Modern UniFi cameras can do far more than record motion, but the useful question is not “does it have AI?” It is which camera can reliably produce the detection, search result, alert, or access-control action you actually need.
AI security cameras are getting marketed with increasingly broad claims. In a real installation, those claims need to be translated into something practical: Can the camera distinguish a person from a vehicle? Can it read a plate at the driveway? Can it recognize a face? Can it search for a red truck? Can a detection trigger a light, move a PTZ camera, or open an authorized gate?
UniFi Protect can now do many of those things, but capabilities vary significantly by camera generation and by whether an AI Key or AI Port is part of the system. We verified the feature details below against Ubiquiti's current documentation before publishing this guide.
Current G4, G5, G6 and AI-series Protect cameras support person and vehicle smart detections, with the exact intelligence depending on model. Older G3 cameras and supported third-party ONVIF cameras can gain smart detections through an AI Port.
This matters because “motion” and “person detected” are not the same thing. Motion can be a branch, snow, headlights, an insect near the lens or changing shadows. Object-aware detections let Protect create alerts and searchable events around the things people generally care about.
For Northern Colorado homes and businesses, that can mean notifying on a person approaching a side entrance while ignoring routine movement elsewhere, or identifying vehicles entering a commercial lot without treating every environmental change as an alarm.
Ubiquiti's current feature matrix lists face recognition on G6 and AI-series cameras, excluding 360-degree models. G4 and G5 cameras can gain face recognition through AI Port; AI Key can also add face recognition to G4/G5 footage, but its processing can be asynchronous depending on queue length. G3 and compatible ONVIF cameras require AI Port for this class of recognition.
That distinction is important. If you need near-real-time recognition, choose the camera and AI hardware around that requirement rather than assuming every Protect camera behaves identically.
Recognition also should not be treated as a substitute for good physical security. Camera angle, lighting, distance and image quality still determine whether a face is usable.
License plate recognition (LPR) is available natively on supported G6 and AI-series cameras, again with model exceptions, and can be added to older camera generations with AI hardware. AI Port provides real-time face and license-plate recognition for supported G4/G5 cameras and can AI-enable compatible G3 and ONVIF cameras. AI Key can add LPR to G4/G5 cameras, but results may be delayed by its processing queue.
The specialized camera still has to be installed correctly. Plate angle, distance, vehicle speed, headlight glare, night exposure and the size of the plate in the frame matter. A wide overview camera mounted high on a building may be excellent for situational awareness but poor for reliably capturing plates.
That is why we normally treat license plate cameras as a design problem, not a checkbox. If plate identification is important, dedicate the right view to it.
Protect supports line-crossing detections on supported fixed cameras. Ubiquiti's current matrix specifically notes that line crossing is not available on PTZ models. Line-cross counting is also model-dependent: supported G6/AI cameras can count crossings, while G4/G5 models may require AI Port for counting.
A virtual line can be useful at a driveway, gate, hallway, loading area or other controlled path where crossing a defined boundary matters more than generic motion inside a large zone.
Microphone-equipped supported cameras can classify audio events such as speech and glass break. That limitation sounds obvious, but it is worth stating because not every camera or deployment should be expected to provide audio intelligence.
AI Key can add speech-to-text for supported detections. Ubiquiti currently says speech transcription is only available with Protect AI-series cameras that support speech smart detections, and AI Key currently supports English for speech transcription and natural-language search.
Audio recording and recognition also raise privacy and legal considerations. We recommend deciding intentionally whether microphones should be enabled for each location rather than turning them on everywhere by default.
Pan-tilt-zoom cameras are useful when an operator needs to look around a large area, zoom into an event, patrol presets or use Protect automations to reposition a camera. Warehouses, large parking lots, yards, campuses and wide commercial properties are common examples.
Protect's Alarm Manager can respond to detections with actions that include adjusting PTZ cameras. That creates useful workflows—for example, a detection on one camera can help direct attention to another view.
But a PTZ camera is looking in one direction at a time. If it is zoomed toward a gate, it is not simultaneously providing the same detail behind it. We usually use fixed cameras for guaranteed coverage and PTZ as an additional investigative tool. Also remember the current limitation above: Ubiquiti does not list line crossing on PTZ models.
UniFi Protect's Alarm Manager is where detections become more than timeline markers. Ubiquiti documents automations that can trigger notifications, webhooks, floodlights, PTZ adjustments and even UniFi Access actions based on supported events.
That opens practical possibilities: alert only when a person enters a defined area after hours; turn on lighting when a relevant detection occurs; reposition a PTZ; or create a workflow around an entry event. The goal should be fewer meaningful alerts, not more notifications.
AI Key is not simply “more AI.” It analyzes smart detections produced by compatible Protect cameras and adds richer computer-vision functions. Ubiquiti currently lists NeXT AI Natural Language Search, AI-enhanced alarms, advanced AI summaries, face enhancement, speech-to-text, and added face/LPR processing for G4/G5 cameras.
Natural-language search is especially interesting for investigation. Instead of filtering only by a predefined object category, Ubiquiti says searches can use concise descriptions such as a delivery truck or a person walking a dog. It works best for descriptions involving people and vehicles.
There are real limits. AI Key processes up to 1,000 smart detections per hour. Ubiquiti says a full queue can delay search processing by up to about 15 minutes. It can pair with an unlimited number of cameras, but high-activity environments may need additional AI Keys. Only detections that the camera first recognizes as smart detections are available for AI Key processing.
AI Key can also create up to five predefined AI search queries for AI-enhanced alarms. That is useful, but it should be designed around specific operational needs rather than treated as magic surveillance.
AI Port serves a different role. It can add Protect intelligence to compatible older UniFi cameras and third-party ONVIF cameras. Ubiquiti currently lists motion, people, vehicles, animals and speech detections; vehicle color and type classification; and face and license-plate recognition.
For G4/G5 PTZ cameras, AI Port can add face and license-plate recognition while leaving the camera's existing pan, tilt and zoom functions intact. Ubiquiti specifically notes that AI Port does not improve PTZ control or tracking. AI Port also is not compatible with Protect AI-series cameras because those models already provide their own AI capabilities.
AI Port is also the bridge required when you want AI Key's richer analysis applied to G3 or compatible ONVIF camera streams.
One of the more useful UniFi integrations is between Protect and UniFi Access. Cameras can be paired with doors or gates so operators can see live video and review recordings tied to unlock events.
On supported deployments, license plates can become Access credentials. Ubiquiti currently supports License Plate Unlock with specified G6 and AI cameras paired directly to compatible Access hubs. An authorized vehicle enters the configured detection zone, Protect recognizes the assigned plate, and Access can unlock the gate or garage door.
This is not universal across every Access product. Ubiquiti specifically says the Retrofit Hub does not currently support License Plate Unlock. Gate deployments also require proper safety design: Ubiquiti requires appropriate gate-position sensing and UL 325-compliant entrapment protection for License Plate Unlock configurations.
For a residence, gated shop, HOA, warehouse yard or commercial property, this can eliminate a separate vehicle credential system—but only when the camera view, gate operator and Access hardware are designed together.
Start with the outcome, not the model number. A useful design might use different cameras for different jobs:
More AI does not fix a poor camera angle. A specialized LPR camera does not fix a plate that occupies too few pixels. A PTZ does not replace cameras that continuously watch critical areas. And an AI Key does not analyze an event the camera never recognized as a smart detection.
Berthoud WiFi designs and installs wired security camera systems and UniFi Protect systems for homes, businesses, shops, warehouses and rural properties across Berthoud, Loveland, Fort Collins, Longmont and Northern Colorado.
If you are comparing camera layouts, read the earlier articles in this security series: why UniFi cameras are a strong security choice, security camera installation cost in Northern Colorado, and how to pre-wire a new home for security cameras.
For current feature details, we relied on Ubiquiti's official documentation for Protect AI detections, AI Key, AI Port, and License Plate Unlock. Features can change with Protect, Access and device firmware, so confirm requirements before buying hardware.
Tell us whether the goal is identification, license plates, a gate, after-hours alerts, PTZ coverage or simply dependable recording. We will design the camera views and UniFi hardware around the outcome.
Share the property, the problem areas, and what you need the network to support. We will review the details and follow up with practical next steps.
