Ring is adding AI search to its home camera app, giving subscribers a faster way to find moments buried in recorded video. The feature, called Smart Video Search, is meant to answer everyday questions such as whether a package was picked up, without forcing users to swipe through long timelines.
The early picture is useful but uneven. In WIRED’s tests, the tool surfaced many relevant clips, yet it also missed some obvious cases, blocked certain searches and confused objects or actions in ways that show the limits of AI video understanding.
What Smart Video Search Does
Smart Video Search is rolling out by November 5 to all Ring Home Pro subscribers. That plan costs $20 monthly or $200 annually. Users on the plan can opt into the feature and choose which cameras are included, and Ring says it works with all versions.
The basic idea is simple: type a phrase into Ring’s app and get matching video clips. Liz Hamren, the CEO of Amazon’s Ring camera business, described using a search like “package today” to find whether her husband picked up a recurring delivery of frozen goods at home.
That is the kind of task Ring wants to make easier. The company has built a large base of video doorbells and other security devices, leaving many customers with extensive footage and limited ways to search it. The source article notes that users may have up to six month’s worth of recorded video to sort through.
Ring’s search can filter results by relevance or time. It can recognize broad categories such as packages, people, vehicles, animals, weather and actions such as running. It can also pick out some vehicle types, including “police” or “minivan,” but it is not built to distinguish a Kia from a Mazda or to read a specific license plate.
“We're not currently focused on license plate reading,” Hamren says.
Useful Results, With Clear Gaps
WIRED tested Smart Video Search over the past few days before the source article was published. Some searches worked well enough to show why people may want the feature. Queries such as “dog pooping,” “gardeners,” and “rollerblading” produced mostly accurate results.
Other tested searches also performed decently, including “backpack,” “scarf,” “speeding cars,” “ice cream truck,” and “shopping cart.” Hamren said she was impressed when “volleyball” found someone wearing a generic volleyball uniform.
But the feature did not reliably answer every household question. WIRED said it could not help with the recurring question, “Did you remember to lock the door?” It also could not spot the name of a football team on a shirt, because the feature does not directly recognize text.
The errors were sometimes understandable. Headlights from turning cars were interpreted as “fireworks.” “Blonde women” returned people matching the description, but also golden retrievers. A hand near a face could make the system think someone was “smoking.”
Some misses were stranger. “Purple hair” returned someone wearing a purple sweater and a bus with purple paint. Ring did not identify a decorative pumpkin, but connected the gourd with an orange-vested landscaper bundling leaves into a white sack. “Rollerblades” worked, while “roller blades” did not, because Ring associated “blade” with weaponry.
Blocked Searches And Privacy Limits
Ring is placing boundaries around what people can search. Hamren, who also oversees three other Amazon home technology units, said the company blocks searches it considers “offensive, inappropriate, or harmful.” She declined to provide a detailed list, but said names of weapons and “qualitative terms about people or situations” are among the barred terms.
WIRED’s searches for “Black,” “turban,” and “gun” returned no results, even though matching footage existed. “Disabled” returned people scooting by in wheelchairs and mothers pushing strollers. Ring said it will now block searches for “disabled” because of the risk of misuse.
Weapon-related searches showed the tension in those limits. Attempts to find someone swinging a wooden baseball bat or carrying a hammer failed. Results for “Hammer” included a leaf blower and hedge trimmer. “Water gun” found two instances of people holding water guns, while “gun” did not return results.
There is also a privacy limitation. Smart Video Search does not work when users have enabled end-to-end encryption, which the source describes as a higher form of privacy for videos. Hamren said Ring is working to remove that limitation, but did not give a timeline.
Ring says searches are not logged in users’ accounts. However, anonymized search terms may be used to improve the feature.
Why Ring Is Making This Shift
Smart Video Search has been in development for well over a year, according to Hamren. Ring says the system was trained with contrastive learning, using public videos and videos shared by employees and their friends.
The feature is part of a broader change in how Ring presents itself. Under Hamren, who joined in March 2023 after roles at Microsoft and Meta, Ring has moved away from “Tough-on-crime” advertising showing suspicious characters. She also pulled back on partnerships with police that had enabled tools for users to share videos with law enforcement agencies without a warrant.
Ring’s current marketing emphasizes pets, children and lighter everyday moments. The company is framing its cameras less around curbing crime and more around reducing anxiety by helping customers understand what happened in front of a camera.
Search may only be the first step. Hamren said the AI models behind it could later notify users about specific situations, such as a bear in the backyard or anomalous puddling on the patio. Ring also expects to provide summaries of recent events.
“Providing you intelligence and context about what's happening is ultimately where we want to go,” she says.
Ring is also looking beyond homes. In the first half of 2025, it plans to launch a subscription plan for small and medium-size businesses. Hamren said subscribers would be able to oversee multiple locations and give staff access to certain footage.
For now, Smart Video Search is a practical upgrade with visible limits. It can make Ring footage easier to use, especially for ordinary questions about packages, pets, vehicles or visitors. But its blocked terms, missed matches and odd false positives show that AI search in home surveillance remains a tool that needs careful boundaries and continued testing.