Google’s Pet Memory Can’t Recognize My Cats

One of the standout features of my smart home setup is its ability to help me keep an eye on my pets, and security cameras play a crucial role in monitoring my furry friends. However, the flood of notifications can sometimes drown out the important ones. So, when I heard about the new Pet Memory feature for Gemini for Home, I thought it might be just what I needed.

Pet Memory aims to teach your Google Home system to recognize your pets, allowing connected Nest cameras to identify not just that an animal is present, but which specific animal it is. This could help tailor your smart home experience to your pets’ needs.

I envisioned using it to monitor my three cats while I worked in my home office, distinguishing between them and automating their feeding based on who approached the feeder. However, after two weeks of testing, Google’s AI still struggles to differentiate my three cats.

My home sits on a spacious suburban lot, and I’ve taken full advantage of the space to care for a variety of animals (my husband has already vetoed the idea of a goat or horse). Currently, I have a flock of free-ranging chickens, a large Wirehaired Pointing Griffon named Gus, two indoor/outdoor cats, and the newest addition, a playful kitten.

As someone who frequently tests smart home tech, I’ve found that security cameras are best used for monitoring my pets. They help keep the chickens safe from predators and my lunch safe from Gus, while also serving as a digital pet door for the cats. When I receive a notification about an animal at the door from my video doorbell or outdoor camera, it usually means one of the cats wants to come inside.

While I only have one dog and don’t need to identify the chickens individually, I thought Pet Memory could help manage my three cats in three key ways: first, by identifying which cat wanted in and reducing false alerts (like when my neighbor’s cat is lounging on my porch or a chicken is peeking in); second, by confirming that all my pets were safely indoors before nightfall; and third, by enabling personalized automated feeding for each cat.

The ability to recognize individual pets represents a significant advancement in smart security camera technology. While some dedicated pet cameras offer facial recognition, most standard security cameras have only recently begun to provide alerts for motion. Thanks to advancements in machine learning and generative AI, these cameras can now deliver more detailed notifications. Instead of just a generic “motion alert,” they can specify exactly what they detected, eliminating the need to open the app and wait for a clip to load.

Nest has been a leader in this area, particularly with its Nest Cam IQ, which launched in 2017 and could differentiate between people and pets. Google later expanded animal detection capabilities to Nest cameras, and more recently, it introduced AI-generated text descriptions of what the camera sees.

Other brands like Ring, Wyze, and Apple Home have also started incorporating AI-powered descriptions, often identifying the type of animal detected (though not always accurately). However, Google Home is the first to offer personalized recognition.

There are, however, some limitations and costs involved. Pet Memory requires the premium Google Home Advanced Plan at $20 a month, and during my testing, it only functioned with Nest cameras. Additionally, it only works with indoor cameras, which limited its usefulness for my idea of a smarter digital pet door.

I set up several indoor Nest Cams in strategic locations, hoping to use Pet Memory to easily check which cats were inside and ensure everyone was home before dark. I asked, “Hey Google, when did you last see Smokey in the house?”

Despite providing Google Home with descriptions of my three cats, it insisted that they were all Smokey, the first cat I introduced. When I asked the Google Nest Hub about Smokey’s last sighting, it pulled up an adorable video of the new kitten playing in my daughter’s room.

From my experience, Pet Memory doesn’t seem to create individual visual profiles for each pet. There’s no training process, no option to upload photos, and no way to correct errors. You simply enter a name and breed, and it attempts to figure out the rest. I tried providing more details—Boone is a tuxedo cat with white paws, Osa the kitten is a tabby, and Smokey is a large gray-and-white cat—but it couldn’t process the extra information and continued to label every cat as Smokey.

I reached out to Google for clarification on why the system wasn’t functioning as expected, but I haven’t received any response.

Undeterred, I tried to leverage Pet Memory to enhance my Aqara Pet Feeder. I had previously set it up to activate when Aqara’s G3 camera detected an animal, as it couldn’t identify species or individuals.

With a Nest Cam, I wanted to see if I could configure it to dispense the right amount of food when it recognized a specific cat, helping to prevent overeating without needing RF tags or other proprietary solutions (my cats don’t wear collars).

This was also a chance to test Google’s new feature that allows camera detection events to trigger automations, along with its Help Me Create function, which can set up automations based on simple descriptions.

I typed my 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,” and it set it up for me. I created a separate automation for Boone, but I encountered several issues.

The biggest problem was that it still couldn’t distinguish Boone from Smokey, so when either cat approached the feeder, it triggered the automation. Additionally, the Google Home app doesn’t allow you to limit how many times an automation can run (unlike the Aqara automation). As a result, the automation activated every time either cat visited the bowl, leading to an overflowing food bowl by the end of the day.

I also had high hopes for using Google’s Home Brief feature to get a summary of my pets’ activities. This feature provides an overview of activity from connected devices at the end of the day. I requested more detailed pet activity information, but it mostly reported “there were many pet interactions,” which wasn’t particularly useful.

At one point, it indicated that Smokey was eating in the garage, which raised a red flag. Cats eating in the garage can lead to trouble, as there are many things they shouldn’t access. However, when I checked the relevant clip, it was actually Boone eating in the laundry room, where he should be.

I reviewed the garage camera footage, but neither cat had been in there at any point. Misidentification can be more problematic than having no information at all. While getting a cat’s name wrong can be amusing, relying on it for accurate location information can lead to serious issues.

Recognizing individual pets is evidently more complex than simply detecting that a pet is present. This feature hints at the future of AI-powered cameras, but it’s not something I would depend on right now. AI has promised to alleviate the tedious task of sifting through camera footage, but until it can reliably identify events, continuous recording remains more beneficial than unreliable AI summaries.

Fortunately, continuous recording is something Nest Cams excel at as part of that same $20 monthly subscription. Whether you’re monitoring an indoor kitten, a tropical aquarium, or a flock of chickens, being able to scroll through a day’s worth of footage is invaluable for figuring out what might have upset your cat’s stomach or how a chicken went missing (since hawks don’t always trigger motion alerts). Once AI can reliably answer those questions, it may be worth the investment; for now, Boone is enjoying the extra kibble.