Tech hiring has a signal problem. Job descriptions now ask for AI skills by name, and candidates, many of them using AI to write their applications, put the same names back on their CVs. When the job description says "LLM" and the CV says "LLM", the match tells you almost nothing about who can do the work.
This article combines the job description data in Lenny Rachitsky's State of the tech job market: late 2026 with Gartner research on how candidates use AI. It is written for talent acquisition leaders and hiring managers at tech companies. Our view: in 2027, the CV becomes a starting point rather than evidence. Tech employers should move their main hiring signal to work samples and structured interviews, and set clear rules for how candidates may use AI. It expands one part of our guide to the tech job market in 2027.
Two trends that weaken keyword screening
Employers ask for the same AI terms
The report shows AI terms near the top of job descriptions in every core function. In product management, "LLM" is now the most-mentioned skill and "Claude" is third. In engineering, "LLM" has just passed Java. In design, "Claude" became the second most-mentioned skill in spring 2026. These are counts of words in job descriptions across global openings at tech companies.
Candidates write applications with AI
Gartner's July 2025 research release reports three findings that matter here:
| Finding | Source |
|---|---|
| 39% of candidates used AI during the application process; of those, 54% used it to generate CV text, 50% for cover letters and 29% for answers to assessment questions | Gartner survey of 3,290 job candidates, 4Q24 |
| 6% admitted to participating in interview fraud, either posing as someone else or having someone else pose as them | Gartner survey of 3,000 job candidates, 2Q25 |
| By 2028, one in four candidate profiles worldwide will be fake | Gartner prediction |
Put the two trends together and the result is predictable. Candidates read the same job descriptions, and AI tools make it easy to mirror their language. The report does not discuss CV writing or fraud; linking the two trends is our reading.
What CVs still tell you, and what they do not
CVs remain useful for facts that are hard to fake and easy to check: where someone worked, for how long, in what kind of role. They are weak evidence for skills, and weaker still for AI skills, because the vocabulary is now everywhere.
A practical rule for 2027: use the CV to decide whether a candidate's background is plausible for the role, not whether they have the skill. Move the skill decision to later stages, where candidates have to show it.
Build the assessment around evidence
1. A short, realistic work sample
Give candidates a task that resembles the real work, with a time limit that respects their time. Examples:
- Engineers: extend a small codebase and explain the trade-offs, or review a pull request that contains subtle problems.
- Product managers: write a one-page plan for an AI feature, including how to measure its quality.
- Designers: propose a direction for a real, simplified problem with a rough prototype.
Score the work against a written rubric, the same way for every candidate. Write the rubric before you see any submissions, so the standard does not shift to fit the first strong answer.
2. A follow-up conversation on the work sample
Ask candidates to walk through what they did and why, and change one constraint to see how they adapt. This is where someone who relied on AI without understanding the result stands out, and where genuine skill shows.
3. Structured interviews
Every interviewer for a role uses the same questions, the same scorecard with four or five dimensions, and written descriptions of "strong" and "not enough". Scores are submitted on the day of the interview, before interviewers discuss the candidate. Consistency matters more as roles broaden, because interviewers without a shared rubric score the same answer differently.
Set clear rules for AI use
Banning AI from the process is hard to enforce and sends the wrong message for roles that use AI every day. Gartner's recommendation is to set clear expectations and communicate hiring standards, explaining how the employer defines acceptable use of AI. A simple policy:
- Applications: candidates may use AI to help write, but everything must be accurate and their own experience.
- Work samples: state whether AI tools are allowed. For most tech roles, allow them and score how candidates check the output.
- Live interviews: candidates answer in their own words, without assistance, unless the task says otherwise.
Publish the policy on your careers page and repeat it in interview invitations. Candidates who know the rules are less likely to cross them.
Verify identity in proportion to the risk
Interview fraud is a security risk as well as a hiring risk. Gartner quotes its research director: "Candidate fraud creates cybersecurity risks that can be far more serious than making a bad hire." Proportionate steps:
- Check identity at a defined point, such as before the final interview, using a standard process for every candidate.
- Keep cameras on in video interviews, and make sure the person at the final interview is the person from the first.
- Consider one in-person stage for sensitive roles, such as those with access to production systems or customer data. Gartner found that 62% of candidates said they are more likely to apply to a position if the organisation requires in-person interviews (2Q25).
- Tighten background and reference checks for roles with privileged access.
Collect only the identity data you need, tell candidates why, and follow the data protection rules in each market where you hire.
Keep the process fair in candidates' eyes
Stricter assessment can push good candidates away if it feels arbitrary. Trust is already low: in the same Gartner release, only 26% of candidates said they trust AI will fairly evaluate them, and in a 1Q25 survey of 2,918 candidates, 25% said they trust employers less if they use AI to evaluate their information.
Three habits protect the candidate experience:
- Explain each stage before it starts: what the task is, how long it takes and how it is scored.
- Keep work samples short, and pay for longer ones. Experienced tech candidates often have several processes running and will drop one that asks too much.
- Say where people decide. If you use AI anywhere in screening or interviews, tell candidates what it does and confirm that a person makes the hiring decision.
Measure whether screening still works
Two numbers show whether your early stages still separate candidates:
- First technical interview pass rate. If many candidates who passed CV screening fail the first technical interview, screening is letting through the wrong people.
- Agreement between stages. Compare work sample scores with final decisions. If they rarely agree, either the task or the rubric needs work.
Review both every quarter. A falling first-interview pass rate can be an early sign that keyword screening has stopped working.
A note for APAC teams
Candidates in Singapore, Malaysia, Indonesia and Hong Kong apply in several languages, and matching on English tool names is less reliable on CVs written in other languages. That is one more reason to move skill decisions to work samples and interviews, held in the language the job is done in.
Next step
Pick one technical role and add a short work sample and a written scorecard before the next opening. Then write a one-paragraph AI use policy and add it to your interview invitations. Tools that let you define competencies once per role, such as AI interviews from Moka AI, can keep first rounds consistent across interviewers. For the full picture of the 2027 tech job market, read our guide to the tech job market in 2027. For choosing AI recruiting tools for tech hiring, see our guide.

Author
Moka AI Editorial Team
The Moka AI Editorial Team creates insightful, practical, and engaging content on HR, recruitment, and AI innovation. We explore emerging trends, share industry insights, and turn complex ideas into actionable knowledge to help HR professionals navigate the evolving world of work.



