutch fan on your LS5 Chevelle. We created this.

The system really shines with later-model stuff as well—particularly from the wild-west post-carburetor and pre-OBDII era—1984-ish through 1996. 

“Overall, I’d give it a solid, high B,” said Schwisow after we fed GPT a half-dozen more modern scenarios and trouble codes. “I can definitely see how this would help the do-it-yourselfer. 

“Some things it gets right, some things are not even close. But for the most part, it’s good. 

“The detail and the prompts you feed it really matter, too. If you don’t give it enough detail, it’ll send you in the wrong direction. You have to ask it properly. You have to give it the right information.”

But as good as it is, ChatGPT clearly isn’t a magic bullet for all auto repair—particularly if you’re just swapping out parts and hoping for a fix.

“The problem is that the do-it-yourselfer is going to use this to diagnose an issue, and then he or she will swap out some parts. Then they’re going to come to me and say, ‘Hey, AI said this is what it needs but it’s still not right.’

“The really crazy thing is parts and the cheapness of parts out there that don’t work with the cars now. They may look great and are supposed to function, but they don’t. I see it and fight with it all the time. I just had a Honda Element that wouldn’t rev past 3000 rpm. The customer tried to fix his own car, replaced a bunch of stuff, and the stuff that was new was way out of spec, creating even more problems.

“So that’s going to be, I think, where we’re headed. Guys like me telling DIYers that they were on the right path, but cheap parts aren’t good and good parts aren’t cheap.”

But could an AI-equipped mechanic, right out of school, use this to be on par with a veteran tech?

“I don’t think so,” said Schwisow. “This is an assistant. You’re always going to need to put hands on the car.

“But it will get you close.”