The volume of AI-generated CVs hitting job listings in 2025 and 2026 is not a future problem. It’s happening now, at scale, and it’s making keyword-based screening actively counterproductive.
What AI-generated CVs look like
AI-generated CVs are optimised for keyword matching. They contain every skill you listed in the job description, phrased almost exactly as you phrased it. They describe achievements in complete, well-structured sentences. They’re formatted perfectly.
They’re also often hollow. The achievements are plausible but unspecific. The career narrative is coherent but generic. There’s no texture, no unusual detail, no indication of what this person actually built, broke, or learned.
The tell is homogeneity. Real CVs are messy and specific. AI CVs are smooth and general.
Why keyword matching amplifies the problem
If your screening system matches keywords, it will rank AI-generated CVs at the top of every pile. The AI that wrote the CV was explicitly optimising for your keywords. It wins the keyword game every time.
The result is a shortlist full of people who are great at prompting ChatGPT, not necessarily great at the job.
Screen for specificity, not keywords
The counter-move is to screen for evidence of specific impact, not presence of keywords. Numbers. Dates. Outcomes. Names of products, teams, or problems.
“Improved conversion rate by 34% over six months by redesigning the onboarding flow for mobile users” is a real achievement. “Drove significant improvements in key conversion metrics across multiple platforms” is an AI sentence.
Sieve screens for semantic fit and evidence of real impact rather than keyword density, which makes it significantly harder to game with AI-generated content. See how it works at sievecv.com.