For most of the past decade, the story of AI in recruitment has been told from one side of the desk. Employers invested in applicant tracking systems, layered on screening algorithms and, more recently, handed parts of the shortlisting to generative models. The candidate was the one being assessed.
That story is now out of date. Now, candidates have access to the same technology, and they are using it at scale with AI-generated job applications: tailoring CVs in seconds, generating a cover letter for every opening, and in some cases submitting applications they have barely read. The result is a hiring market where both sides have automated their respective processes, but neither is getting what it wants.
The numbers behind the flood
The scale of the problem came into focus when Indeed’s CEO, Hisayuki Idekoba, described hiring as a vicious cycle. Job seekers send out AI-polished applications by the dozen, employers struggle to sort through them, and qualified people hear nothing back.
The data behind that description is stark. An analysis of roughly 640 million applications across more than 6,000 companies found that applications per job more than doubled between 2022 and 2025, while the average number of recruiters per organisation fell by 56%. Far more demand on the pipeline, far fewer people to manage it.
Idekoba framed the core challenge as one of trust. How can an employer tell whether an applicant is a real person, genuinely interested and actually qualified? And how can candidates be confident that the employer on the other side is real too? Fake applicant schemes have added a security dimension to what used to be purely an efficiency problem.
How the AI vs AI hiring loop works
The reversal is worth spelling out, because it explains why adding more technology on the employer side has not fixed anything:
- A candidate uses AI to write a CV and application tailored to the job ad.
- The employer uses AI to screen that application against the same job ad.
- The candidate, rejected or ignored, uses AI to optimise and resubmit, or to apply for fifty more roles.
- The employer, now facing even more volume, tightens its automated filters.
Each turn makes the next one worse. Candidates learn that volume is the only rational strategy when responses are rare. Employers learn that manual review is impossible at that volume, so they automate further. Before long, two AI systems are negotiating with each other over a job that a human will eventually have to do.
Why the loop destroys hiring signal
The underlying issue is not volume on its own. It is that the loop strips out the information employers need to make good decisions.
Applications converge. When everyone optimises against the same job description using similar tools, CVs start to look alike. Keyword matching was always a crude proxy for capability. When candidates can generate a perfect keyword match on demand, it stops working as a proxy at all.
Screening becomes a target. Some candidates have gone beyond polishing. Researchers found that at least 1% of 200,000 real CVs contained hidden prompt injections designed to trick AI screeners into rating them highly. The number is small, but it shows the screen itself is now something to be gamed.
Good candidates disappear into the noise. Idekoba put it bluntly: “You got 1,000 applications, and you think all 1,000 people are not qualified?” If a thousand applicants cannot produce one viable hire, the process is failing, not the talent pool.
Candidate experience collapses. Only 4% of US job changers with at least a high school education landed their last role after a recruiter reached out. Everyone else applied cold, often dozens of times, frequently into silence. That silence damages employer brand long after the role is filled.
Even the platforms are recalibrating
Job platforms are also changing how they use AI. Indeed, for example, tested a feature that automatically applied for jobs on behalf of candidates, but paused it within weeks. Instead, it moved to a model where AI helps draft applications, while candidates review them before submitting. At the same time, Indeed is increasing verification for both employers and candidates, including employer identity checks, phone verification and verified professional licences.
The shift is clear: less automation in the application process, with more focus on verifying candidates, and smarter matching that works more like a recruiter assessing a CV than a system simply scanning for keywords.
This points to a bigger question: how can organisations adapt their hiring processes as AI changes how candidates apply?
Five ways HR leaders can break the loop
Most organisations cannot control what candidates do with AI. They can control how their own process responds.
- Stop treating volume as success. Application counts are a vanity metric. Track signal instead: the share of applicants who reach interview, quality of hire, and time to a confident decision. A requisition with 80 relevant applicants is healthier than one with 1,000 undifferentiated ones.
- Redesign the front door. If a generic CV can be produced in seconds, ask for something that cannot: a short, role-specific question, a work sample, or evidence of a particular skill. Friction, used deliberately, filters for intent.
- Move from keywords to skills evidence. Screening against job-description language rewards whoever is best at mirroring that language. Structured, skills-based assessment is harder to game and fairer to candidates from non-traditional backgrounds.
- Put human judgment where it matters most. AI can triage, but it should not be the only thing standing between a qualified candidate and a conversation. Define the points where a person must review, reinvest the time automation saves into recruiter capacity, and make accountability for automated decisions explicit.
- Be transparent about AI on both sides. Tell candidates how AI is used in your process and what you expect from them in return. Clear guidance on acceptable AI use reduces the incentive to game the system and signals that the organisation values substance over polish.
Where the conversation goes next
The AI vs AI hiring loop is not a technology problem that better technology alone will solve. It is a design question about what hiring is for, where judgment sits, and how trust is built between strangers at scale. There is no single answer, and the organisations making progress are often learning from each other.
That is the thinking behind the “AI Sees Talent Differently Than You Do” theme at Horizon Summit 2026. For leaders working out how AI is reshaping talent decisions on both sides of the market, this table-top peer discussion is a chance to compare what is actually working.
