Alternatives & comparisons
SMOWL Alternative for the AI Cheating Era (2026)
SMOWL is a lightweight browser-extension proctoring tool, popular with European universities, built on webcam monitoring and keystroke/mouse biometrics. It does behavioral identity checks well. But Cluely isn't visible to any of it. Here is what sits below the observation layer.
Other exam-security alternatives to consider
No single tool covers every layer. Here is how the rest of the lockdown-browser and proctoring-service field compares on the same AI-cheating threat model.
What SMOWL Does
SMOWL is a browser-extension remote proctoring tool widely used by European universities. Rather than a full lockdown browser or live human proctor, it runs as an extension inside the candidate's existing browser, capturing periodic webcam and audio snapshots and building a behavioral profile from keystroke and mouse patterns over the course of the exam. That profile is compared against a baseline captured at enrollment to flag identity mismatches or unusual behavior for later review. It works well for the things a webcam snapshot and a typing pattern can reveal: confirming the enrolled candidate is likely still the one typing, an obvious camera violation, a login from an unexpected pattern.
In short: SMOWL is built to confirm who is likely typing based on behavioral patterns and periodic webcam checks, not what else is running on the machine underneath the browser.
What SMOWL Cannot See
A browser extension only ever sees what happens inside the browser tab it is installed in, plus whatever a periodic webcam snapshot catches. An invisible AI overlay like Cluely runs at the OS layer, entirely outside the browser process, and marks itself excluded from the screen-capture APIs a webcam snapshot or extension-level check would rely on. Keystroke and mouse biometrics add a behavioral signal, but a candidate simply reading an answer off an overlay before typing it themselves produces a normal-looking typing pattern: there is no keystroke signature for words the candidate never had to compose.
The same gap applies to on-device LLMs. A model running in local memory produces no browser-extension artifact and no network traffic to intercept. Aiseptor detects the technique instead of the appearance: a screen-capture-exclusion flag, GPU VRAM deltas, and DNS/SNI to AI endpoints, regardless of process name.
Behavioral biometrics and browser-level checks solve a real, separate problem: confirming who is likely at the keyboard. Neither one reaches the OS layer where these two techniques actually run.
Comparison: SMOWL vs. Network-Layer Enforcement
| Capability | SMOWL | Aiseptor (Network Layer) |
|---|---|---|
| Webcam / audio monitoring | Yes | No |
| Keystroke / mouse biometrics | Yes | No |
| Invisible AI overlays (Cluely, Pluely, unknown forks) | No | Yes |
| On-device LLMs (Ollama, LM Studio) | No | Yes |
| OS-wide network / DNS enforcement | No | Yes |
| Data collected | Webcam, audio, keystroke/mouse patterns | Network access signals only |
Complementary Architecture
SMOWL = browser-extension, webcam, and behavioral-biometrics layer. Aiseptor = network and device layer. Aiseptor sits beneath SMOWL, not in place of it. SMOWL's keystroke and mouse biometrics catch fraud Aiseptor does not address: a substitute test-taker, an obvious camera violation. Aiseptor catches the device and network AI surface SMOWL's browser extension cannot reach: the overlay running at the OS layer, the local model in memory. We don't claim to replace identity verification or behavioral monitoring.
See also: Respondus alternatives · Safe Exam Browser alternatives
Frequently Asked Questions
Does SMOWL detect Cluely?
SMOWL runs as a browser extension, watching the candidate through a webcam and building a behavioral profile from keystroke and mouse patterns. Cluely is an OS-layer overlay that excludes itself from the same screen-capture APIs a browser extension would read. A candidate can have an overlay open on screen while SMOWL's extension sees nothing unusual in the browser window it monitors.
Does SMOWL detect AI cheating generally?
SMOWL's webcam checks and keystroke/mouse biometrics can flag some AI-assisted behaviors that change typing rhythm or attention: pasted text, unusual pauses, looking away from the screen. They cannot see an OS-layer overlay that hides from screen capture, or an on-device language model running in local memory with no network traffic and no browser footprint.
Why can't SMOWL's biometric monitoring catch AI overlay use?
Keystroke and mouse biometrics are built to catch identity fraud (a different person typing) and gross behavioral anomalies, not to inspect what else is running on the machine underneath the browser. An invisible overlay that only surfaces answers for the candidate to read produces no keystroke signature at all.
Is there a SMOWL alternative without webcam?
Yes. Aiseptor enforces exam integrity at the network and device layer: no webcam, no keystroke biometrics, no browser extension. The audit record is metadata about what was and wasn't reachable during the session, not a recording or behavioral profile of the candidate. Full data model on the trust page.
Where Aiseptor Fits: Beneath, Not Instead Of
Aiseptor is a layer, not a rip-and-replace. It sits beneath SMOWL, and beneath any lockdown browser or proctoring service, owning the device and network layer those tools architecturally cannot reach. Keep what you have for browser control or webcam proctoring; add Aiseptor for the OS- and network-level AI threats it was built to stop. The exam page can open in a normal browser while Aiseptor enforces the machine boundary.
What Aiseptor does not do: physical and environment security is out of scope. Aiseptor does not verify identity, watch the room, or catch a phone, a paper note, or an in-person accomplice off-camera. For those, pair Aiseptor with a live proctor or an identity-verification step. Aiseptor secures the device and the network path, not the physical room around it.
Typing pattern verified. The device, still unwatched.
Aiseptor pairs with SMOWL's behavioral and webcam layer to close the network and device gap underneath. The audit record is metadata about what was and wasn't reachable, not a recording of your candidates.