There is a story job seekers tell each other a lot at the moment: send off a CV, hear nothing, and assume a robot rejected you in half a second flat. It is a satisfying story because it explains the silence. It is also, on the evidence, more complicated than that.
The more accurate picture is this. An ATS in 2026 is best understood as a workflow system with optional screening automation bolted on, not a single universal gatekeeper that reads every CV and passes judgement. It stores applications, parses CVs into structured data, applies whatever eligibility rules a recruiter has set, supports the recruiting team’s day-to-day work, and increasingly layers AI on top for matching, sourcing, scheduling and summarising. How much of that pipeline actually decides your fate, rather than simply organising your application for a human, varies enormously from one employer to the next.
That distinction matters, because it changes what you should actually do about it.
What an ATS is doing behind the scenes
Underneath the single word “ATS” sit several quite different jobs. It captures your application and any screening questions, then parses your CV into searchable fields, such as employer names, dates, qualifications and skills, a stage that is genuinely error-prone: information that reads perfectly clearly to a person can go missing if the document layout confuses the parser. Eligibility filters get applied next, things like location, right to work, salary expectations or minimum qualifications. Ranking and scoring may follow, though whether that happens by AI, by simple rules, or not at all depends entirely on the platform and how the employer has configured it. And underneath all of that sits the unglamorous but important administrative layer: moving people between stages, sending acknowledgements and rejections, scheduling interviews, and keeping an auditable record.
An ATS does not necessarily reject every application on its own. Automatic rejection usually comes from explicit knockout criteria the employer has set, from high-volume filtering rules, or from a dedicated AI screening module, rather than from the ATS acting as a blanket judge of every applicant.
How much of the process is actually automated
It helps to think of this on a spectrum rather than as one setting.
The ATS is mainly an administrative database. It stores and parses applications and lets a recruiter search by skill or keyword, but a person is still doing the deciding, even if they only ever look at candidates a search has surfaced.
Probably where most mid-sized and larger employers now sit. The system does more of the narrowing itself: rejecting candidates who fail knockout questions, filtering by eligibility criteria, suggesting or ranking candidates by apparent fit, and handling scheduling and routine communication automatically. Humans still make the final call, but only on a shortlist the system has already shaped.
Found particularly in high-volume sectors like retail, hospitality, healthcare and logistics. AI may score CVs directly against a job description, infer skills from job titles and history, rank the entire applicant pool, run chatbot or video-based screening, and in some cases trigger automatic progression or rejection with limited human review.
It is worth noting that even “AI adoption” figures are inconsistent, precisely because they are measuring different things. One 2025 SHRM study put the share of organisations using AI to support recruiting at just over half, while separate research from iCIMS and Aptitude found a similar overall adoption figure but a much smaller share, under a fifth, using AI broadly across the majority of their hiring. A tool being switched on somewhere in the business is a different thing from it running on every requisition.
Which systems are involved, and how
There is no single dominant platform. Large enterprises tend to cluster around a handful of names, alongside a long tail of specialist and regional systems, and market share estimates shift considerably depending on which sample of employers a study uses. The cards below cover the platforms most likely to affect a professional applicant in the UK, Europe or North America, along with what tends to matter for the person on the other end of the application.
The main platforms at a glance
Watch for: strict, well-configured eligibility questions. Answer them carefully rather than relying on the CV to compensate.
Watch for: terminology needs to hold up consistently across regions and business units.
Watch for: rigid requisition and eligibility rules; clunky forms. Follow instructions to the letter.
Watch for: make sure your CV parses cleanly; assessment stages are common before any human contact.
Watch for: interview structure and scorecard criteria matter as much as keyword matching; strong human review stage.
Watch for: includes candidate rediscovery, so your profile may resurface for future roles even after rejection.
Watch for: accuracy of your skills and experience fields matters a lot, since the system is built to search on them.
Watch for: automated matching can surface you against roles that are a poor fit; check what you’re being matched to.
Watch for: process varies hugely by employer, so avoid assuming one company’s approach tells you how another works.
Watch for: conversational and chatbot screening is common; treat chatbot answers as seriously as a form.
Watch for: be explicit about skills, not just job titles, since the system infers competencies from language.
Watch for: chatbot questions can end an application quickly; keep answers clear, complete and honest.
Watch for: AI features are expanding, so don’t assume less scrutiny than a large employer.
Watch for: a well-written, clear CV tends to go a long way here.
Watch for: straightforward system; a human is likely reading applications directly.
Watch for: formatting cleanliness still matters for the parsing stage.
Watch for: GDPR-aware and UK-oriented; screening tends to be deliberately set up rather than default-heavy.
Watch for: less about beating an algorithm, more about a clear application and a good candidate journey.
Watch for: your profile may be reused across multiple client roles, not just the one you applied for.
Watch for: built for compliance and redeployment; expect your details checked against several roles over time.
Watch for: very rules-driven; incomplete or imprecise answers to formal questionnaires can disqualify a strong candidate.
Watch for: limited transparency by design; treat every field in the application as potentially load-bearing.
Where AI actually sits in the stack
AI itself now shows up in several distinct layers rather than one place. It sits inside the ATS for parsing, matching and summarising. It sits inside broader HR platforms, connecting recruitment data to workforce planning. Specialist tools handle skills inference, sourcing, automated video interviewing, and conversational screening. The important distinction for a candidate is between AI that assists a recruiter, drafting a job advert or suggesting interview slots, and AI that makes or heavily shapes a decision about you. Those two things get talked about as if they are the same, and they are not.
What actually goes wrong, and what is still contested
The genuine complaints from job seekers are consistent across the research: qualified candidates rejected with no explanation, CVs misread by parsers, repetitive application forms, chatbots that cannot handle anything unusual, and a general absence of human contact anywhere in the process. Those are real and well documented.
The contested numbers
Some of the more dramatic figures circulating, such as claims that 75 or 80 percent of CVs are rejected automatically before any human sees them, are worth treating with real caution. They tend to trace back to single studies or preprints rather than independent, representative audits of the ATS market as a whole, and they get repeated far more confidently than the underlying evidence supports. It is fair to say a meaningful share of applications never reach a person. It is not currently fair to say that a fixed, universal percentage does.
Where the real risk lies
The fairness question is better established and matters more. Because these systems learn from historical hiring patterns, they can reproduce and sometimes amplify the biases already present in that data, disadvantaging candidates on the basis of gender, race, age, disability, career breaks, non-traditional education or simply an unfamiliar job title. In the UK, government guidance is explicit that AI recruitment tools can perpetuate bias and create digital exclusion, and that employers remain legally responsible under the Equality Act 2010 for the outcomes their tools produce, including the duty to provide reasonable adjustments. In the US, the EEOC has warned that algorithmic screening tools can unlawfully screen out qualified disabled applicants if accommodation processes are not built in.
Underneath all of this sits a transparency problem. Most applicants have no way of knowing whether their CV was simply stored, ranked by an algorithm, or reviewed by a person, what criteria led to a rejection, or how to request a human review or a reasonable adjustment. That opacity is arguably the more pressing issue for candidates right now, more so than the exact percentage of CVs an algorithm supposedly discards.
What this means for how you approach an application
Given all of that, it makes sense to treat every serious application as though it has to get past three different filters, because in practice it often does.
The three filters
Technical: can the system read your CV correctly at all. This is the easiest one to control. Use a conventional layout, standard section headings such as Profile, Experience, Education and Skills, and keep your contact details in the main body of the document rather than in a header or footer, which some parsers skip entirely. Avoid text boxes, columns and graphics in the version you submit through a portal, even if a more designed version exists for networking purposes. Where a platform lets you check how it has parsed your details, it is worth actually looking.
Relevance: does your evidence genuinely match what the employer has asked for. This is where tailoring earns its keep. Use the employer’s own terminology where it accurately reflects your experience, spell out both the full term and any abbreviation, such as artificial intelligence and AI, and make sure your strongest, most relevant experience is not buried under less relevant detail. Knockout questions deserve honest, careful answers; do not assume a strong CV will override an eligibility question you have skipped over.
Human credibility: the one most candidates underestimate. Once your application reaches a person, whether that is at application stage or interview stage, generic language does you no favours. Quantify what you actually changed and what happened as a result. “Managed a team” is forgettable. “Managed a team of eight through a systems migration that cut processing time by a third” is not. AI tools can help you get to a first draft or sharpen a sentence, but the final version needs to sound like you, with your own specifics and your own judgement in it, because that is precisely what a generic, AI-smoothed application cannot convincingly fake.
Beyond the ATS itself
Beyond the document itself, it is worth remembering that the ATS is only ever one route in. A genuine referral or a direct introduction to someone on the hiring team frequently sidesteps the initial screening stage altogether, which is exactly why networking has become more valuable, not less, in a more automated hiring market.
The bottom line
The technology behind hiring has changed a great deal, but the fundamentals of a strong application have not. Systems that read CVs reward clarity and relevance. Human beings, who still make the final decision in the great majority of cases, reward specific, credible evidence that you can do the job. The most useful response to a more automated process is not to try to outsmart it, but to make sure your genuine experience is as easy as possible for a machine to find and for a person to believe once they see it.
If you would like a proper audit of how your CV is likely to read on both counts, that is exactly the kind of work I help clients with.
Resources
- SHRM — The Role of AI in HR Continues to Expand (2025 Talent Trends)
- iCIMS & Aptitude Research — AI Adoption Report 2026
- State of ATS 2026: Which ATS Do Fortune 500 Companies Use?
- ATS Market Share 2026: What 3,222 Top Employers Use
- State of AI Recruiting 2026: How TA Teams Are Actually Deploying AI
- GOV.UK — Responsible AI in Recruitment (DSIT guidance)
- GOV.UK — Reduce Unconscious Bias in CV Screening
- U.S. EEOC & DOJ — Warning against Disability Discrimination in AI Hiring Tools
- Rabczuk, R. (2025) — The Resume Parsing Crisis of 2025 (unreviewed preprint; source of the contested 75–80% rejection figure, treated with caution in this article)
[…] first: the two-column CV myth, because the formatting advice most people follow is wrong, and what job seekers need to know about modern ATS and AI recruitment, because what happens to your CV before a human sees it has changed […]