How AI Resume Screening Tools Score Your CV (and How to Raise Your Score)

November 9, 2026 · 17 min read

Quick answer: AI resume screening and scoring tools read your CV into fields, drop applications that fail the employer's must-have filters (work permit, location, minimum years), then rank the rest against the job description, mostly by how many of the posting's skills and title words your CV shows. There is no universal "ATS score": each employer sets its own filters and weights. To raise your score, write each skill the way the posting writes it, keep a complete skills line and clear the filters honestly. When we scored seven wordings of the same resumes against 338 live postings, writing "Forecasts" where the posting said "forecasting" lowered the match level on 12% of the postings that asked for it, on our own checker, until we fixed it. You can check your own CV with Seera's free resume match checker.

Image credit: Tima Miroshnichenko / Pexels

How we got these numbers (November 9, 2026): the people are the eight fictional job seekers from our resume tailoring test: data, business and financial analysts, a software engineer and a data engineer, a product designer, a customer success manager and a senior product manager. Each starts from the one-page resume they would send to every job. We changed only its wording, seven ways, never its facts: short forms only, long forms only, both forms, other word endings, US spelling, UK spelling, and no skills line. The postings are a fresh random draw from the 400,000+ live jobs Seera indexes, collected from employers' own career sites: 50 per person with exactly their job title and no more than two from any one employer, of which 338 carried a readable skills list. Every version went through the scoring a resume pasted into Seera's free checker gets, three times over, because the AI step that reads a resume is not perfectly repeatable; three readings of the same unchanged resume disagreed on 0.6% of levels. We then counted how 5,662 live postings with the same eight titles, from 2,172 employers, write the same skills. No employer is named.

Key Takeaways

• Screening software compares words. Before our fix, a resume that said "Forecasts" and "Wireframes" lost the match level on 11.7% of the postings asking for "forecasting" and "wireframing". Same person, same skill.
• Postings use the short form. Of 5,662 live postings, the ones naming AWS wrote only "AWS" 98% of the time, ETL 97%, KPI 87% and CI/CD 82%. UX was the one split down the middle: short in 48%, spelled out in 39%.
• Brackets didn't protect anyone. "Amazon Web Services (AWS)" did exactly as badly as the long form alone, because the AI that reads the resume kept only one of the two.
• The skills line is the list. Removing it cost the business analyst a level on 42% of the postings asking for skills only that line named, and helped the software engineer, whose bullets named tools the short line left out.
• We fixed our own checker. It now treats endings, UK spelling and common abbreviations as one skill. Other screening tools may not, so write skills the way the posting does.

Table of Contents

• How AI resume screening works, stage by stage
• What an ATS score is (and what it isn't)
• The test: same facts, seven wordings
• What changed the score
• We fixed our own scorer. Here's what changed
• How to raise your resume screening score
• Common mistakes
• Frequently Asked Questions
• Useful sources and further reading

How AI Resume Screening Works, Stage by Stage

Resume screening is several steps, not one. The software that tracks applications and the software that ranks them are often sold together, but they do different jobs. A Harvard Business School and Accenture study of 2,275 executives in the US, the UK and Germany draws the line clearly: an applicant tracking system "helps organizations manage and track the pipeline of applicants", while a recruiting management system automates steps "such as automated candidate scoring and interview scheduling". Of the employers that used one, more than 90% said they used it to filter or rank candidates first. Here is what happens to your CV, in order.

1. The file becomes fields

Your PDF or Word file is turned into text and split into name, contact details, roles, dates, education and skills. Anything the reader can't pull out is invisible to everything after it: contact details in a page header, text inside a picture, a two-column layout read in the wrong order. Our parser test of one resume in 14 formats covers which layouts survive.

2. Knockout filters remove applications outright

This is where most automatic rejections happen, and it has little to do with how you write. The employer sets yes/no questions or rules: Are you authorized to work here? Can you work on site in this city? Do you have five years of experience? A wrong answer ends the application. Some employers filter on the CV itself:

"Almost half the companies surveyed weeded out resumes that present such a ‘work gap.’ If an applicant’s work history has a gap of more than six months, the resume is automatically screened out by their RMS or ATS, based on that consideration alone."

You can't word your way past a knockout, and you shouldn't try: a false answer surfaces at the interview or the background check. Check the gates before you write anything; our five-minute eligibility check walks through them.

3. AI resume screening ranks what's left

The applications that pass are scored against the job description and sorted. The details differ by tool and employer, but the ingredients are much the same: how many of the posting's skills your CV shows, how close your job title is to theirs, your seniority and years, and how similar your experience reads to the work as a whole. Our explainer on how AI matches resumes to job descriptions shows the weights Seera uses for each. The skills part is the one you control most directly, and it is the one this post tests.

4. A person reads the top of the list

A recruiter or hiring manager opens the highest-ranked applications first. That's the reader your wording is ultimately for, and the reason keyword stuffing fails even when software rewards it: in our tailoring test, pasting in skills a person didn't have scored as well as honest tailoring, and no interview would agree.

These tools are regulated in some places. The EU AI Act lists "AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates" as high-risk (Annex III, point 4(a)). In New York City, Local Law 144 has required, since July 5, 2023, that an automated employment decision tool have a bias audit within a year before use, and that candidates be told about it 10 business days in advance.

What an ATS Score Is (and What It Isn't)

There is no standard ATS score. An applicant tracking system is a database with a workflow: it stores applications and moves them between stages. When an employer's system ranks candidates, the employer chose the filters and the weights, and you never see the result.

The ATS resume score you see online is something else: a resume checker's own estimate of how well your resume fits one posting, usually the share of the posting's keywords it finds in your text. That is useful for one thing, finding skills you have but didn't write down. It is not a number any employer receives, and two checkers will give the same resume different scores. Seera's checker gives a level (strong, good, stretch) rather than a percentage for the same reason; how accurate those levels are is measured separately.

Two numbers about screening are worth knowing, and one isn't. The one that isn't: "75% of resumes are rejected by an ATS before a human sees them." We could find no published study behind it; the earliest source usually cited is a 2012 sales claim by a resume-optimization company that has since closed. The two that are worth knowing come from the same Harvard study. Most employers used software to filter or rank candidates first, and most believed it filtered out good people:

"A large majority (88%) of employers agree, telling us that qualified high-skills candidates are vetted out of the process because they do not match the exact criteria established by the job description. That number rose to 94% in the case of middle-skills workers."

"Exact criteria" is the key phrase. So we tested how exact a scorer is about words.

The Test: Same Facts, Seven Wordings

Each person's one-page resume was rewritten seven ways. Only the summary, the bullets and the skills line changed; the headline, job titles, employers, dates and education stayed the same.

1. Short forms only: every term with a common abbreviation written short: "KPI", "AWS", "CI/CD", "FP&A", "QBRs".
2. Long forms only: the same terms written out: "key performance indicator", "Amazon Web Services".
3. Both forms: the long form with the short one in brackets, "financial planning and analysis (FP&A)", the advice most resume guides give.
4. Other word endings: skills written as nouns instead of activities, or the reverse: "Budgets, Forecasts, Financial models" for "Budgeting, Forecasting, Financial modeling"; "Wireframes, Prototypes" for "Wireframing, Prototyping"; "REST API" for "REST APIs".
5. US spelling throughout: "modeling", "analyzed", "center".
6. UK spelling throughout: "modelling", "analysed", "centre".
7. No skills line: the SKILLS section deleted; the skills appear only where the summary and bullets name them.

A rewrite can only matter for a posting that asks for the words it changed, so we report each one on those postings: 91 asked for a skill whose ending we changed, 46 for an abbreviated term or a spelling we changed, and 213 for a skill the deleted line had named. Here is the share where the rewrite lowered the match level, on our checker as it was on the morning of November 9:

Across all 338 postings, the original resumes were a strong match for 61%. The word-endings rewrite alone took that to 56%, with nothing about the people changed.

What Changed the Score

Word endings: "forecasts" is not "forecasting" to a word matcher

The biggest effect came from the smallest change. A posting asks for "financial modeling"; your resume says you built "financial models". A person reads those as the same skill. A scorer that compares words sees "modeling" in one and "models" in the other, and counts a miss. On our checker, writing skills as nouns ("Budgets", "Forecasts", "Wireframes", "Prototypes", "Roadmaps") instead of the activity the posting named lowered the level on 11.7% of the postings that asked for those skills. The product designer lost a level on 16% of them and the financial analyst on 9%.

The fix on your side is boring: in your skills line, use the posting's form. If it says "wireframing", write "wireframing", not "wireframes".

Short form or long form: write it the way postings do

We counted how live postings write ten common terms:

For most technical terms the short form is the norm, so a resume that only spells them out speaks a different language from almost every posting. That cost the most where abbreviations are everywhere: the software engineer lost a level on 4 of the 12 postings that asked for AWS, CI/CD, JavaScript or TypeScript when the resume spelled those out. Short forms written short cost nothing in our test, because that is how the postings wrote them. The exceptions are terms like UX and UAT, where postings are split; there, match the one you're applying to.

Both forms in brackets didn't help

"Amazon Web Services (AWS)" is the standard advice: cover both spellings at once. On our checker it did no better than the long form alone, 8.7% of levels lost on the affected postings. The reason was not the matcher but the step before it. The AI that reads a resume into a profile wrote each skill down once, and when it saw both forms it kept the long one. Any tool that turns your CV into a list of skills before comparing can do the same. Brackets are still fine for a human reader, but don't count on them for software: put the posting's form in your skills line.

UK or US spelling: match the posting

Spelling mattered less than we expected, for a simple reason: few skill names differ. The one that did was "modelling", which cost the financial analyst a level on 4% of the postings that used one of the words we changed. Of the live postings using any of the words we tracked, 85% used US spelling only and 9% UK spelling only, so if you apply to US employers, "modeling" is the safer spelling; for a UK employer, copy theirs.

Your skills line is the list software reads

This result cut both ways. Without a skills line, the business analyst lost a level on 42% of the postings that asked for skills only the line had named, such as requirements gathering, process mapping and user stories. The bullets describe that work ("gathered requirements and wrote user stories", "mapped 12 payment-operations processes"), but the AI reader didn't list it as skills. The software engineer went the other way and gained a level on 16%: the bullets named tools such as Node.js, PostgreSQL, JavaScript and SQL that the eight-item skills line had left out, and with a line present the reader took the line as the complete list.

So keep a skills line, and make it complete: every tool and method your bullets show, in the posting's words. A short "top 8" line can hide what the rest of your resume proves.

We Fixed Our Own Scorer. Here's What Changed

These results describe Seera's own checker, and we didn't like them. A person with "Forecasts" on their resume has the forecasting skill; a scorer that disagrees is wrong. So we changed how the checker and Career Mode compare skills: plurals and -ing forms now count as the same word ("forecasts" and "forecasting"), UK and US spellings are matched, and common abbreviations are matched to their long forms ("Amazon Web Services" and "AWS", "key performance indicators" and "KPI"). Words that only look alike stay apart: "accounting" does not match "account management", and "programming" does not match "program management". We checked every skill word in our live index for merges like those before switching it on.

We then ran the same test again on the same postings:

The skills line is still the list. That one is about what the AI reader writes down, not how words are compared, and this fix doesn't change it, so the advice above stands.

What we can't tell you is how any other screening tool behaves. We can only test our own. A system built to match "the exact criteria established by the job description" has no reason to know that "forecasts" means "forecasting", so the safe assumption is that it doesn't.

How to Raise Your Resume Screening Score

Work through the stages in order, because a strong ranking is worth nothing if a filter removed you first.

1. Clear the gates honestly. Read the posting for work authorization, location, years and required licences before you write a word. If one rules you out, move on. The eligibility check takes five minutes.
2. Make the file readable. One column, real text (not an image), contact details in the body rather than a page header, standard section names: Experience, Education, Skills.
3. Name each skill in the posting's form. Same abbreviation or long form, same word ending, same spelling. "AWS" if they say AWS; "quarterly business reviews" if they spell it out; "wireframing" if that's their word.
4. Make the skills line complete. Every tool and method your bullets show, the posting's must-haves first. Don't let a short line hide what the bullets prove.
5. Back each skill with a bullet. Software may only read the skills line, but the person who reads the top of the list wants to see where you used it.
6. Put the posting's job title in your headline if it's honestly what you do. Title is the heaviest single signal in Seera's matching, and the first thing a recruiter reads.
7. Leave out what you don't have. Pasting in missing skills can raise a score, as our tailoring test found, and it fails at the interview.
8. Check, then send. Run the version through a checker against the posting and read which of its skills were found and which were missed.

Our guides to checking a resume against a live posting and tailoring a resume to a job description with AI go through steps 3 to 8 in detail.

Checking one posting at a time is slow when you're searching. Seera's Career Mode matches your profile against live openings on employers' own career sites, with the missing skills listed for each job.

Find your job matches →

Common Mistakes

• ❌ Writing skills as results instead of skills. "Forecasts" and "Wireframes" lost matches that "forecasting" and "wireframing" kept.
• ❌ Spelling out every abbreviation. Almost every posting writes AWS, ETL and KPI short.
• ❌ Trusting brackets. "Amazon Web Services (AWS)" did as badly as the long form alone.
• ❌ A "top 8" skills line. The reader can take it as the whole list and ignore the tools your bullets name.
• ❌ No skills line. Methods your bullets describe but never name may not be counted at all.
• ❌ Chasing an "ATS score". No employer sees the number a checker gives you. Use it to find missing skills, then stop.
• ❌ Believing "75% are rejected by robots". Most automatic rejections come from knockout answers, not your wording.
• ✅ The posting's words, a complete skills line, honest gates, checked against the posting before you send.

Frequently Asked Questions

How do AI resume screening and scoring tools score a CV?

They read your CV into fields, apply the employer's must-have filters, then rank the applications that pass against the job description. The ranking mostly compares the skills and job title your CV shows with the ones the posting names, plus your years of experience and how closely your experience reads to the work. Each employer sets its own filters and weights, so the same CV can rank high for one job and low for another.

What is an ATS score?

There is no standard ATS score. An applicant tracking system stores and moves applications; when one ranks candidates, the employer configured how, and you never see the result. The percentage a resume checker shows you is that tool's own estimate of how well your resume fits one posting: useful for finding missing skills, but not a number any employer receives.

Do applicant tracking systems automatically reject resumes?

Some do, on the employer's rules rather than your writing. Most automatic rejections come from knockout questions such as work authorization, location or minimum years, and some employers filter out employment gaps: in a Harvard Business School and Accenture survey, almost half of the companies said their software screened out resumes with a gap of more than six months. The often-quoted claim that 75% of resumes are rejected by software has no published study behind it.

Should I write acronyms or spell them out on my resume?

Write each skill the way the job posting writes it. In the 5,662 live postings we read, AWS appeared only as AWS in 98% of the postings that named it, ETL in 97% and KPI in 87%, while UX was split: 48% short only, 39% spelled out. When our test resumes used only the long form against postings that used the short one, the match level dropped on 8.7% of those postings on our checker before we fixed it, and writing both forms in brackets did not prevent it.

Does AI resume screening read the skills section?

Yes, and on our checker it treated the skills line as the full list. Without the line, a business analyst whose methods appeared only there lost a level on 42% of the postings that asked for them, while a software engineer gained on 16%, because the bullets named tools the eight-item skills line had left out. Keep a skills line, and make it list every tool and method your bullets show.

Related Reading

• Resume Match Checker: Your Resume vs One Job Description
• AI Job Search That Starts From Your Resume
• How AI Matches Resumes to Job Descriptions, Explained
• How to Tailor Your Resume to Any Job Description With AI
• How to Check Your Resume Against a Live Job Posting
• How Accurate Is AI Job Matching? Match Scores Explained
• Should You Apply? How to Check Your Eligibility for a Job in Five Minutes
• One-Page PDF Resume With a Portfolio Link, Parser-Tested
• How to Get Better AI Job Recommendations: What We Changed and What Moved

Useful sources and further reading

• Harvard Business School and Accenture: Hidden Workers: Untapped Talent (2021) — survey of 2,275 executives and 8,720 workers in the US, UK and Germany; source of the two quotes above on work-gap filters and qualified candidates screened out
• Regulation (EU) 2024/1689, the AI Act — Annex III, point 4(a) lists AI used to filter job applications and evaluate candidates as high-risk
• NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools — Local Law 144's bias-audit and candidate-notice rules
• Seera: Resume Match Checker — free check of one resume against one job description, with the skills it found and missed
• Seera: AI Job Search (Career Mode) — your profile matched against live openings, with the missing skills for each
• Seera wording test, run November 9, 2026 — 8 fictional people, 7 rewordings of each resume, 3 readings each, scored against 338 live postings with their exact job title, before and after the fix; plus a count of how 5,662 live postings from 2,172 employers write ten common skills (aggregate counts only; no employer named)
• Seera: Editorial standards — how we use our own data and write about our product