AI in recruitment has been overpromised since 2015. What changed in 2024-2026 is specific and real. Here's what AI actually does well in hiring now - and what it still does not.
In 2015, AI was going to eliminate CV screening. In 2018, chatbots were going to replace recruiter outreach. In 2020, AI matching was going to make the hiring decision automatic.
None of that happened in the form promised. What happened instead: incremental improvements in search and matching, a lot of "AI" labels on keyword filters, and a persistent gap between the marketing and the reality.
Something genuinely changed in 2024-2026. The change is specific and worth understanding clearly — because it is also being overhyped in new ways.
Transcription quality crossed a threshold. Audio transcription of interview recordings is now accurate enough to be used as the primary source for evaluation documentation. This was not true in 2020. It is true now. The practical consequence: AI can read what was said in an interview and draft structured documentation from it. This is the highest-leverage change in recruiting technology in the last decade.
Language models can extract structured information reliably. Given a CV and a job brief, a current language model can produce a match assessment with specific reasoning that is consistently useful — not occasionally useful. The output quality of AI matching in 2026 is meaningfully better than in 2022.
Semantic search is production-ready. Vector-based search that finds candidates based on contextual relevance rather than keyword matching is no longer a research project. Tools like Spott have deployed it. It works. For senior and specialist roles where the right candidate is not always the most obvious one, this is a genuine capability improvement.
AI cannot evaluate cultural and contextual fit. Whether a candidate's working style will function in a specific team dynamic, whether their approach to ambiguity fits a company that is currently in chaos, whether they will navigate the political landscape of a particular organization — these require human judgment from people with context.
AI cannot manage the relationship. The reason a strong candidate accepts your offer over an equivalent one is almost always relationship-based. The recruiter they trusted. The process that felt respectful. The honest answer to a hard question. None of this is automatable.
AI cannot make the final call. Match scores, process diagnostics, risk flags — AI produces useful inputs. The hiring decision integrates everything you know about role, team, candidate, and timing. That judgment remains human and should stay that way.
Use AI for: transcript-based scorecard drafting (highest ROI), match scoring across large candidate pools, process analytics to identify bottlenecks, communication drafts at standard process stages, scheduling automation.
Keep human: evaluation judgment, candidate relationship management, offer negotiation, the final hiring decision.
The recruiters gaining the most from current AI are using it to eliminate 3-4 hours of daily administrative work — and spending those hours on relationship and judgment work that AI cannot replicate.
AI scorecard pre-fill from interview transcripts. This solves the most persistent problem in structured hiring — scorecards that are not submitted because the effort cost is too high — in a practical, low-friction way. Submission rates go from under 50% to over 85% when the effort drops from 45 minutes to 8 minutes.
AI tools that numerically rank candidates for hiring decisions create potential discrimination risk if the ranking criteria are not transparent and documented. The best approach: use AI for administrative automation and as decision support, not as a decision-maker. Ensure all AI-assisted evaluations are reviewed and approved by a human.
Final hiring decisions, evaluation of cultural and contextual fit, candidate relationship management, and offer negotiation should remain human. These are the tasks where human judgment and relationship skills have the highest impact on outcomes.
Run your free hiring process audit
Connect your ATS in 3 minutes. Get a full breakdown of where your process loses time and candidates — with three prioritised fixes.
Start free audit →Andreas Gruber
Founder of Pickr and ScalingPPL. Former recruiter who placed engineers and operators into European startups and scale-ups for four years before building the tool he wished had existed.