Dana has been in talent acquisition for eleven years.

She built her instincts through phone screens, panel interviews, and reading a room. One Monday in March she opened her applicant tracking system to 312 new applications for two open roles. Her company had turned on AI screening eight months earlier, so the system had already ranked everyone, flagged the top 40, and emailed the rest. By the time Dana logged in, the machine had made the first cut.

Marcus had submitted one of those applications.

Laid off six weeks earlier when his employer replaced its entire content team with AI, he spent weeks learning to use AI for his search. He ran each job description through Claude to see which skills carried the most weight, benchmarked himself with ChatGPT, and rebuilt his resume around the language most likely to score well. He sent 34 applications in six days.

How candidates are using AI

Candidate use breaks into three categories, and employers who understand all three make better decisions than those who only know about resume writing:

• Discovery and fit research.** Candidates arrive at interviews having done research that used to require an inside contact.
• Upskilling on demand.** The gap between “I have this skill” and “I have used this skill for years” is narrowing in ways keyword screening cannot detect.
• Application tailoring.** LinkedIn reported a 45% year-over-year jump in applications alongside AI-assisted tools.

How employers are using AI

Employer use has expanded well beyond screening. Research from HR.com found the top five use cases: automating job descriptions (61%), candidate communication (55%), resume filtering (45%), interview scheduling (36%), and candidate discovery (35%).

The efficiency is real. AI screening can cut time-to-shortlist by up to 75%. Mastercard reduced interview scheduling time by more than 85% and scheduled 88% of interviews within 24 hours of a request.

The gap underneath the speed

Here is the problem. Most employer AI was built to optimize existing processes, not to ask whether those processes still work. Dana’s system scores candidates against a rubric built four years ago, before her most important roles changed shape. It is faster. It is also optimizing for a job that no longer exists.

They made the funnel faster. They did not ask whether the funnel was selecting for the right things.

Read the full report

Section One of the 2026 AI for Recruitment Report lays out the full picture on both sides of the table, with the data and the action items that follow from it. Start from the beginning to get the full story.