Border control at an international airport works on a simple premise. A camera looks at your face, matches it against a profile, and decides what you’re allowed to do next. Petlibro has now built roughly that same system and pointed it at a food bowl, because the biggest unsolved problem in a three-cat household is that the fat one eats the prescription diet meant for the skinny one.
That is the actual pitch behind Petlibro’s new AI-powered feeder, and I want to be precise about what it does and does not do, because the gap between those two things is where most smart pet gear falls apart.
What the feeder actually does
Three capabilities are confirmed. The feeder uses camera recognition to identify individual pets and track them separately. It offers dual feeding modes. And it sends alerts when something needs maintenance.
The recognition piece is the interesting one. CNET’s John Carlsen tested a version that uses an AI camera to work out which pet is eating and automatically log that meal against the correct profile. No collar tags, no RFID chips, no microchip reader. Just a camera making a judgment call about which face is in the bowl.
If that works consistently, it solves a genuinely annoying problem. Anyone tracking food intake across multiple cats currently does it by standing in the kitchen and watching, or by guessing. Automated attribution turns guesswork into a log.
Where I’d slow down
Petlibro describes dual feeding modes. A competing review of PETKIT hardware points out that PETKIT’s dual-hopper design puts two separate food compartments in one unit, so two different formulas can be dispensed, and describes that capability as absent from Petlibro’s current lineup.
Those two claims can coexist. Feeding modes and food compartments are not the same thing. Modes typically means something like scheduled versus on-demand dispensing. Hoppers means physically separate food. For a household where one cat is on a renal diet and another is on kitten food, the hopper question is the only one that matters.
So the honest read is this. The feeder appears to recognize which cat is eating and record it correctly. Whether it can dispense two genuinely different foods from the same unit is not established by anything I can verify. If you’re buying specifically to separate diets rather than to separate data, that’s the spec to confirm before you pay.
Recognition accuracy is the whole product
Camera-based identification lives or dies on edge cases, and cats are an unusually hostile test set. Two grey tabbies from the same litter. A cat that approaches the bowl at an angle, in the dark, or with its head mostly inside the housing. A cat that gains two pounds over a year and changes shape.
No accuracy figure has been published that I can point to. Until one exists, treat the tracking log as a strong signal rather than a medical record. Directionally useful, not evidentially airtight.
The vet assistant framing needs pushback
One 2026 buying guide describes the modern feeder as a 24/7 virtual veterinary assistant and suggests the data collected becomes the most valuable thing you bring to your annual vet visit. I’d hold that at arm’s length.
Intake data is legitimately useful. A cat that stops eating is one of the earliest signals that something is wrong, and a feeder that notices before you do has earned its shelf space. But a stream of gram counts is not diagnosis, and no vet is going to change a treatment plan because an appliance has opinions. The device produces one input. Framing it as an assistant oversells the machine and undersells the vet.
The unglamorous feature I like most
Maintenance alerts. Nobody writes headlines about them, and they are the difference between a working feeder and a plastic box full of stale kibble.
Automatic feeders fail quietly. Food bridges over the dispensing chute, a hopper empties while you’re away, a mechanism jams. The cat is hungry and nothing in your house knows about it. A feeder that flags its own problems is doing the single most important job that a feeder has, which is not failing silently while you’re at work.
Who should buy this
Multi-cat homes that want per-pet intake data and don’t want to put hardware on their cats’ necks. That’s the clean use case, and the recognition camera is a real answer to it.
Multi-cat homes that need two physically different foods dispensed from one unit should verify the hopper configuration first, because the marketing language around feeding modes does not settle that question.
Single-cat homes get very little from the identification layer and should buy on reliability and alerting instead. Paying for face recognition when there is only one face is a tax on novelty.
The broader pattern here is worth watching. Recognition models are cheap enough now to put inside a $100 kitchen appliance, and that is where this kind of AI stops being a demo and starts being plumbing. A camera that knows which cat is eating is a modest achievement. It is also exactly the kind of modest achievement that actually ships.
🕒 Published: