How to Spot AI-Written CVs, and What to Do Instead

You can no longer trust a polished CV on its own. AI writing tools let any applicant produce a clean, keyword-matched CV in minutes, so the CV now tells you less about the person than it used to. Trying to catch AI-written CVs is a losing game. A better move is to bring the real evidence earlier: hear every candidate answer a few structured questions before you build the shortlist.

Why CVs stopped being a reliable signal

For years, a clear and well-written CV was a rough signal of effort and communication skill. Generative AI broke that link. An applicant can paste your job ad into a chatbot and get back a CV that mirrors its keywords, uses confident action verbs and has no typos.

That does not make the applicant dishonest. Many people use AI the way they once asked a friend who writes well. But it does mean two CVs that look equally strong can come from people with very different skills.

High-volume roles feel this most. When hundreds of CVs for a customer service or sales role all read like your job description, a CV scan or keyword filter ends up ranking applicants on how well their tool wrote, not on how well they will do the job.

Warning signs that a CV was mostly written by AI

None of these signs proves anything on its own. Treat them as prompts for questions, not as reasons to reject.

  • Phrases copied almost word for word from your job ad, sometimes including your own internal wording.
  • Achievements with no numbers, names, tools or context, such as "drove customer satisfaction" or "spearheaded cross-functional initiatives".
  • The same polished tone in every section, including parts that are usually rough, like a first internship.
  • A skills list that matches every requirement in the ad, including skills that rarely appear together at that level.
  • A CV and cover letter that read like one template, with no personal detail you could check.
  • Vague dates, titles or responsibilities exactly where you would expect detail.

Why AI detectors are not the answer

Tools that claim to detect AI-written text return a probability, not proof. Research from Stanford found that popular detectors wrongly labelled more than half of a set of essays by non-native English writers as AI-generated. In Egypt and the Gulf, where many strong applicants write English as a second language, that is a serious risk.

A detector also answers the wrong question. What you need to know is not who typed the CV, but whether this person can do the job.

What to do instead: hear the candidate early

  1. Keep the CV as context, not as the filter. Use it to decide what to ask about, not who gets asked.
  2. Ask every applicant the same few role questions by voice. Three to five questions about real situations in the job show how a person thinks and speaks in a way a CV cannot. For example: "A customer shouts at you on the call. What exactly do you do?"
  3. Ask about their own experience. A question like "Tell me about the busiest week in your last job and what you did" quickly shows whether the experience is real and understood.
  4. Score every answer against the same rubric. Decide in advance what a strong answer includes, so every candidate is judged on the same evidence.
  5. Keep a person in the decision. Use the screen to choose who moves on to a human interview, with the answers in front of the reviewer.

An example screen for a customer service role

Illustrative example

A BPO is hiring 40 customer service agents for an English-speaking account and receives 900 applications. Instead of scanning CVs, the team sends every applicant a short voice screen: an English self-introduction, one angry-customer scenario, one question about shift work, and one question about the most recent job on their CV. Reviewers compare the scorecards, listen to the key answers of the strongest candidates, and invite them to a final interview.

How Gahiz fits

Gahiz does not read, score or rank CV documents, and it does not try to detect AI-written CVs. It helps you hear every applicant instead.

Candidates take a short, structured AI voice interview from their phone, in Arabic, English and other languages your roles need. If you choose, the CV can shape which experience questions are asked, but scores come only from what the candidate says. Each answer is scored against your rubric with the evidence attached, and your team makes every hiring decision.

Common questions

Can you reliably detect an AI-written CV?
Not reliably. Detection tools give probabilities, and research has found they wrongly flag many non-native English writers. Treat signs of AI writing as a reason to ask more questions, never as proof.
Should we reject candidates who used AI to write their CV?
Usually not. Many strong candidates use AI to format and polish a CV. What matters is whether they can explain their experience and handle the job, which you can check by hearing them answer.
Does Gahiz score or filter CVs?
No. Gahiz never scores, filters or ranks CV documents. A CV can give the interview context, but scores come only from the candidate’s spoken answers, and people make the decisions.

Interview question banks for these roles

Free interview scorecard templates

More hiring guides

Sources

  1. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E. and Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7).