A practical guide to AI-resistant hiring assessments

How to use this field guide

An AI-resistant assessment starts with a clear boundary between permitted tools and the independent skill you need to observe. This guide connects five decisions: setting those boundaries, writing useful scenarios, interpreting reasoning evidence, respecting the candidate experience and keeping the overall hiring process focused on work. No single monitoring feature resolves all five.

Use the guide with one proposed assessment in front of you. For every task, write the purpose, allowed tools and expected evidence. Then list the events you could actually observe and the conclusions those events cannot support. This simple separation helps prevent a technical observation from becoming an unsupported accusation. It also makes candidate instructions more concrete.

These chapters are edited selections from the HireValid editorial library, with links to the complete articles. They are gathered as an implementation workbook, not presented as independent research. All scenarios are hypothetical, and the draft requires expert review. The marketing demo does not capture camera data or establish validated detection performance.

Assign a person to review concerns before collecting them. A detailed event timeline is of little value if no one knows how to interpret it or ask a neutral follow-up. Agree what context will be considered, how a candidate can explain an interruption and when another work sample would be more informative than additional monitoring. Keep the response proportionate to the actual role and uncertainty.

At the end of the guide, use the review worksheet to examine both the assessment and the process around it. The goal is not an unqualified claim that assistance is impossible. The goal is relevant evidence under clearly communicated conditions, supported by a human decision and an honest explanation of the system’s limits.

Chapter 1: How to reduce AI assistance in online assessments

Adapted from How to reduce AI assistance in online assessments.

Decide what independent work means

Reducing unauthorized AI assistance starts with a decision about the job, not a detector. Some tasks genuinely require a person to recall information, reason independently or respond without outside help. Others are normally completed with documentation, colleagues and AI tools. An assessment should explain which capability it is trying to observe. Otherwise, a candidate may follow normal workplace habits while the employer interprets that behavior as a rule violation.

Write a short tool policy for each stage. State whether search, calculators, code completion, translation and generative AI are allowed. Explain whether the candidate may take notes and whether those notes can be digital. Give a concrete example when a boundary might be unclear. “You may use the provided reference sheet, but not an external assistant” is more useful than “complete this honestly.” Make the policy available before an invitation becomes a timed activity.

A useful process can include both unaided and assisted tasks. For example, a junior developer might explain a small function without assistance, then use documentation during a separate debugging exercise. The second exercise can assess how they verify a suggested solution. This distinction makes the evidence easier to interpret and avoids pretending that every real workplace task happens in isolation.

Design a task that reveals thinking

A task that can be answered with a generic paragraph is difficult to interpret. Give candidates specific information and ask them to make a decision from it. For a support role, provide a short order history and a return policy. Ask for the next action and a brief explanation. For an analyst, provide a small table with one ambiguous value and ask what needs checking before a conclusion is shared.

The aim is not to make a puzzle unnecessarily obscure. The candidate should have enough information to produce a defensible answer. What makes the task useful is the connection between the evidence and the decision. A strong response identifies relevant facts, acknowledges missing information and explains a proportionate next step. A fluent answer that ignores the supplied facts becomes easier to question without invoking an AI detector.

Randomized item pools can reduce repeated exposure, but randomization does not automatically make versions equivalent. Each version still needs comparable coverage and difficulty. Per-question time limits can constrain outside assistance, but they also introduce speed as part of the measurement. Only use that constraint when it is appropriate, and preserve a process for accommodations. An assessment should not become a test of panic under unexplained pressure.

Treat browser signals as context

A browser can observe some events, such as losing focus or receiving pasted text. It cannot reliably tell you everything happening around the candidate. A second device may be invisible. An overlay may not produce a detectable event. A focus change might come from a password manager, an operating-system message or an accessibility tool. These technical limits matter when explaining the product and interpreting a result.

An integrity report should therefore describe what was observed before suggesting what it might mean. “The assessment lost focus for twelve seconds” is a narrower statement than “the candidate used AI.” The first can be checked and discussed. The second introduces a conclusion that the event alone does not support. Timing patterns and answer similarities also need context rather than automatic accusations.

The proposed HireValid integrity approach combines several signals and keeps decisions with a person. Its website demo uses illustrative weights. It does not represent validated detection performance or monitor your browser activity. Before a production rule is used, the team needs to understand what it measures, how often innocent behavior triggers it and what explanation a reviewer will receive.

Review a concern without making an accusation

Imagine that a candidate pastes a paragraph into a written response. There are several possible explanations: they drafted it elsewhere, used an assistive tool, copied an external answer or misunderstood the rules. Start by checking the instructions they received. If the policy allowed drafting in a separate document, the event is not evidence of breaking that policy. If the rule was unclear, improving the instruction may be the most important action.

A neutral follow-up might ask the candidate to explain their approach and how they prepared the response. A short discussion of the actual content can be more informative than asking whether they cheated. For a technical task, ask why a particular condition was included or what would happen with a different input. Keep the question relevant and comparable with the process used for other candidates.

Record the observed event, the applicable rule, the explanation and the evidence considered. Do not record an unsupported diagnosis of dishonesty. If you decide the evidence remains insufficient, use another proportionate assessment stage rather than inventing certainty. A hiring record should distinguish an unresolved concern from a demonstrated failure to follow a clearly communicated instruction.

Chapter 2: Situational judgement tests: examples and better practice

Adapted from Situational judgement tests: examples and better practice.

What makes a situational question useful?

A situational judgement test presents a scenario and asks the candidate to choose, rank or explain possible responses. Its usefulness depends on the quality of the situation and the reasoning behind the scoring. A believable story alone is not enough. The question needs a clear purpose, sufficient context and a defensible relationship between the response and the work.

Start with a decision that occurs in the role. A sales representative may need to clarify an objection. An administrator may need to prioritize conflicting requests. A team member may need to respond to feedback. Identify the behavior you want to examine and the information available to the person making the decision. Do not write the answer first and then invent a story that makes it appear inevitable.

The format is a way to gather evidence, not a personality diagnosis. A response to one hypothetical situation does not prove how someone will behave in every workplace. It can help structure a follow-up conversation and reveal what the candidate notices. Interpretation should remain tied to the scenario, with attention to alternative reasonable answers and the limits of the evidence.

Example one: a price objection

Suppose a prospect tells a junior sales representative that the proposed service is too expensive. The representative has not yet clarified the prospect’s desired outcome or current process. A useful question asks what they would do next. One option might be to pressure the prospect for an immediate signature. Another might be to ask what outcome would make the investment worthwhile. A third might be to offer an unauthorized discount.

The second response is defensible because it gathers relevant information before deciding whether the offer fits. It does not guarantee a sale, and the assessment should not imply that it does. The point is the quality of the next step: clarify value and constraints rather than guessing at the objection or promising something outside the representative’s authority.

A follow-up could ask what information would change the candidate’s approach. That reveals whether they can adapt the principle to context. If the prospect has already explained a strict budget limit, repeating a generic discovery question may be less useful. Good scenario design includes the facts that determine the choice rather than rewarding a memorized sales phrase.

Example two: conflicting deadlines

An administrator receives two requests due at the same time and cannot complete both. The scenario should state the consequences, dependencies and available authority. Without those facts, “do the most important task first” merely restates the problem. A useful response may involve explaining the conflict and agreeing priorities with the responsible person, rather than silently missing a deadline.

Ask the candidate what they would communicate. Would they identify the capacity constraint, propose options and make the trade-off visible? Would they check whether part of the work can be completed earlier or reassigned? The answer should show a practical approach to managing the situation rather than a claim that excellent time management makes every conflict disappear.

The scoring should allow more than one reasonable sequence if the scenario supports it. A candidate may first clarify a dependency or first notify the requester. Evaluate the reasoning and the result of that choice. An overly narrow answer key can punish sensible judgement because it differs from the author’s preferred ordering of actions.

Example three: difficult feedback

A colleague appears upset after receiving feedback. A workplace judgement item might ask how to respond. Offering a private conversation and asking how they understood the feedback can be more constructive than assuming they are unable to accept criticism or discussing their reaction with the whole team. The rationale concerns respect, privacy and understanding, not a diagnosis of anyone’s emotions.

Keep the scenario within the candidate’s role. A peer and a manager may have different responsibilities. If the question does not specify the relationship, several answers may be defensible. Avoid scoring the candidate on an unstated assumption about authority or confidentiality. Add the minimum context needed to make the decision meaningful.

This kind of task should not be marketed as a clinical measure. The Emotional Intelligence test is planned around workplace scenarios. It does not diagnose mental health or determine a person’s overall character. Use the response as a basis for a relevant discussion, and be cautious about drawing broad conclusions from a small number of hypothetical choices.

Chapter 3: Cognitive ability tests: what they measure and how to use them

Adapted from Cognitive ability tests: what they measure and how to use them.

Understand the question a reasoning test answers

A cognitive ability test is intended to gather evidence about reasoning, learning or solving unfamiliar problems. Depending on the design, it may include numerical information, written passages, visual patterns or a mixture of formats. A result tells you something about performance on those tasks under those conditions. It does not provide a complete measure of a person’s potential, character or suitability for every job.

That distinction matters because a single score can look more comprehensive than it is. A candidate may reason well in a visual format while finding a language-heavy format difficult. Another may understand the problem but work slowly under a strict time limit. Those differences deserve interpretation in relation to the role rather than a quick conclusion about who is “smartest.”

Start by asking why this kind of evidence belongs in the process. If the job involves learning new systems, interpreting unfamiliar information or comparing options, reasoning may be relevant. Document that connection. If the test is included merely because another employer uses it, you have not yet explained its purpose. A recognizable assessment name is not a substitute for a job analysis.

Separate reasoning from job knowledge

A reasoning task and a knowledge test measure different things. A person may know a spreadsheet shortcut because they have used a particular application for years. They may also be able to work out a new formula from documentation. Both can matter, but they are not interchangeable. Decide which ability is required immediately and which can be developed through onboarding.

For a junior analyst, a numerical reasoning question might reveal whether they interpret a percentage correctly. A SQL task might reveal whether they can filter rows or reason about a join. A short work sample could show whether they check missing values before explaining a result. These methods complement each other when each has a clear purpose. They become redundant when several versions merely reward the same familiarity.

The planned General Cognitive Ability test combines several reasoning areas. Its public sample is a practice illustration, not a validated instrument. The Junior Data Analyst bundle shows how reasoning can sit beside role-specific technical skills. The employer still needs to judge whether that proposed mix matches the actual work.

Interpret a percentile carefully

A percentile describes relative standing within a reference group. A result at the 75th percentile means the score is above 75% of that documented group under the relevant scoring convention. It does not mean 75% of questions were correct. It also does not mean a 75% probability of succeeding in the role. Those are different quantities that require different evidence.

Always look for the norm group, sample size, collection method and version. A broad working-population reference may answer a different question from a group of applicants for a specific role. If the reference sample is small or poorly matched, a precise-looking percentile can be misleading. Reports should explain those limitations instead of hiding them behind a confident visual.

Comparing scores across test versions also requires care. Changes to items, timing or scoring can alter interpretation. A useful report records the version used for the candidate’s attempt. If norms are updated later, the earlier report should remain understandable. HireValid’s benchmark page describes the intended documentation and explicitly notes that live norm data is pending validation.

Examine language and timing requirements

A reasoning question may contain a substantial reading component even when it is labeled numerical. Complex wording can turn a task about interpreting a table into a test of advanced language proficiency. Review the instructions and item text for unnecessary complexity. If the job requires plain workplace communication, use plain workplace communication in the assessment wherever possible.

Timing can also change the skill being observed. A strict limit may emphasize processing speed, familiarity with the format or comfort under pressure. That may be relevant in some contexts, but it should not be assumed. Tell candidates the timing in advance and explain whether practice is available. Consider appropriate accommodations rather than treating one clock setting as inherently fair for everyone.

Visual reasoning is not automatically free of barriers. Small shapes, low contrast or reliance on color can make a task inaccessible. A lower reading load does not establish cultural neutrality or universal accessibility. Evaluate each format on its own merits. The right question is whether the task offers an appropriate way to demonstrate the skill for the intended population.

Chapter 4: Give candidates a better testing experience

Adapted from Give candidates a better testing experience.

Make the purpose visible

A candidate should understand why an assessment belongs in the hiring process. “Complete this test” asks for time without explaining what the employer hopes to learn. A better invitation describes the relevant skills, the expected duration and the next stage. It also makes clear that a person will review the evidence rather than implying that software will decide the application on its own.

Explain the connection to the role in ordinary language. A support candidate might be asked to interpret a policy and write a helpful response. A virtual assistant might prioritize a set of requests and explain what needs clarification. When the task resembles the work, the candidate can understand its purpose even if the format is unfamiliar. A long battery of unrelated tests is harder to justify.

Do not promise a particular outcome in exchange for completion. Completing an assessment does not guarantee an interview or an offer. You can still explain the selection stages and communicate respectfully. Clarity does not require certainty about every future decision; it requires being honest about what is known and what the candidate can expect next.

Give an honest time estimate

An assessment’s time cost includes reading instructions, checking equipment and trying practice questions. If the timed portion lasts twenty minutes but setup takes another ten, the invitation should not imply that the entire activity fits into twenty minutes. Ask someone unfamiliar with the workflow to try it and note where extra time is needed.

Keep the assessment proportionate to the stage of hiring. A brief initial screen and a substantial later exercise create different expectations. Avoid asking every applicant to complete extensive work before the employer has reviewed basic suitability. When a longer task is necessary, explain why and consider how to make the request reasonable. Do not use the exercise to obtain unpaid commercial output.

State whether the candidate must finish in one sitting and what happens if the connection drops. Explain deadlines with a time zone. If the system supports resuming, describe the limits accurately. A calm statement about recovery is more useful than a vague promise that everything will be fine, especially when a timer continues during a disconnection.

Explain tools and monitoring before the start

Candidates need to know which tools are permitted. Search, calculators, notes, translation software and AI assistance can be appropriate in some tasks and restricted in others. Write the rule for the specific activity rather than relying on an undefined instruction to work independently. Give an example when a boundary is likely to be misunderstood.

Monitoring also needs a clear explanation. If an employer enables browser-event logging or optional camera checks, describe what is collected and why before the candidate consents. Do not hide a camera requirement behind a generic privacy link or reveal it after the timed session begins. Provide a route to discuss an alternative where appropriate.

The HireValid candidate page describes the intended experience. The website’s sample tools do not activate a camera or submit practice results to an employer. That distinction is important: a marketing preview should not create anxiety about hidden monitoring or suggest that a practice score has entered a real application record.

Make accommodations easy to request

A general commitment to fairness is not enough if a candidate cannot find the right person to contact. Put the accommodation route in the invitation and before the start screen. Assign someone responsibility for responding. Explain how the candidate can ask for an adjustment without sending unnecessary sensitive information through an ordinary form.

Focus on the skill being measured and the barrier the person faces. Extra time, an accessible format or an alternative to a camera check may be relevant in different circumstances. A single standard adjustment will not fit every need. Employers should understand their applicable obligations and seek qualified advice where necessary rather than expecting software defaults to resolve every situation.

Review the interface as well as the policy. Can the task be completed with a keyboard? Are instructions readable at increased text size? Is information communicated by color alone? Can someone understand an error message and recover from it? Accessibility is a practical part of whether the assessment measures the intended skill instead of an unrelated interaction barrier.

Chapter 5: Start skills-based hiring without an HR team

Adapted from Start skills-based hiring without an HR team.

Replace a vague profile with a useful task list

Skills-based hiring starts with the work, not with a new label on the same job advertisement. If a small business asks for a degree, several years of experience and a familiar job title without explaining the responsibilities, it may be filtering for background rather than capability. Some credentials are genuinely necessary. Others are habits inherited from an old template. Review each requirement instead of assuming it belongs.

Imagine hiring an office administrator. Begin with the tasks: reconcile order records, coordinate a shared calendar, respond to routine requests and flag exceptions. Ask which tasks need independent competence on arrival and which can be taught. This creates a more practical profile than “organized self-starter with excellent communication skills.” It also makes the eventual interview easier to prepare.

A useful task list names an action, an object and a standard. “Check daily order entries against the source record and escalate unexplained differences” tells a candidate more than “attention to detail.” You do not need a lengthy competency framework to write that sentence. A manager who knows the work can draft it, then ask a colleague to check whether it reflects the real role.

Separate essential requirements from preferences

An essential requirement should have a defensible connection to the job. A preference may make onboarding easier but should not quietly become a screening rule. List them separately. If your team can teach a particular software interface, say that equivalent experience is welcome. If a legal qualification is required, explain that clearly and verify it through an appropriate process.

Be careful with requirements that sound neutral but may be vague proxies. “Native speaker,” “young and energetic” or “a perfect cultural fit” can distract from the actual capability needed and create fairness concerns. Describe the communication task, working conditions and observable behavior instead. Obtain appropriate advice for employment wording in your location, especially where requirements might exclude protected groups.

The job description templates provide a structure with placeholders. They do not know your compensation, working hours or local obligations. Replace every placeholder and remove anything that does not apply. A template is useful when it prompts thought, not when it allows the employer to publish an unexamined list of demands.

Choose one direct demonstration per major skill

For each essential requirement, ask what a candidate could show in a reasonable amount of time. A brief record-checking exercise can reveal how someone handles mismatches. A scheduling scenario can reveal whether they notice dependencies and ask useful questions. A short written response can reveal clarity and tone. The demonstration should resemble the work without requiring confidential information or a lengthy unpaid project.

Use synthetic data and clearly fictional situations. Do not ask candidates to solve a live customer problem that your business can use commercially. Provide enough context to make the task fair to people who have not worked at your company. Explain any terms that are specific to your process. Hidden knowledge rewards familiarity rather than the intended skill.

A general assessment can complement a work sample, but it should not dominate the process merely because it produces an easy number. The Office Administrator bundle suggests several relevant areas. Review its scope against the task list. If the proposed bundle is longer than the decision requires, remove tests rather than treating the default as mandatory.

Write a small scoring guide before reviewing

A rubric turns an impression into a set of questions about evidence. For the record-checking exercise, you might consider whether the candidate found important mismatches, avoided inventing corrections and explained the next step. For a scheduling task, you might consider whether they recognized constraints and communicated a conflict. Keep each dimension distinct enough that reviewers can explain it.

Use plain descriptions for rating levels. A middle level might mean the answer is mostly correct but misses an important check. A stronger level might require accurate work plus a clear explanation of uncertainty. Avoid labels such as “excellent” without saying what makes the response excellent. The label adds confidence without helping someone apply the scale consistently.

Try the rubric on two or three fictional answers before using it on candidates. Ask whether two reviewers identify similar evidence and whether disagreements reveal ambiguous wording. This is a calibration exercise, not a validation study. Its purpose is to make the process understandable and reduce avoidable inconsistency before real people are affected by it.

Your integrity review worksheet

Create four columns for each proposed signal: observed event, possible explanations, follow-up question and decision relevance. For a focus change, the event is leaving the assessment window. The explanations may include a notification, assistive software or an external tool. A neutral follow-up asks what happened rather than asserting which explanation is true.

Review the task itself before adding another signal. A generic essay with no supplied facts may tell you less than a short, specific decision exercise. A follow-up explanation can show how the candidate used the information. The strongest improvement may be a better question, not a larger collection of monitoring data.

Finally, review the privacy and accessibility consequences of every setting. Define the purpose, access and retention before enabling collection. If a requirement is unresolved, keep the feature disabled and use a proportionate alternative. Planned capabilities should remain labeled as planned until implemented and verified.

Final decision questions

• What must be demonstrated without assistance, and why?
• Which tools are explicitly permitted?
• Does each signal describe an observation rather than a verdict?
• Can a candidate provide context through a clear process?
• Are limitations communicated without promising universal detection?
• Is a person responsible for every consequential decision?

Continue with the integrity overview, sample test and candidate guidance.