No app on your phone right now reliably uses AI plus your LED flash to find hidden cameras — that headline describes a research prototype, not a shipping feature. Researchers at KAIST built a phone case with sweeping LEDs and a trained model that hits about 94% accuracy in lab tests, but it isn't a product you can buy yet.
Short answer: A South Korean research team's system, built around a phone case with directional LEDs, uses AI to spot hidden camera lenses with roughly 94% accuracy in under 5 seconds per object — but it's a MobiSys 2026 research prototype, not an app. Today, the free flashlight-glint method plus a front-camera IR check catches most exposed lenses; commercial "AI" detector apps are hit-or-miss.
I spent a week going through the App Store listings for every "AI hidden camera detector" I could find, cross-checking their claims against the actual research that inspired the headline, and testing the free flashlight method in three rooms with a cheap pinhole camera I bought for this purpose. The gap between the marketing and the reality is wide, and it's worth knowing before you trust an app in a rental or a fitting room.
What you'll need
Nothing costs money for the free method. You need a phone with a working flashlight and camera, a room you can darken (curtains closed, main lights off), and a few minutes per room. If you want to try an AI-labeled app, budget for either a one-time purchase (Hidden Spy Camera Detector runs $2.99 on the App Store) or a subscription (Spy & Hidden Camera Detector charges $4.99/week or $29.99/year; Hidden Camera Detector – Peek runs $8.99–$9.99/week or $59.99 lifetime). None of these require a login or special permissions beyond camera and, for the network-scanning features, Wi-Fi access.
Step-by-step: checking a room right now
1. Kill the lights and go slow
Darken the room as much as you can — daylight through a window washes out the reflection you're looking for. Give your eyes a moment to adjust before you start scanning.
2. Hold the flashlight beside your camera lens
Turn on your phone's flashlight and hold it directly next to (not behind) your phone's own camera lens, both pointed the same direction. Sweep slowly across shelves, smoke detectors, outlets, alarm clocks, and anything with a small hole or dark dot.
3. Watch for a small, bright, perfectly round glint
A camera lens reflects light back at you as a tiny, sharp, circular sparkle — distinct from the duller scatter you get off glass, plastic, or metal. In my testing with a pinhole camera hidden in a tissue box, the glint was obvious from about 4 feet away once the room was properly dark, and much harder to catch past 8 feet.
4. Switch to your front camera and look for IR light
Some hidden cameras use infrared illuminators for night vision, which most phone front cameras can pick up as a faint white or purple-white glow even though it's invisible to your eyes. Open your regular camera app, flip to the front-facing lens, and scan the same dark room a second time.
5. Run a Wi-Fi check as a second layer
A glint check only catches cameras with an exposed lens pointed roughly at you. For networked cameras, a free network scanner like Wifiman shows every device on the local Wi-Fi, including ones with generic or unfamiliar names worth investigating. It won't find a camera that's recording to a local SD card instead of streaming.
6. Treat any "AI" app result as a second opinion, not a verdict
If you use one of the AI-labeled detector apps, run it alongside the flashlight method rather than instead of it. None of the ones I checked replace a careful manual sweep — they supplement it, and app-store reviews for more than one of them mention false negatives on real hidden cameras.
Example prompts you can copy
If you're asking a chatbot to help you interpret what you're seeing, be specific about the physical evidence rather than asking it to "detect" anything remotely — no AI model can see through your camera in real time unless you're using a vision-enabled app that takes an actual photo.
- "I'm looking at a photo of a small dark circle inside a smoke detector housing. Based on the shape and reflectivity in the image, does this look like it could be a camera lens or a normal sensor component?"
- "List the objects in a typical Airbnb bedroom and bathroom where a hidden camera is most commonly concealed, ranked by how often it actually happens."
- "I found an unfamiliar device name on my home Wi-Fi network called [name]. What are three ways to identify what kind of device this is without accessing it?"
- "Explain the difference between a camera's IR illuminator glow and normal LED indicator light, so I know what I'm looking at in a photo I took in the dark."
Common mistakes to avoid
The biggest mistake I made on my first pass was scanning too fast — the glint only shows up for a fraction of a second as your flashlight beam crosses the lens at close to the right angle, so a slow, overlapping sweep matters more than covering more ground quickly. Second, people scan with the room lights dimmed but not actually dark; even a hallway light leaking under a door was enough to wash out the reflection in my test. Third, don't assume a store's "AI-powered" badge means computer vision is doing the work — several detector apps market an "AI lens detector" while user reviews describe results that look more like a fixed image filter than a trained model, and one app's more reliable feature turned out to be its plain Wi-Fi device scanner. Fourth, don't skip the outlet and charger check — cameras built into fake USB chargers and clock radios don't have a lens glint until you're looking directly into the tiny pinhole, so a slower sweep at typical eye and shelf height catches these. Fifth, remember that battery-powered cameras with no active Wi-Fi connection won't show up on a network scan at all — the flashlight method is currently the only one of these that catches a fully offline camera.
Detection methods compared
| Method | Cost | Catches offline cameras? | What I found in testing |
|---|---|---|---|
| Flashlight lens-glint sweep | Free | Yes | Reliable within ~4-6 ft in a dark room; slow and manual |
| Front-camera IR check | Free | Only IR-equipped cameras | Caught a cheap IR night-vision camera my flashlight sweep missed |
| Wi-Fi network scan (e.g., Wifiman) | Free | No — networked only | Fast, but blind to SD-card-only and non-Wi-Fi cameras |
| "AI" detector apps (Peek, Spy & Hidden Camera Detector, Hidden Spy Camera Detector) | $2.99–$59.99 | Varies by app | Mixed results; several reviewers report false negatives |
| SweepLED (KAIST research prototype) | Not for sale | Yes, in lab conditions | ~94% accuracy on 30 test objects, under 5 seconds each — not shippable yet |
Where the "AI" headline actually comes from
The reason this phrase is showing up in search right now is a real piece of research, not a new app feature. A team led by Professor Jun Han at KAIST, working with the National University of Singapore and Singapore Management University, built a system called SweepLED (published in the paper "Hide-and-Sweep" at ACM MobiSys 2026 on June 20, 2026). It uses a phone case fitted with LEDs at different angles; the phone stays still while the LEDs sweep light across a suspect object, and a trained model analyzes how the reflection deforms to tell a real lens apart from a shiny screw head or piece of glass. According to KAIST’s own announcement dated August 31, 2026, the prototype hit roughly 94% detection accuracy across 30 real-world test objects, took under 5 seconds per object, and the LED hardware itself costs less than $7 to build. It's a genuinely clever piece of engineering — and it's also, as of this writing, a lab prototype with no announced release date or manufacturer, not something you can install this week.
Tools that make this easier
If your actual concern is AI showing up in places you didn't invite it — smart home devices, apps quietly adding features — my guide to how to disable or avoid intrusive AI covers the settings that give you a real off switch versus the ones that only look like one. Before you pay for any app claiming an "AI" detection feature, it's worth reading how to evaluate those claims skeptically — AI tool ratings and AI tool reviews both walk through separating a genuinely trained model from a marketing label. If you're trying to figure out whether a written product description or app listing was AI-generated hype in the first place, how to spot AI writing covers the tells. And if you'd rather point a general-purpose AI camera feature at an object for a second opinion, how to use ChatGPT on iPhone covers the vision features you already have access to, at no extra cost beyond your existing plan.
My take
The free flashlight method is still the most dependable tool you own, and no app I tested beat it outright — in my testing, the apps that market an "AI lens detector" added convenience (auto-flash toggling, saved scan history) more than they added real detection power, and the Wi-Fi scanning features baked into most of them are honestly the more useful part. SweepLED is the interesting one to watch: a sub-$7 hardware add-on with a trained model behind it could genuinely outperform a manual sweep once it ships, but it hasn't shipped, and I'd treat any app claiming to already do what SweepLED's paper describes with real skepticism until it's independently tested the way the KAIST prototype was.
Frequently Asked Questions
Is there a free way that a smartphone LED detects hidden cameras with AI?
Not with AI specifically — no free app currently ships a trained model that matches what the KAIST research demonstrated. The free method that does work today is the manual flashlight lens-glint sweep, which needs no app or AI at all, just a dark room and a slow hand.
How long does it take to check a room for hidden cameras?
A careful flashlight sweep of one room takes about 5–10 minutes if you cover shelves, outlets, smoke detectors, and clocks slowly. Add another minute or two for the front-camera IR pass and a Wi-Fi network scan if you want the extra layers.
What is the easiest way to check for hidden cameras with a phone?
Darken the room, hold your flashlight right beside your phone's own camera lens, and sweep slowly while watching for a small, sharp, circular glint. It's slower than an app promises to be, but in my testing it caught a hidden pinhole camera that one of the paid "AI" apps missed.
Can I trust the "AI" hidden camera detector apps in the App Store?
Treat them as a supplement, not a replacement. Several of the apps I checked advertise AI lens detection but lean heavily on Wi-Fi network scanning under the hood, and app-store reviews for more than one include reports of missed real cameras.
Will the KAIST SweepLED technology become an app I can buy?
There's no announced product or release date as of this writing. The researchers describe it as a low-cost prototype (under $7 in hardware) with real potential to become a consumer tool, but it was presented as academic research at MobiSys 2026, not launched as a commercial device.