SHL Proctoring in 2026: What It Records and Flags

SHL is one of the largest talent assessment vendors in enterprise hiring, and its proctoring is not a single fixed configuration. It is a menu. The employer that invited you chooses which monitoring features are switched on, and SHL's own documentation describes two quite different envelopes depending on that choice: a light one built from periodic webcam stills and event counters, and a heavier one that can include live audio-video monitoring, face verification, and detection of objects such as a mobile phone in frame.

This article reconciles what SHL's administrator guide, its candidate-facing data protection notice, and its release notes each say, as of September 2026. The practical answer is that you cannot infer what your assessment is doing from the SHL brand alone — but you can narrow it down considerably, and you can read a flag for what it actually is.

"An SHL assessment" is four different products

The invitation email rarely says which one you have. SHL's product catalog separates at least these:

• Verify cognitive and behavioral assessments — timed, question-based, usually taken alone in a browser.
• Smart Interview On Demand — a one-way recorded video interview. SHL describes candidates as recording "responses at their convenience" and says its AI "uses validated data sets to assess candidate competency for your role" and "standardizes scoring using pre-defined evaluation parameters."
• Smart Interview Live — a live video interview with a human on the other side.
• Smart Interview Live Coding — a live technical interview with a code editor.

Proctoring features attach mainly to the first category. A one-way video interview is already a recording of your face and voice by design, so "proctoring" in that context means something different from a monitored aptitude test. Establishing which product you have been sent is the first step in knowing what is being captured.

The five features in SHL's standard remote proctoring

SHL's TalentCentral Remote Proctoring Guide, written for the administrators who configure assessments, lists five selectable monitoring features:

Four of those five are counters, not recordings. They produce a number, and in the case of the image features, a set of stills. The guide describes administrators receiving "a combined proctoring report for each candidate" with flag counts, snapshots, and downloadable reports.

Three things that guide does not describe in this mode are worth naming explicitly, because candidates routinely assume all of them: continuous screen recording, keystroke logging, and audio capture. It also states meaningful limits — the features run on desktop and laptop only, not tablets or phones, and not every SHL assessment supports them.

The three event counters are conventional implementations of ordinary web APIs rather than anything proprietary to SHL. If you want the underlying mechanics — including why the browser-off-focus figure is a percentage of elapsed time rather than a count of tabs — see how interview platforms detect tab switching, which covers the same visibility events SHL is reading here.

What the candidate privacy notice adds

SHL's Candidate Online Assessment Data Protection Notice describes a considerably wider set of proctoring data than the administrator guide covers. It lists "live audio-video monitoring of the candidates; Intermittent or periodic capture of audio, still images, or audio-video recordings from your device and camera; multi and no face detection; suspicious objects detected in your image capture (i.e., mobile phone); face verification; and recording of your identification documentation."

That is the outer boundary of what SHL may collect, not a description of what any particular assessment does. The two documents are not in conflict; they are written for different readers. The administrator guide covers the commonly configured feature set, while the privacy notice must enumerate everything SHL is capable of processing across all of its offerings and regions.

Two entries in that list matter more than the rest for a technical candidate. "Suspicious objects detected in your image capture (i.e., mobile phone)" means at least some configurations run object detection over the webcam frames rather than only counting faces. "Face verification" means at least some configurations compare you against an identity document. Neither appears in the five-feature administrator list, which is why you cannot assume the light configuration.

The notice also records the mundane technical collection — "browser types; operating systems; and date / time stamps" — and states a retention period of "a maximum of seven years" where SHL acts as controller. Crucially, it puts the employer in the driving seat: the employer decides "which Assessments you take, what type of Personal Information is collected, the purposes for which it is collected and who your Personal Information is shared with." SHL is generally the processor. If you want to know what was enabled for your assessment, the employer is the party who knows.

How flags become a High or Low proctoring index

An SHL release note dated 7 July 2025 documents a "customized proctoring index," described as a flag indicating "the likelihood of a candidate exhibiting suspicious behavior or using unfair means."

The mechanics are worth understanding precisely:

• The employer selects which proctoring features contribute to the index.
• The system weighs "the type as well as the frequency of the violation" against predefined thresholds.
• Those per-feature thresholds are fixed. SHL states "the Individual feature thresholds cannot be altered/customised. These are set & maintained based on studies."
• The output is two bands, High or Low, replacing an earlier three-band High/Medium/Low scheme.
• It appears "under a dedicated proctoring section in reports" and can be pushed into an applicant tracking system through integrations.

So the index is not a continuous suspicion score and not a human judgment. It is a threshold comparison over employer-selected counters, collapsed into one of two values, delivered to a recruiter who may see it inside their ATS alongside your name.

What a flag does not establish

SHL is unusually direct about the limits of its own output. Its proctoring guide states: "There is no 'one size fits all' interpretation of the flags thrown by the system, and SHL is unable to provide further guidance on the likelihood of dishonest behaviour."

That sentence should shape how you react if you are flagged. The vendor is saying the signal does not carry a built-in meaning; the employer supplies the interpretation. A browser-off-focus percentage rises for a dismissed notification, a second display waking, an accessibility tool, or a genuinely switched tab, and the counter does not distinguish between them. What happens next is a human policy decision, which varies widely by company — we have written separately on what actually happens after an assessment integrity flag, including when it is worth contesting.

SHL's stated approach to AI use

SHL published a position on generative AI in assessments on 25 May 2023. Its argument is partly that the risk is overstated for some formats — it notes that "using ChatGPT to answer questions may help in some cases, but it may just as easily hurt your score" — and partly that it layers three mitigations: proctoring flags on "print screen, copy-paste, browser toggle" plus face detection; plagiarism-style "pattern matching" across constructed responses; and a stated development effort to "automatically categorize a candidate response as either written by a human or an AI model," based on analysis of over one million responses.

Treat the third as a vendor claim. It was described as in development in 2023, and we found no published independent validation of its accuracy, false-positive rate, or current deployment status as of September 2026. Text classifiers of that kind have a documented tendency to misfire on non-native English writing and on heavily edited prose, which is a real risk to legitimate candidates rather than only to dishonest ones. For contrast on how a different vendor documents the same problem, see what HackerRank proctoring detects and records.

If you use a desktop assistant

Control is a desktop AI interview assistant for Windows and macOS. It runs as a native desktop overlay with real-time transcription, screenshot analysis, hotkey-driven responses, and optional phone remote pairing. Control documents that the overlay stays out of screen captures and does not interfere with a shared screen. There is no browser-only or hosted web mode, and no Linux build.

Set against SHL's feature list, that distinction cuts both ways, and it is worth being blunt about which way it cuts in each configuration.

The three event counters — browser off focus, print screen count, copy paste attempt — observe activity inside the assessment's own browser context. A separate native application is a different surface from a browser tab, and screen-share capture is a different mechanism again from a webcam still.

The camera-side features are the opposite case. Periodic image capture, multi-face detection, object detection for mobile phones, and face verification observe the room, not the computer. Where an employer has enabled those, a phone-remote workflow is a poor fit by construction: the feature exists specifically to notice a phone in frame. No desktop architecture changes what a camera sees. We do not claim otherwise, and any tool that tells you it does is selling you something it cannot deliver.

Because behavior varies by operating system, meeting client, and version, the only trustworthy check is one you run yourself on your own machine before it matters. Control publishes a focus test for exactly that purpose. Verify rather than assume, and re-verify after OS updates.

Practical steps before an SHL assessment

• Ask the recruiter which SHL product you have been sent and whether proctoring is enabled. They can find out, and asking is not suspicious.
• Run SHL's Browser Check, which it describes as verifying "if your device settings are compatible," on the machine you will actually use.
• Assume desktop or laptop, since SHL states the proctoring features do not run on tablets or phones.
• Reduce accidental counter noise: close messaging apps, silence notifications, and use a single display where practical.
• Request accommodations in advance if a disability affects gaze, movement, or the use of assistive software, rather than explaining a flag afterward.
• If you are in a jurisdiction with AI hiring disclosure or deletion rights, exercise them before the assessment rather than after.

The short version

For a standard SHL aptitude assessment with proctoring enabled, expect periodic webcam stills and counters for tab-switching, screenshots, copy-paste, and extra faces in frame — rolled into a two-band High or Low index that your recruiter sees. For a video interview product, expect your face and voice to be recorded and machine-scored by design. For heavier configurations, expect that live audio-video monitoring, identity document capture, and phone-in-frame object detection are all on SHL's documented list of what it may collect.

What you cannot do is determine which of those applies from the invitation alone. Ask, run the browser check, and prepare on the assumption that the camera-side features may be active. If you are evaluating whether a desktop assistant fits your interview format at all, start with the focus test and decide from observed behavior on your own hardware rather than from anyone's marketing page, ours included.