How reliable is a standard LLM/AI in CV Analysis? We Put It to the Test.

We analysed a large set of CVs multiple times using the same LLM and identical prompts—and compared the results.

What we found:

  • Up to 24% of skills were missing or misclassified
  • Up to 40% variation between repeated extractions

Meaning: the same CV can produce very different profiles.

⚠️ For recruiters, that’s a real risk.

At Actonomy, we combine LLMs with domain-specific models to deliver accurate, stable, and explainable talent intelligence.

Want the full analysis and methodology?
Request our complete report.