The robot pulled off its demonstration. The lock opened without a key. The mirror commented on how you looked. All very well. But who will replace their batteries, what will they do without the Internet, and what problem do they actually solve? In September 2026, looking back at the promises of CES with a few months’ perspective means shifting the spotlight: less focus on the “wow” factor, more on the product’s life after unboxing. Useful innovation is not necessarily spectacular. It is what remains desirable once the novelty has worn off.
A note on methodology: this analysis does not claim to provide a verified assessment of the products presented at CES 2026. It draws on previously established trends and events; the potential developments discussed for 2026 are forward-looking. The aim is to offer a framework for assessing announcements at the show, not a ranking based on unverified performance.
The show reveals potential, not everyday life
CES knows how to turn a technical possibility into a compelling performance. Perfect lighting, a controlled network, a short scenario: the conditions are designed for success. At home, Wi-Fi struggles to penetrate a wall, several people use the same device, and the manufacturer eventually launches its next model. It is in this less photogenic environment that value is measured.
Previous editions have established several trends: the proliferation of connected home devices, voice interfaces, health sensors and the addition of artificial intelligence features. Since the rise of generative AI, almost any product category can promise an assistant. Yet recognizing a request is not the same as executing it correctly or saving time. The decisive question is not “is it smart?” but “is it better than the existing solution?”
1. Repairability: look for the part beneath the casing
The first test is to imagine the most ordinary fault. A worn-out battery, a torn-out connector, a jammed wheel. Can the component be replaced without destroying the device? Is the part available to buy, at what price, and with what documentation? Visible screws are a clue, not a guarantee: an accessible component may still be impossible to source or depend on software pairing restricted to the manufacturer.
France introduced its repairability index in 2021. At the European level, the repair directive adopted in 2024 also marks a shift, with its impact depending on the categories covered and its implementation. These frameworks do not certify every device at the show. They nevertheless establish an essential principle: maintenance is no longer simply a matter of manufacturers’ goodwill.
For a robot vacuum, ask about the price of brushes, the battery and wheels. For a camera, ask about the mount or power supply. The duration of software updates also needs scrutiny: a physically intact device can become unusable when software support is abandoned. A credible promise of longevity must therefore cover both hardware and software, with time-bound commitments.
2. Independence: lasting without a socket—and without a server
Battery life is not just about the hours listed on a specification sheet. A watch may last a long time on standby and much less with its sensors active. A device that needs charging every evening adds a small chore; five such devices create a logistical task. Battery life must be compared with equivalent features enabled, and real-world testing should be sought out.
There is also functional independence. What remains when the Internet goes down? Does the switch still control the light? Does the lock retain a way to open it locally? The shutdown of services for certain connected devices has already demonstrated the model’s fragility. In 2024, Spotify notably announced the discontinuation of Car Thing, its in-car accessory: owning the hardware did not guarantee continued service.
The Matter standard, launched in 2022 to facilitate smart-home interoperability, is a useful step forward. But its logo guarantees neither every feature across every ecosystem nor complete independence from the cloud. A good habit is to request an offline demonstration, then check whether essential settings and automations still work. Fallback operation deserves almost as much attention as the headline features.
3. Privacy: follow the data all the way
A sensor in a bedroom or a microphone in a kitchen does not collect abstract information. It can reveal schedules, habits and conversations. Before applauding personalization, identify what is captured, what leaves the home, how long it is retained and who receives it. An accessible privacy policy is worth more than a reassuring “your data is secure.”
Local processing offers a potential advantage: certain operations can be performed without transmitting raw data to a server. Yet it is no guarantee of privacy. A device may process voice locally while still sending usage logs. Conversely, a remote service may meet a legitimate need, provided that data collection is proportionate and the choices are understandable.
Three actions should be simple: genuinely switching off a sensor, deleting its history and passing the device on without handing over your account. For family products, the rights of different users also need examining. The person who installs the app should not, by default and without transparency, become the administrator of the entire household’s privacy.
4. Practical value: count the burdens as well as the benefits
The most demanding test comes down to one sentence: which task disappears? A connected dispenser that prevents a recurring oversight may make sense. A refrigerator that requires constant corrections to its inventory turns a promise of assistance into extra work. The assessment must factor in setup, learning, notifications, errors and cleaning—not just the seconds saved during the demonstration.
Context also changes the verdict. A voice command that is unnecessary for some may significantly improve the independence of someone with reduced mobility. A detector that makes sense in an isolated home may be redundant elsewhere. Usefulness is not universal: it is measured for a particular use and a particular person. The price must therefore include subscriptions, consumables, essential accessories and foreseeable battery replacement.
A short checklist for more informed purchases
- Repair: available parts, publicly listed prices, documentation and duration of software support.
- Stay independent: realistic battery life, local controls and operation without a subscription.
- Protect: limited data collection, explained data flows, accessible deletion and sensors that can be disabled.
- Deliver value: observable benefits, acceptable burdens and an explicit total cost.
There is no point in manufacturing a scientific score from incomplete answers. It is better to distinguish what has been demonstrated, what has been promised and what remains unknown. An unknown concerning a decorative accessory does not carry the same weight as one concerning a lock. The more a device affects security, health or access to the home, the less a buyer should be willing to accept opaque dependencies.
What next? For future generations, progress could take a less theatrical form: replaceable batteries, local AI where appropriate, essential functions accessible without an account and contractually guaranteed maintenance. These are prospects, not established achievements. The real leap in innovation would be to make these qualities central selling points at CES. A device deserves its place when it makes everyday life easier over the long term without locking its owner in.
