The most questionable AI claims we heard at CES 2026

CES 2026 did show real progress. Screens look better, gadgets are lighter, and sensors are sharper. But the loudest theme on the show floor was not a breakthrough. It was a habit. “AI” is now the word brands reach for when they want a product to sound finished, even when the idea is still half baked. The most revealing moments were not the slick demos. They were the pitches that wobbled the second you asked a plain question. Does it work without the cloud? What data does it collect? What happens when it gets it wrong? Who is responsible when it guesses?

When the Clippers want to be a coach

Glyde’s smart hair clippers tap into a real anxiety, most people do not know how to pull off a clean fade at home. The pitch is that the clipper adjusts as you move, and an AI coach guides you while you cut. But haircuts are physical and personal. They depend on feel and small judgment calls, not just instructions. A tool that sounds confident does not make the advice correct, and the penalty for being wrong is instant. There is no undo button, only a patchy fade and a long wait.

When health language gets slippery

SleepQ sells the idea of pills plus AI. It pulls data from a smartwatch or sleep tracker and suggests when you should take a sleep related supplement. On paper, that is a reminder system with a personal schedule. The problem is the framing. When a product uses clinical-sounding language, it can make readers assume medical value that is not there. Timing advice is not treatment, and wearable data is not diagnosis. If a brand wants to sit near health, it should state clearly what it can do, what it cannot, and what evidence supports it.

Appliances that turn “smart” into spending

Deglace’s Fraction stick vac claims it can predict issues early and assign “health scores” to parts so replacements are easy to order. Modular repairability is a good direction. The worry is the score itself. If the same system that declares a part “unhealthy” also sells the replacement, the incentive is obvious. Without a clear explanation of how those scores are calculated, you are being asked to trust a black box. The same trust gap shows up in other show-floor ideas too, like microwaves that promise AI cooking guidance or drink machines that try to judge age and sobriety from a camera. The message is clear, but if the night photos don’t deliver, the whole launch falls apart.

Calm products turned into prompt machines

An E Ink picture frame should be simple. A low-power display for art or photos that you set up once and forget. Add AI image generation, credits, and paid top-ups, and the product changes shape. It stops being a frame and starts acting like a meter. Instead of showing what you already care about, it nudges you into generating more. That might be fun, but buyers should be honest about the trade. Are you buying a device, or signing up for ongoing usage.

AI for children deserves the toughest questions. A device that invites kids into open ended chat, and can react to a camera feed, is not a harmless toy feature. It is a system that can shape trust. This is not only about bad replies. It is about confidence without judgement. If companies want to sell this category, they need plain answers. What data is stored, who can access it, can parents fully turn off the camera, and what happens to chat logs.

AI is not the problem. Vague AI is. If a brand cannot explain what the system does, what data it needs, and how it fails, “AI powered” should read like a warning label. Before you buy, ask three things. Where does the data go? What still works offline. What happens when the AI makes a wrong call? If the answers sound like copywriting, walk away.