Here is the finding most recruiting blogs do not lead with: cost-per-hire and time-to-hire have both gone up in the three years since generative AI became widely available, not down. That is from the SHRM 2025 Talent Trends data, and it is the single most counterintuitive number in the whole AI-in-work conversation. Adoption of AI in HR went from 26% to 43% over a year. The metrics the AI vendors promise to fix have moved in the wrong direction.
The mechanism is not mysterious once you see it. Candidates now use AI to generate applications at a scale recruiters cannot screen at human pace. Recruiters use AI to screen those applications, but the screening models are themselves trained on human-readable resume signals candidates’ AI tools optimise against. The funnel is bigger, the signal-to-noise is worse, and the time required to actually find the right person has gone up rather than down. The work has not been automated. It has been multiplied.
The adoption figures, sourced
The same SHRM report shows AI use inside HR teams climbing from 26% to 43% year over year, with 64% of AI-using HR teams applying it to recruiting, interviewing, or hiring specifically. The LinkedIn Future of Recruiting 2025 report puts active recruiter use of generative AI at 37%, with sourcing and outreach as the dominant use cases. Both figures are in the same direction, with the SHRM number capturing broader use across HR and the LinkedIn number capturing recruiter-specific generative AI use.
The split by company size is sharp. Roughly 60% of extra-large organisations use AI in HR, against 33% of small companies and 35% of midsize. The pattern there is consistent with most other workplace AI adoption data: enterprises move first, midsize follows, small businesses lag, but the gap closes faster than it has for any previous productivity technology.
What recruiters are actually using AI for
Three workflows pulled most of the recruiter adoption.
Sourcing and outreach is the largest. Tools like LinkedIn Recruiter, hireEZ, and general-purpose models are routinely used to draft outreach messages tuned to a candidate’s profile, generate first-pass shortlists, and rewrite job posts to match audiences. The savings here are real and recruiters report them honestly.
Resume screening is the second. AI-assisted screening, where the tool ranks candidates against the role’s stated requirements, has been part of ATS platforms for years and has gotten meaningfully better with generative AI in the last 18 months. The downside is the same as the upside: the better the screening tool, the more candidates apply, because the candidate-side AI tools optimise for whatever the screening tool reads.
Interview support is the third. Note-taking, transcript summarisation, and structured-interview rubric scoring have moved from manual to AI-assisted at most large recruiting teams. The actual interview judgement is still human. The administrative wrap is increasingly not.
Where the productivity gain went
Recruiters who use AI well are not getting their week back the way the AI vendor pitches imply. They are using the time AI saves on outreach to deal with the volume AI created at the top of the funnel. The net effect on hours worked has been roughly neutral. The composition of the hours has shifted, with less typing and more triage.
The recruiters who are getting genuine commercial advantage from AI are doing one thing differently. They are using AI to do reach work, not screening work. They are sourcing wider, contacting more passive candidates with personalised outreach, and pre-qualifying pipelines for roles before the requisition opens. The work AI compounds is the work that produces a better hire from the same hours, not the work that produces the same hire from fewer hours.
What’s not getting automated
Three categories of recruiter work have remained stubbornly human.
The first is the senior leadership search. The work of mapping a market for a CFO, the discreet conversations with sitting executives at competitors, and the trust required to be told why the candidate is actually open to a move, none of that has been touched by AI. Retained executive search firms have been the most resistant adopters of recruiter AI tools, for understandable reasons: the value is the relationship, and the relationship is the firm.
The second is closing complicated offers. When a candidate is weighing two competing offers and a counter from their current employer, the recruiter’s job is psychological, financial, and timing-sensitive in a way models do not currently handle.
The third is internal stakeholder management. Most recruiters spend more time managing the hiring manager than managing the candidate. The hiring manager who keeps changing the spec, who is slow to give feedback, who has unrealistic expectations about market salary, none of that is an AI use case in any meaningful sense.
The productivity gain, sourced honestly
The closest peer-reviewed evidence on AI productivity for adjacent knowledge work is the Harvard Business School and BCG study showing roughly 25% reduction in time-to-completion on consulting tasks within scope, and the GitHub Copilot RCT showing 55.8% speedup on a benchmark coding task. The realistic productivity gain on the parts of recruiting AI is good at (drafting outreach, summarising calls, ranking resumes) sits in the 20 to 40% range.
The “saves 9 hours a week” figures floating around recruiter AI marketing are not from controlled studies. The honest number is that a recruiter who builds AI drafting and AI summarisation into their week will recover several hours, with the time mostly being absorbed back into the higher-volume top of funnel rather than being available for fewer working hours.
What the next 12 months look like
Three movements are likely to be visible by the end of 2026. First, the cost-per-hire trend will start to reverse, but only at organisations that change how they recruit, not the ones that just add AI to existing processes. The teams that genuinely benefit from AI are the ones that redesigned the funnel, not the ones that retrofitted it.
Second, AI-on-AI authenticity questions will get more pointed. When the candidate’s cover letter is AI-generated and the recruiter’s first response is AI-generated, both sides know it, and the working assumption that the human emerges later in the process will become explicit. Some employers will start advertising AI-free interview processes as a differentiator. Some candidates will respond by sharing their actual writing earlier in the process.
Third, the regulatory layer around AI in hiring will harden visibly. New York City’s local law 144 was the leading edge in 2023. By the end of 2026, expect similar disclosure and audit requirements at the EU level, in California, and in several other jurisdictions. Recruiting teams that have been treating AI compliance as a vendor problem will discover it is a process problem.
If you recruit, the question is not whether AI changes your job. It already has. The question is whether your team is the one quietly closing more roles per quarter with better hires, or the one busier than ever and not entirely sure why.