Answer as Yourself at Work, Not the Ideal Employee

Should You Answer Honestly or Say What They Want?

Answer honestly, in the narrow sense of answering as yourself at work rather than yourself on a Saturday. That is not a moral instruction. It is what leaves you with a process you can stand behind later. Where a questionnaire is followed by people who ask you to describe how you actually work, the answers have to belong to one person.

The pressure is real and worth naming. A hundred agree-disagree items where one response is plainly the preferred one does not feel like a measurement. It feels like a quiz with the key taped to the wall.

The research agrees with you about the key.

Where it stops agreeing is on what the key is worth. A questionnaire score is a thin instrument even when nobody shades it. The employer-side writing on this site is direct about one reason: a validity coefficient is measured against one job's performance data and travels no further. The number a vendor quotes may have been earned somewhere else, on other people, doing other work.

That guidance is worth reading in its own words. Do assessment scores still predict how someone performs once AI is in the workflow is written for the people buying the test, not for the people sitting it. That is part of why it is useful to you.

Nobody writes reassurance for an audience that is not in the room.

What the Faking Research Actually Found

Instructed faking moves scores, and by more than a little. Buehl and colleagues had 111 psychology students answer the same twenty-question structured interview twice, once honestly and once told to answer as an applicant would. Mean ratings rose from 3.20 to 3.62, a gap the authors report as a clear faking effect ¹. Those were students with no job on the line.

That last clause is not a throwaway. It was a within-subjects design in a research setting, and the authors say their effect came out far larger than an earlier interview study's and put that down to how the conditions were built ¹.

Read it as a direction that is real, not a size to expect.

The comparison that matters more here is between formats, because a questionnaire and an interview are not equally easy to shade. Van Iddekinge and colleagues ran mock structured interviews built to measure agreeableness, conscientiousness and emotional stability, and had the same people complete matching self-report scales under both honest and applicant-like instructions. Scores in the applicant-like condition rose significantly more for the self-report ratings than for the interviewer ratings ². In that mock setting, the questionnaire was the easier of the two to inflate.

Employer-side guidance ranks a related set of formats, though for a different question: whether someone who cannot do the work passes anyway with an assistant open. Work sample, structured interview, or AI-skills assessment: which one catches AI dependence puts the unwatched take-home easiest and the structured interview hardest of the three, and no personality questionnaire is on that list.

So the prep vendors are not lying about the mechanism.

What they leave out is who the mechanism belongs to. In the Buehl study, the size of each person's gain was modestly related to cognitive ability and to how well they could identify what a question was assessing ¹. That describes who gains, not a technique anyone can sell you.

It also points somewhere more useful than it first looks. Gains ran somewhat larger among people who understood what a question was reaching for, and that understanding is also what having done the work gives you. Someone who has run a project to a deadline knows what an item about planning is asking, and can answer it truthfully and specifically. The two routes end up in the same neighborhood, and only one of them holds up in the conversation that follows.

Will the Test Catch an Engineered Answer?

Nobody can tell you, and an answer that sounds certain in either direction is worth distrusting. The common claim is that a battery repeats each item in different words and scores you on the gap between the two. No verified source here documents which battery does that, or what it does with the result, and a vendor's description of its own product would not settle it.

What is documented sits next door, in research on cheating rather than faking. In one experiment, 315 participants took a knowledge test unproctored and then took it again proctored, and the self-report measures, including an honesty scale, predicted nothing about who had cheated ³.

The stakes there were entry into a gift-card draw rather than a job, and cheating was inferred from the score gap rather than observed ³.

The countermeasures on the employer's side have not been shown to close the gap either. The standard set is specific instructions, honor codes and announced follow-up tests. Reviewing them, the same authors report that a 2020 meta-analysis found those measures could not prevent score differences between proctored and unproctored ability tests ⁴, a null result that is weak evidence rather than proof that the measures do nothing.

An unproven countermeasure is still not a green light.

One large program tested the mechanism from the other end. Between 2006 and 2008, 71,985 applicants to a UK selection program sat unproctored reasoning tests, having signed an honesty contract and been told a supervised verification test would follow. Among those who reached that stage, the share scoring significantly below their earlier result ran from 0.3% to 2.2%, under the 2.5% that chance alone would put in that tail ⁵. Nothing in that system watched anyone during the test.

That was cognitive ability testing with rotating item banks, which is the easy case, and the authors are explicit that the result is descriptive rather than causal. Only about 5% of candidates reached the supervised stage, so nothing is known about the rest ⁵.

None of it transfers to a personality questionnaire, which has no verification form to sit. What transfers is the shape. An unsupervised stage held together in a system where a later stage existed and the same claim had to survive a second look.

For a personality questionnaire, the later stage that matters is any conversation where a person asks how you handle deadlines, disagreement and detail, in your own words.

Why the Engineered Profile Costs More Than It Buys

An engineered profile costs you a fit: you have described a different person to an employer, and the employer may then plan your first months around the description. If the profile says you are energized by constant interruption and cold calls and you are not, that shows up once the work starts, not during the hiring process.

The questionnaire is over in one sitting. The job is not.

A second cost belongs to the employer rather than to you. Traits of the kind many batteries ask about are the subject of a formal complaint. In May 2024 the ACLU filed a Federal Trade Commission complaint against Aon over marketing its hiring assessments as bias-free. The complaint argues that tools measuring traits such as positivity, emotional awareness and liveliness are not clearly job related for most jobs ⁶.

The sharper point in the same filing is that those traits are directly linked with core aspects of the medical understanding of autism and of conditions such as depression and anxiety ⁶. So a tool can be clean on race and sex and still fall hardest on one group.

That is an allegation, not a finding.

The ACLU's announcement reports no agency ruling, Aon disputes the characterization, and the FTC theory is about marketing claims rather than about whether the assessments discriminate. Job-relatedness is decided against a particular job, one at a time.

The engineered answer papers over a question that may not have been worth asking.

If an item is hard for you because of a disability rather than because you are trying to look good, that is different ground with a different route. Does asking for an accommodation hurt your chances covers where that request goes and what the panel is told.

Collapsing that question into this one costs people options they actually had.

On the day, answer as the person your last manager would recognize, at work, on an ordinary week. Then keep the examples ready, because the profile you hand in is a list of things somebody may ask you to prove.

See what gets scored >