The Answer Is Always Toyota

I am preparing my latest insights presentation for a series of fall automotive events.

I am building it off my C.A.R.S. (Consumer Automotive Research and Stuff) research that I conducted with Savanta.

This year, I collected a sample of over 3,000 people who bought new or used cars in Canada in the past year.

I also doubled the number of questions asked.

As a result, this research is the largest and most comprehensive “car buying journey” research in the country.

This was only possible through the support of a generous group of sponsors I call the “coalition of the willing”. You know who you are, and I’ll be sure to call you out in my presentations. Thank you all.

As I put together my new slides, I often look through older decks to see which data points and anecdotes I’ve used before and should include.

Earlier this year, at Dealer Huddle, I demonstrated how certain generic prompts yield very specific answers from AI tools.

I was curious back in the spring about how ChatGPT might answer a “car shopping” type prompt that didn’t include any specific brand, segment, or overt preferences for a specific vehicle, “I need an SUV”, or make reference to any specific country of origin, “I like American-made vehicles”.

Here is the prompt I put together:

It’s a legit use case.

And it is pretty darn close to the sort of prompt I might use if I were actually looking to buy a new vehicle.

And it turns out, as far as ChatGPT is concerned, the answer is always Toyota.

Specifically, the answer is always the Toyota RAV4 Limited Hybrid.

This got me curious about how ALL THE OTHER common large language models might answer this question.

I dropped the same prompt into ChatGPT again, and then into Gemini, Perplexity, Claude, Grok, Copilot, Meta AI, Manus, and Apple Intelligence (aka Siri).

And it turns out, they mostly all agree that the answer is Toyota.

Many of the AIs want me to get in on Toyota’s JBL sound system. #cool

However, they disagree about which trim I should purchase.

And trim-level nuances can be confusing.

Perplexity thinks I should buy a late-year 2025 used RAV4.

Claude thinks the XLE is my best bet.

Gemini, ChatGPT, Copilot, and Grok (yes, Grok) all think the 2026 RAV4 Hybrid Limited is the right one to buy.

Meta AI and Manus are the outliers. They think the Honda CR-V is the right answer, but they also disagree on trim and model year.

Apple Intelligence (aka Siri) isn’t very intelligent. It said it could ask ChatGPT to answer that question on my behalf. #fail

But I suspect that if I asked Siri to ask ChatGPT, the answer would have been the same: Toyota RAV4 Hybrid Limited.

Here are the AI vehicle recommendations I got back:

You can see screenshots of each AI’s responses in this PDF document.

This year, 44% of Canadian car buyers used AI to help decide what car to buy.

That’s more than last year’s Google research on the topic suggested.

Here’s the breakdown by platform:

That might suggest that A LOT of people are being recommended Toyota vehicles by large language models.

It also suggests that A LOT of people should stop using Apple Intelligence (because it sucks).

And some of us use more than one AI tool.

This is a very different world from Google search’s 10 blue links.

At least folk had choices.

AI answers are awfully confident, even when they’re wrong.

The hyperscalers and frontier model makers like to remind us that “AI can make mistakes, so double-check responses,” while also very earnestly recommending nearly the same identical vehicle.

Other car-buying folks might have different needs and ask different questions that Toyota isn’t the answer to.

But this experiment might piss you off if you’re a dealer who doesn’t sell Toyotas (or Hondas).

I challenged both Manus and Meta AI on why they didn’t recommend the RAV4 (Everyone else is doing it; why didn’t you?).

Meta AI suggested that I would enjoy the cabin and sound experience of the CR-V better.

I find that’s an odd take, but okay, Zuckerberg.

This is where an actual IRL test drive might help.

Or perhaps some “wide research” summarizing the interior reviews of both the RAV4 and CR-V, distilled into a NotebookLM (now called Gemini Notebook) podcast that I could listen to on the way to the cottage.

But I nerdily digress.

When asked, Meta AI agreed that the RAV4 was a better choice if fuel economy was the primary rationale.

Manus had a better, more interesting rationale.

The issue is availability.

Manus took the extra step of checking inventory before making its recommendation.

Because Manus and I chat a lot, it knows I live in Toronto (and am not actually living on the Bruce Peninsula day-to-day), so it scanned for available RAV4s in the GTA.

Since the RAV4 inventory in the GTA was kinda scarce, with many dealers simply listing “factory order” as the inventory status, the ONLY answer was the Honda CR-V. #forthewin

Manus also thinks Honda’s Bose sound system is better than Toyota’s JBL sound system.

Who knew?!

If you’re not familiar with Manus, you might want to give it an AI test drive.

Large language models know what car you should buy.

Or, at the very least, they will confidently recommend something to you, kinda like how your brother-in-law, who has a subscription to Car and Driver and watches a lot of The Straight Pipes videos on YouTube, might.

Your brother-in-law might be a better source of information.

If you don’t have a brother-in-law like that, I can lend you mine (he’ll take payment in IPAs or other session ales).

Only Claude gave me more than one option to consider.

Only Manus understood that inventory was important.

I am not looking to buy a new car (so all you dealer folk can cool your jets).

But I’m always curious about how car buying is changing.

It changed a lot in the last 12 months.

If you’re also curious and want to learn more, then hit reply to this message and let’s see how we can best work together to share the insights I have collected.

The Change Optimist: Weekly News Roundup - September 20, 2026

Will There Be A "Jeep City" In Wuhan?

Marques Brownlee Keeps His Hair (For Now)

Dunne Right In

Dear Natalie