Google Pet Memory has a practical promise: make a smart home understand not just that an animal appeared on camera, but which animal it saw. For anyone using Nest cameras to monitor pets, that could turn a stream of generic alerts into something far more useful.
In testing by Jennifer Pattison Tuohy at The Verge, however, the feature did not deliver that level of precision. After two weeks, Gemini for Home still could not reliably distinguish three cats, which weakened the case for pet-specific smart notifications and automations.
What Pet Memory Is Supposed To Do
Pet Memory is part of Gemini for Home and is designed to teach Google Home who a user’s pets are. The goal is to move beyond a Nest camera saying it saw an animal and toward a system that can identify a named pet.
That matters because smart home cameras already produce a large number of notifications. For pet owners, the useful alert is often not simply that something moved, but whether the animal at the door, feeder, or room is the right pet.
The source article describes three intended uses for the feature:
- Identify which cat wanted to come inside.
- Reduce false alerts caused by a neighbor’s cat or a chicken at the door.
- Support personalized feeding automations for each cat.
The household used for the test included a flock of free-ranging chickens, a Wirehaired Pointing Griffon named Gus, two indoor/outdoor cats, and a new kitten. The cats were the main test case because the feature’s value depended on telling them apart.
Why The Test Exposed A Bigger Smart Home Problem
Security cameras have been getting better at describing what they see. The article notes that many cameras once focused mostly on motion alerts, while newer systems use machine learning and generative AI to provide more descriptive notifications.
Nest had already moved in this direction. The Nest Cam IQ, launched in 2017, could distinguish people from pets, and Google later added animal detection to Nest cameras. More recently, Google added AI text descriptions that explain what a camera sees.
Other smart home platforms have followed a similar path. Amazon’s Ring, Wyze, and Apple Home through HomeKit Secure Video have also added AI-powered descriptions, including animal types in some alerts. The important distinction in this case is personalization: Google Home is presented as the first to make those animal alerts specific to a named pet.
That is a meaningful step if it works. A generic animal alert can be useful, but a named-pet alert can support decisions and automations. It can help answer whether Smokey is inside, whether Boone is near the feeder, or whether the kitten Osa is the animal moving through a room.
The Limits Were Clear Early
Pet Memory comes with restrictions. It requires the top-tier Google Home Advanced Plan at $20 a month. Although it is part of Google Home, the source article says it currently only works on Nest cameras.
It also only works on indoor Nest Cams. That limitation immediately ruled out one of the more appealing ideas in the test: using it as an upgraded digital pet door. If the relevant camera is outside, Pet Memory is not available for that use.
The tester instead placed indoor Nest Cams in strategic locations to see whether Google Home could track which cats were inside. One intended question was simple:
“Hey Google, when did you last see Smokey in the house?”
The answer was not reliable. Google Home kept treating the cats as Smokey, the first cat entered into the system. When asked when Smokey was last spotted, the Google Nest Hub showed video of the new kitten playing in a daughter’s room.
Naming A Pet Was Not Enough
The article’s most important finding is that Pet Memory did not appear to create a strong individual visual identity for each cat. The tester found no training process, no photo upload option, and no way to correct mistakes.
The setup process allowed a name and breed. When the tester tried to add more descriptive information, the system did not use it. The details were specific: Boone is a tuxedo cat with white paws, Osa the kitten is a tabby cat, and Smokey is a large gray-and-white cat.
That extra information did not solve the problem. Pet Memory said it could not process the added detail and continued labeling every cat as Smokey.
The Verge asked Google why the system did not seem to be working as intended, but the company had not responded with details according to the source article. Without that explanation, the test leaves a practical conclusion: the feature may know that pets exist, but in this case it did not reliably know which pet was which.
Automations Need Accuracy To Be Useful
The feeding test showed why pet recognition has to be dependable before it can become a serious automation trigger. The tester wanted to make an Aqara Pet Feeder smarter by using a Nest Cam to identify a specific cat and then run the right feeding scene.
There was already a prior setup using Aqara’s G3 camera when it saw an animal, but that camera could not identify species or individual pets. With a Nest Cam and Pet Memory, the goal was more precise: dispense food for a specific cat without relying on RF tags or other proprietary options. The cats did not wear collars.
This also tested two newer Google Home capabilities: camera detection events as automation triggers and Help Me Create, which can generate an automation from a description. The tester typed this request into the Ask Home chatbot in the app:
“Run my Aqara ‘Feed Smokey’ scene whenever the Nest Cam in the laundry room sees Smokey,”
Google Home created that automation, and the tester then made a separate one for Boone. The core issue remained unchanged. Because Pet Memory could not tell Boone from Smokey, the feeder automation ran when either cat approached.
That is the central lesson from the test. AI-powered smart home features can sound transformative when they move from detecting categories to recognizing individuals. But when recognition is unreliable, the failure is not cosmetic. It directly affects alerts, safety checks, and connected-device behavior.
For Google Pet Memory, the promise is clear: named-pet awareness inside Google Home could make Nest cameras much more useful for pet care. The current test result is equally clear: for three cats named Boone, Osa, and Smokey, the system was not accurate enough to make pet-focused automation dependable.