Key takeaways
- AI wearable devices that have stayed on the market do one job well, usually recording or note-taking, rather than promising to replace a phone or an assistant.
- Battery life varies enormously across the category, from about two and a half hours of active use on one device to roughly 100 hours on another, and the published number is a real design spec, not a marketing detail.
- Most of these devices depend on a companion app and a subscription tier for their core AI features, which changes what happens to the device if the company behind it shuts the service down.
- An always-on AI voice recorder can record conversations in places where the other person’s consent is legally required, which is a real compliance question for both buyers and builders.
AI wearable devices that last are the ones built around a single job, usually recording and summarizing conversations, with battery life that matches how someone actually wears it all day. The devices that tried to do more, replacing a phone or a voice assistant inside a pin or a pendant, have had the roughest two years on the market. These devices sit inside the broader category of AI in IoT devices, where the same tradeoffs around cost, battery and cloud dependency apply.
A fast two years for AI wearables
The category moved quickly between 2023 and 2026. Humane unveiled the AI Pin in November 2023 at $699 plus a mandatory monthly subscription, began shipping it in April 2024, and was reviewed harshly that same month. The company shut the product down in February 2025 and sold its technology to HP. The fuller story, including what happened to buyers’ data and refunds, is covered in why did Humane AI Pin fail.
The devices that have kept shipping took a narrower approach. Plaud’s Note and NotePin models are dedicated AI recorders. Rewind pivoted into the Limitless pendant in 2024, a dedicated conversation-recording wearable, and Meta acquired Limitless in December 2025. Amazon-owned Bee relaunched its wearable line in October 2026 with two models built around ambient note-taking. Friend, a conversation-recording pendant from an independent maker, moved through several price points between 2024 and 2026 while keeping the same single job.
Free tool: check if your invention is ready to build, in 3 minutes →Why the single-job devices are the ones still shipping
Every device that has stayed on the market promises one thing: capture what was said, and summarize it. The devices that struggled hardest, by contrast, were pitched as broader assistants or phone replacements, a much harder job for a battery-powered wearable to do well. That pattern, drawn from how these specific products have performed rather than from any outside industry count, is the clearest lesson in the category so far: a wearable that does one job can get that job right, while one promising to replace a general-purpose computer in your pocket is competing with a device that has a much bigger battery and a much faster processor.
Battery life is a real spec, not a marketing line
Published battery figures for AI wearables vary by a factor of more than ten across the category. Every figure below is the manufacturer’s own published spec, not an independently tested number, and real-world results depend on how actively a device is recording:
| Device | Published battery life | Charging time |
|---|---|---|
| Plaud Note Pro | Up to 30 hours (Enhance mode) or 50 hours (Endurance mode) | About 2 hours |
| Plaud NotePin | Up to 20 hours recording | About 2 hours |
| Rabbit R1 | About 4 hours light use, 2.5 hours heavy use, 36+ hours standby | Not published |
| Limitless Pendant | About 100 hours | Not published |
| Friend pendant | About 15 hours | Not published |
A ten-to-one spread in the same product category is unusual, and it reflects a real design tradeoff rather than a difference in engineering quality. At Inventornest, battery optimization is often underestimated: a prototype can perform its main recording or AI function successfully in a demo and still deliver inadequate battery life once someone wears it for a full day. Battery performance should be treated as a core requirement from the beginning and verified through practical testing. A bigger battery is also not a free upgrade. Increasing battery size requires additional internal space, which is exactly the constraint that makes a wearable a wearable in the first place, so capacity, wear comfort and charging frequency have to be decided together, not bolted on after the shape is set.
The app and subscription question
Almost every device in this category needs a phone app and a cloud subscription to do its core AI work, which means the device’s usefulness is tied to a company staying in business and keeping its servers running. Current examples show different shapes of this tradeoff: before its December 2025 acquisition by Meta, Limitless offered 10 hours of AI-feature use per month free with unlimited storage, and a $19 monthly plan to remove that cap; subscription terms for the pendant may have changed since the acquisition, so check the current offer before buying. Bee’s free plan keeps a rolling 14-day history, with a $12 monthly Bee Pro plan for unlimited summaries. Friend offers an optional subscription of about $10 a month specifically to extend how long the AI retains memory of past conversations, beyond a 30-day default.
At Inventornest, we advise against adding connectivity when it adds cost and complexity without solving a meaningful problem for the user. A standalone version is often the more appropriate design whenever a product can perform its essential function reliably without an app or a cloud connection. For an always-on AI recorder, the cloud connection is usually doing real work, running the transcription and summarization, so the dependency is harder to avoid. The design question worth asking early, covered in depth in edge AI vs cloud AI, is how much of that work could run locally, and what happens to the device’s core function the day the subscription or the company itself goes away.
The recording-consent question buyers and builders both miss
An always-on AI wearable records other people, not just its wearer, and recording law in the United States is not uniform. Federal law sets a one-party consent baseline for intercepting an oral communication under 18 U.S.C. § 2511: if one party to the conversation consents, the recording is generally lawful. A substantial number of states go further and require all parties to consent for at least some recordings, including California, Florida, Illinois, Massachusetts, Pennsylvania and Washington among others. State wiretap statutes change, so anyone buying or designing one of these devices should check the current law in their own state, and should talk to an attorney about compliance in jurisdictions where they plan to sell or use the device. This is a genuine product-design constraint, not a footnote: several makers in this category already build an explicit consent-request step into the device for exactly this reason.
Frequently asked questions
What is the biggest difference between AI wearables that succeeded and ones that failed?
Scope. The devices still shipping do one job, usually recording and summarizing conversations. The devices that struggled hardest were pitched as broader phone or assistant replacements.
How long does an AI wearable’s battery actually last?
It depends heavily on the device and the mode. Published figures in this category range from about two and a half hours of heavy active use to roughly 100 hours, so check the manufacturer’s specific figures rather than assuming a category-wide number.
Do all AI wearables need a monthly subscription?
Most need an app, and the ones offering real-time AI summarization typically gate the heaviest usage behind a paid tier, though several offer a free tier with limits on history length or monthly AI-feature hours.
Is it legal to record conversations with an AI wearable?
It depends on your state and who you are recording. Several US states require consent from everyone being recorded, not just the wearer. This is not legal advice; check your state’s current law and consult an attorney if you plan to use or sell one of these devices.
What happens to an AI wearable if the company behind it shuts down?
It depends on the company’s wind-down plan. Humane’s shutdown in 2025 included a limited refund window and a hard cutoff date for server access and data deletion, after which the device’s cloud features stopped working.
Is a single-purpose AI wearable a better bet for a new hardware company to build?
This category’s track record so far suggests a tightly scoped single job is easier to deliver well on wearable-grade battery and hardware than a broad assistant promise, though the right scope always depends on the specific product and users.
Where Inventornest fits
An AI wearable lives or dies on decisions most buyers never see: how much battery the enclosure can actually hold, how much of the AI workload has to run in the cloud, and how the device behaves the day its subscription or its company goes away. At Inventornest, we treat battery performance as a core requirement from the first stage of design, not a late check, and we advise against adding connectivity that does not solve a real problem for the wearer. We work with founders on this kind of connected hardware through IoT product design services, and examples of finished devices are in our portfolio. If you are designing a connected wearable and want that kind of scrutiny applied early, the next step is to book a consultation.
