Candidates are producing polished AI generated CVs that pass keyword filters but fail technical screens.
The proportion of technology job applications containing AI-generated or AI-heavily-assisted content has grown significantly in 2026. While current adoption remains relatively low, the trend toward AI-enhanced applications is expected to accelerate, with AI-polished resumes and cover letters becoming increasingly common in job applications. The pattern is already visible in the data: application volumes are up 40 percent year-on-year, but the proportion of applications that convert to interviews has not increased proportionally. The gap is filled by applications that look polished but do not represent genuine capability. Tallenxis
Understanding what recruiters and hiring managers have changed in response, and what specifically makes an application stand out in a pool of AI-polished submissions, is directly actionable for technology candidates who want their genuine capability to be visible in a market where surface-level polish is no longer a differentiator.
A specialist technology recruiter who reviews hundreds of CVs monthly develops a pattern recognition for AI-generated content that is accurate enough to inform screening decisions. The markers are not individual features but combinations: unusually consistent bullet point structure across all roles (real CVs have more variation because different employers and roles were written at different times), achievement statements that are plausible but generic (reduced costs, improved efficiency, increased team performance, all without the specific numbers or mechanisms that genuine experience produces), and vocabulary that precisely matches the job posting language without the contextual depth that genuine familiarity with a technology or domain produces.
The recruiter response is not to reject AI-assisted CVs categorically. It is to apply heavier qualification in the first conversation, asking specific and contextual questions that require genuine experience to answer rather than questions that AI-polished preparation has likely covered.
The characteristics that consistently identify genuine applications from AI-generated or AI-heavily-assisted ones are specificity, narrative coherence, and the presence of uncomfortable truths.
Specificity: genuine experience produces numbers, contexts, and mechanisms that AI cannot generate accurately without the information in the first place. “Reduced pipeline latency from 340ms to 47ms by implementing Redis caching with TTL parameters tuned to our specific access pattern” is specific because only someone who did the work knows those specific numbers and that specific approach. “Improved system performance by implementing caching strategies” is generic because it could describe any caching implementation in any system.
Narrative coherence: a CV written by a genuine professional shows a career trajectory that makes sense, with role changes that are explained by logical progression, skill development that follows a visible path, and a technical focus that becomes increasingly specific over time. AI-generated CVs often have a polished but incoherent career narrative because the tool is optimising for language quality rather than professional coherence.
Uncomfortable truths: real CVs include roles that ended for complex reasons, periods of growth that involved learning from failure, and honest assessments of what was challenging. The AI-polished CV tends toward relentless positivity that is readable as artificial by anyone who has hired extensively.
The answer to the AI generated CVs problem is not to avoid AI tools entirely. It is to use them for specific, legitimate purposes while ensuring the content remains genuinely yours.
AI tools are productive for: grammar and style checking (the equivalent of a good editor), structural suggestions (does this CV communicate clearly and is it in the right order), and initial drafts that you then revise significantly with specific details from your actual experience. The draft is a scaffold, not a submission.
AI tools are counterproductive for: generating achievement statements from a vague description of what you did (the result is generic because the input is generic), describing technical work in terms that are technically accurate but not grounded in your specific implementation, and writing a professional summary that does not reflect how you would actually describe yourself in a conversation.
The test for whether AI has been used productively is whether the final output is something you could speak to in detail in a first phone conversation with a recruiter. If you received a call about your CV and were asked to explain the achievement statement in detail, can you? If no, the statement should not be in your CV.
The first recruiter conversation has always served a qualification function. In 2026, it has added a specific purpose: confirming that the CV represents genuine experience rather than AI-polished presentation.
The questions that specialist technology recruiters use for this purpose are contextual and specific. “Walk me through the Databricks implementation you described on your CV. What was the specific data volume you were handling and what was the latency requirement the business had for the pipeline?” A candidate with genuine experience answers this with specific numbers and mechanisms. A candidate whose CV was AI-generated from a vague description of working with Databricks pauses, generalises, or produces answers that do not match the specificity of the CV.
This is not a trick question designed to expose dishonesty. It is a reasonable request for the context that confirms genuine capability. The candidates who ace this conversation are those whose CV accurately represents their genuine experience at the right level of specificity.