We had an AI write a review for the Pixel 9 Pro. Here’s what it said:

There’s a lot to look forward to in Google’s new Pixel 9 Pro and Pro XL, from a refreshed design to a suite of new AI skills. And while many of those skills are image-based (the wild generative AI Pixel Studio is one of them), Google’s Gemini Advanced can provide detailed (though sometimes inaccurate) answers to just about any question you might have.

So I thought I’d take on the ultimate challenge: a review of the camera itself. I’ve already written a review of the Pixel 9 Pro XL, highlighting the good things about it, like the updated design, longer software support, and better daytime photography, while also addressing the not-so-good things, like the monthly Gemini Advanced subscription and disappointing low-light image quality.

My reviews are based on my 13 years of experience as a technology journalist specializing in mobile and testing smartphones in various real-world scenarios. The Pixel review is based on information likely hashed from the internet. I asked Pixel to “write a 200-word review of the Pixel 9 Pro XL” and here is the result:

“The Pixel 9 Pro XL is Google’s biggest, boldest phone ever. With a sleek new design, beautiful 6.8-inch display and powerful Tensor G4 chip, it’s truly a flagship device. The camera system is outstanding, allowing you to take beautiful photos and videos in any lighting. Battery life is also impressive, easily lasting a full day on a single charge.

“But it’s not perfect. The phone is fairly expensive, and the lack of expandable storage may be a turn-off for some. Plus, while the Tensor G4 is fast, it doesn’t match the pure power of some of its competitors.”

“Overall, the Pixel 9 Pro XL is a fantastic smartphone with a lot to offer. If you’re looking for a top-of-the-line Android device with an incredible camera and great battery life, it’s definitely worth considering.”

To be honest, this isn’t bad. Like me, the review points out the “sleek new design” and acknowledges that the phone is expensive and the Tensor G4 processor isn’t up to par. But the review pushes the camera performance far beyond what I consider to be accurate, and it reads more like marketing material than an experienced and honest review.

The example above is a direct quote from one of the responses the Google Assistant gave me, but I’ve asked the Google Assistant to review the Pixel 9 Pro XL multiple times, wording it differently each time, and interestingly, the results have changed quite dramatically.

Some reviewers pointed out the poor macro shots, while others lamented inconsistent battery life. One reviewer concluded, “If you’re looking for the absolute best in terms of performance and design, other options may be better for you.” Another said, “The design is dated and the price is too high for what you get,” and called the AI ​​features a “gimmick.”

The results were inconsistent and generated different opinions each time the query was run. This is simply the nature of AI search – factors like where the information comes from and how the algorithm puts it together are inconsistent. This is why if you ask someone to write a poem about a circus horse 100 times, you’ll get 100 different poems.

But we saw more than just simple inconsistencies: We also saw factual errors. Some reviews misrepresented the Tensor G3 processor (instead of the actual Tensor G4 used) or got key specifications wrong. In our testing of Gemini Advanced, we found general misinformation: the service recommended that we listen to songs that didn’t exist, or told us to drink beer at a bar that had long since closed.

The lesson? Be careful how much you trust AI-generated answers. Sure, we’d love to say you should trust CNET experts over an AI if you’re planning on buying a product, but this applies more broadly: trust your doctor over AI-generated medical advice, trust a real electrician when it comes to rewiring your house, and use your common sense when an AI tells you to eat glue and rocks.