How ATS Screening Really Works in 2026

Applicant tracking systems have changed. But the biggest change isn't a mysterious robot that now rejects every CV. It's that employers are layering ATS data with stricter filters, AI-assisted search and ranking, assessments and human verification, all to cope with a flood of applications. This guide walks through each filter, what actually gets people eliminated at each stage, and what to do about it. Exercises are built in, and your answers save in your browser as you go.

The real picture is less dramatic and more useful. When 25 recruiters working across Workday, iCIMS, Greenhouse, Lever and other platforms were interviewed in late 2025, 23 of them said their systems don't auto-reject CVs for formatting, content or design. Rejections were either manual or triggered by eligibility filters [3]. So the ATS isn't where most people lose. The real problem is volume [1][2][5]:

Read together, those numbers explain almost everything about modern screening. Employers have two problems at once: too many applications, and too many that sound qualified but are hard to trust. Every filter below exists to solve one of those two problems. Once you see which one, you know how to get through it.

Part 01 — The funnel, stage by stage

A typical hiring funnel now has seven stages. Not every employer uses all seven, and the same ATS can be configured very differently by two companies. But this is the shape. Open each stage to see what happens there and what eliminates people:

Part 02 — Knockout questions: the strongest automatic filter

Knockout questions decide far more than keyword density does. They're yes/no or multiple-choice gates on the application form, and a disqualifying answer can end the application before anyone opens your CV. Typical ones: Do you have the right to work in the UK? Will you need sponsorship? Can you work from this location, or be in the office this many days? Do you hold SC or DV clearance? Do you meet the minimum years of experience? Is your salary expectation within the range?

Candidates often blame “ATS keywords” when the actual cause was an application-form rule. Recruiters see it from the other side: in the same set of interviews, 100% of recruiters said they use eligibility filters for things like work authorisation, required licences and location [3].

Exercise — the five-minute eligibility check

Before you spend an hour tailoring an application, pick one job ad you're considering and run this check. It saves as you go. If you can't tick an item, that's your answer: skip the role or find out more before applying.

• I meet the right-to-work or sponsorship requirement as written.If the ad says “no sponsorship”, believe it.
• I can genuinely work the location and hybrid pattern.Three days in Manchester is a rule, not a preference.
• My salary expectation overlaps the advertised range.If there's no range, check comparable roles so your answer is realistic.
• I hold, or am eligible for, any clearance it requires.SC and DV roles often filter on this first.
• I've separated the real must-haves from the wish list.“Essential” and “desirable” sections exist for a reason. Knockout questions usually map to the essentials.

Part 03 — Parsing is not rejection

Every ATS parses your CV into structured fields: employers and job titles, employment dates, education, skills and certifications, location and contact details. A complicated layout can make that extraction harder, and a badly parsed profile is harder to find in search. But the claim that every CV must be a plain, single-column Word document is overstated. The question that matters is whether your key information is still readable once the CV becomes plain text.

Exercise — the plain-text test

Open your CV PDF, select all, copy, and paste it into a plain-text editor (Notepad, TextEdit in plain-text mode, or any code editor). That's roughly what a parser starts from. Then check:

• Each job title sits next to its employer and dates, in reading order.Two columns that interleave line by line are the usual failure.
• Dates use one consistent format (e.g. Mar 2023 – Jun 2025).“Summer '23 – now” is easy for a human to read and hard for a parser.
• Nothing important lives only inside an image, icon or skill bar.A five-dot Terraform rating becomes nothing at all as text.
• Section headings are recognisable: Experience, Education, Skills, Certifications, Projects.Creative headings like “My Journey” can confuse section detection.
• Your contact details and location appear as real text, not in a header graphic.

Myth or reality? Flip each card

Part 04 — Recruiters search; ranking is getting semantic

With hundreds of applicants, recruiters don't read in order of arrival. They search and filter. A query for a cloud role might look like this:

AWS AND Terraform AND ("Cloud Engineer" OR "Platform Engineer" OR DevOps)

They can also filter by location, experience, current employer, qualifications, application date, or answers to screening questions. If a word isn't in your CV, you don't show up for that search. That isn't a rejection; nobody looked. Newer tools soften this. Greenhouse, for example, says its AI suggests related search terms so recruiters cast a wider net, and states that it never uses AI to rate candidates or auto-reject applications [4]. So exact keywords still help, but context matters more every year.

Older matching relied on literal overlap: the ad says “AWS Lambda”, the CV says “AWS Lambda.” Newer matching can weigh adjacent and transferable skills, similar job titles, seniority and recency, and whether a skill appears in paid work, education or a project. Implementation varies hugely, though. One company uses sophisticated matching; another uses the same ATS as a filing cabinet. In the late-2025 recruiter interviews, 44% had AI fit-scoring available, but only 8% used it to auto-reject. Most treated scores as a rough guide or ignored them [3]. That's a small US sample, so treat the percentages as a signal, not a census.

Exercise — be the recruiter

Take a real job ad for your target role. Write the Boolean search you'd type if you were the recruiter with 300 applications to get through. Use the job title, platform and tools the ad repeats most.

Now check your own CV against it. Tick a term only if it appears inside a sentence of evidence, not just in a skills list:

Part 05 — The CV gets you into verification, not the job

Because anyone can now produce a polished application in minutes, employers trust the CV less as proof. Greenhouse's 2025 AI in Hiring research found 91% of US recruiters had spotted candidate deception, and 34% were spending up to half their week filtering spam and junk applications [5]. The response has been to add evidence stages after the CV: technical screening questions, timed coding or cloud exercises, work samples, structured interview scorecards, live architecture discussions, identity and credential checks, and more in-person or supervised steps.

The UK has clear examples. After applications rose by nearly 30% year on year, with many of the written tasks easy to complete using AI, Teach First sped up its move to live, task-based assessment, including “micro-lessons” taught to an assessor [6]. Across graduate recruitment, the Institute of Student Employers recorded 140 applications per vacancy, a 14% rise in a like-for-like sample of employers [2].

This creates a contradiction worth understanding. Employers use more AI to screen applications, and at the same time they're more suspicious of AI-polished candidates. So a CV that lists “AWS, Terraform, Docker, Kubernetes, CI/CD, Python” is easy to generate and therefore increasingly weak. A strong bullet shows the decision, the environment and the result:

Provisioned an ECS-based deployment pipeline with Terraform and GitHub Actions, reducing manual release steps from eight to two while keeping production credentials in Secrets Manager.

That one bullet works at three levels. The ATS extracts the technologies. The recruiter sees relevance at a glance. And the hiring manager gets something concrete to question you on, which is exactly what they want.

Exercise — rewrite one bullet

Exercise — can you defend it?

For every major claim on your CV, a hiring manager can pull any of these threads. Flip each card and answer it out loud for your strongest bullet. If you can't, the bullet is a liability, not an asset.

Part 06 — Why the career plan comes before the CV

Put the whole funnel together and a pattern appears: every filter rewards alignment with one specific role. That's why rewriting your CV, however well, is the wrong first move. You can't make relevance findable, or claims defensible, until you've decided what you're relevant for. A career plan isn't a nice extra that sits alongside the job search. Every other step depends on it. Here is how it maps onto each layer:

Layer 1 — Pass eligibility

Knockout questions filter on location, right to work, clearance, salary and minimum experience. A plan starts from those realities: which roles you can genuinely be hired into now, in which places and at what level, and which ones need a stepping-stone first. Without it, you keep applying for roles a rule will reject in forty minutes.

Layer 2 — Make relevance retrievable

Recruiters search by job title, platform, tools, certifications and sector. A CV written to cover cloud engineering, DevOps, security, data and AI all at once matches every search a little and none of them strongly. A plan picks the target title family, and with it the vocabulary: the words that go in your headline, your bullets and your LinkedIn profile, placed inside evidence rather than a giant keyword list.

Layer 3 — Make every major claim defensible

Depth takes hours, and you have a limited number each week. You can make maybe three or four claims truly defensible in the next few months: the project you can talk about for 45 minutes, the certification that backs it, the failure you can explain. A plan decides which ones, in what order, so your study time turns into evidence that holds up in the verification stage instead of a spread of half-finished courses.

There's a volume argument too. At 110 applications per tech hire, sending another hundred generic applications only adds to the pile. With a plan, you can send ten targeted applications that pass the rules, match the search and hold up under questioning. That's more work per application and far less work per offer.

Exercise — the plan-readiness check

Tick each question you can answer right now, in one sentence, without hesitating:

Your next seven days

Here's the whole guide as one week of work. Tick items as you go; they save.

• Day 1 — Write down your target role in one line.If you can't, that's the first job, and the Orientation Plan exists for it.
• Day 2 — Collect 10 real job ads for that role and highlight the terms that repeat.
• Day 3 — Run the plain-text test on your CV and fix anything that scrambles.
• Day 4 — Write the recruiter search for your role and close the gaps with evidence bullets.
• Day 5 — Rewrite your three weakest bullets as action + tools + environment + result.
• Day 6 — Rehearse the six “defend it” questions out loud for your strongest project.
• Day 7 — Send two targeted applications, each after the five-minute eligibility check.

Pass the rules → appear in the recruiter's search → prove the match quickly → survive technical verification.

Sources & data notes

Every figure above was checked against its source in September 2026. Where a statistic comes from one country's respondents or a small sample, the text says so. The widely shared “75% of CVs rejected by ATS” figure is deliberately not treated as data: no study behind it was ever published.

1. [1]Technology employers receive 51% more applications per opening, 110 applications per hire, 0.7% offer rate (45% below average). SmartRecruiters, Technology Benchmark Recruiting Metrics 2025 (drawn from ~90M applications for 1.5M jobs across 95 countries). smartrecruiters.com/resources/article/technology-benchmark-rverified
2. [2]140 applications per graduate vacancy; +14% in a matched sample (123:1 → 140:1). Institute of Student Employers, Student Recruitment Survey 2025. ise.org.uk/knowledge/insights/552/the_application_explosion_verified
3. [3]23 of 25 recruiters said their ATS doesn't auto-reject for formatting, content or design; all use eligibility knockout filters; 44% had AI fit scoring, 8% used it to auto-reject. Enhancv, “Does the ATS Reject Your Resume?” (25 US recruiters, Sept–Oct 2025, small sample). enhancv.com/blog/does-ats-reject-resumes/verified
4. [4]Greenhouse “never uses AI to rate candidates or auto-reject applications”; AI suggests relevant search terms. Greenhouse, “How does Greenhouse use AI?” (January 2025). my.greenhouse.com/blogs/how-does-greenhouse-use-ai-heres-eveverified
5. [5]91% of recruiters have spotted candidate deception; 34% spend up to half their week filtering spam and junk applications; 41% of 1,200 US job seekers admit using prompt injection / hidden text. Greenhouse 2025 AI in Hiring Report (November 2025; the figures quoted are from the US sample). greenhouse.com/newsroom/an-ai-trust-crisis-70-of-hiring-manaverified
6. [6]Teach First: nearly 30% more applications year on year; accelerating in-person, task-based assessment such as micro-lessons. The Guardian, 13 July 2025. theguardian.com/technology/2025/jul/13/graduates-teach-firstverified
7. [7]Some AI recruitment tools allowed filtering by protected characteristics or inferred gender and ethnicity from names; almost 300 recommendations issued. ICO, AI tools in recruitment — audit outcomes report (November 2024). ico.org.uk/action-weve-taken/audits-and-overview-reports/202verified
8. [8]Employers using automated decisions should be transparent, monitor for bias, and tell candidates how to challenge a decision and request human review (following the Data (Use and Access) Act 2025). ICO, “Here's what jobseekers need to know about automated recruitment decisions” (March 2026). ico.org.uk/about-the-ico/media-centre/news-and-blogs/2026/03verified