Creating an ATS‑Friendly Resume in 2026: Practical Rules That Actually Work
Why ATS still matters in 2026
Almost every mid-size and large employer runs applications through an Applicant Tracking System. Workday, Oracle Taleo, SAP SuccessFactors, iCIMS, Greenhouse, Lever, Ashby and SmartRecruiters cover most of the market between them, and staffing agencies add Bullhorn on top. If you applied through a careers page rather than emailing a hiring manager, your resume went into one of these.
The part most advice gets wrong is what the system does with it. An ATS is a database with a workflow attached, not a robot judge. It very rarely reads your resume and decides to reject you. What it does is convert your document into structured records, and those records are what a recruiter searches, sorts and filters on later.
An Applicant Tracking System stores job requisitions and applications, parses uploaded resumes into fields (name, contact, employers, titles, dates, education, skills), and gives recruiters a queue they can search and filter.
Recruiters then work from that parsed record. Most will open your original PDF only after something in the parsed version catches their eye.
So the failure mode is not usually "the robot rejected me." It is quieter than that: your resume parsed into a mangled record, you never surfaced in the recruiter's search, and nobody ever formed an opinion about you at all.
There is real evidence that this filtering costs employers good candidates. The 2021 Harvard Business School and Accenture report Hidden Workers: Untapped Talent found that a large majority of executives surveyed believed their own hiring systems were screening out qualified applicants because those applicants did not match the exact criteria in the job description. Employers know the filters are blunt. They use them anyway, because the alternative is reading thousands of documents by hand.
Your job is to make the parsed record accurate and the human-readable version worth a recruiter's time. Both, not one.
How your resume actually gets read
Understanding the pipeline makes the rest of this article obvious rather than arbitrary.
- Text extraction. The system pulls a raw character stream out of your PDF or Word file. This is where design-heavy documents fall apart, because a PDF stores positioned text runs, not a reading order you would recognise.
- Segmentation. The parser splits the text into blocks and guesses which block is which section, mostly by looking for familiar headings.
- Field mapping. Inside each block it looks for patterns: something that looks like a date range, something that looks like a company, something that looks like a job title.
- Indexing. The extracted values go into searchable fields, plus a full-text index of everything else.
Most ATS vendors do not write step 1 to 3 themselves. They license a parsing engine, and a handful of engines (Textkernel, which absorbed Sovren, along with Daxtra, HireAbility, RChilli and Affinda) sit underneath a large share of the market. That is genuinely useful to know, because it means "ATS-friendly" is not twenty different rulebooks. Get the document clean for a good parser and you are close to fine everywhere.
The engines have improved a lot. A modern parser handles a two-column layout better than the advice from 2019 suggests. But the old systems have not disappeared, some large enterprises are still running Taleo instances configured a decade ago, and you have no way of knowing which one is behind the careers page you are looking at. That asymmetry is the whole argument for a plain layout: a clean single column costs you nothing with a good parser and saves you with a bad one.
Choose an ATS-safe layout
Keep the layout simple and predictable. Not because plain looks better, but because a plain document has only one possible reading order.
Layout choices that help or hurt
| Element | Safe choice | Risky choice |
|---|---|---|
| Overall layout | Single column, left-aligned | Multi-column, magazine-style, or sidebar designs |
| Fonts | Arial, Calibri, Helvetica, Georgia, Times New Roman, embedded in the file | Decorative or script fonts, or fonts the exporter did not embed |
| Graphics and icons | None | Logos, skill rating bars, icons, shapes, background images |
| Section boundaries | Plain text headings | Content inside text boxes or shapes |
| Tables | Avoid; use bullets | Nested tables, merged cells, builder grids |
| Headers and footers | Page numbers only | Name, email, phone, or links placed there |
| Bullet glyphs | Standard round bullets | Wingdings or symbol-font glyphs |
| Headings | Normal letter spacing | Letter-spaced headings like W O R K E X P E R I E N C E |
Two of those deserve an explanation, because they are the ones people argue about.
Columns. The risk is not the column itself, it is what happens when the extractor reads across the page instead of down each column. A job title from the left column can end up glued to a skill from the right, and the record that reaches the recruiter is nonsense. Good parsers detect columns. You are gambling that this employer has one.
Text boxes and header areas. Many parsers ignore Word headers and footers outright, and treat floating text boxes as separate from the document flow. Put your phone number in a header and it can vanish completely.
If your name, email or phone lands in a header, a text box, or a graphic, the parser may drop it. A record with no email address is a record nobody can contact, and you will never find out that is why.
Put contact details in normal body text at the top of page one.
Letter-spaced headings deserve a mention too, since they show up constantly in designer templates. E X P E R I E N C E does not match the string "Experience" in the parser's heading list, so the section below it may never be recognised as work history.
When in doubt, imagine pasting your entire resume into a blank text file. If the order still reads sensibly, your layout is safe.
Use section titles ATS actually understands
Parsers match headings against a list of expected labels. Stay on that list.
- Summary, or Professional Summary
- Skills, or Technical Skills
- Work Experience, or Professional Experience, or Employment History
- Education
- Certifications
- Projects, Publications, Volunteer Experience
Skip the creative alternatives. Where I have been and My journey so far read fine to a person and read as nothing to a parser, which means the roles underneath may not register as employment at all.
1. Summary
2. Skills
3. Work Experience
4. Education
5. Certifications, and anything else relevant
This matches how recruiters skim: fit first, then skills, then recent roles, then education last.
One caveat on the Skills block. Putting it above your experience is good for human skimming, but a few parsers assign skills to whichever job appears above them. If your skills section sits second, that ambiguity disappears, which is another small reason to prefer this order.
Get the machine-readable details right
This is the layer almost every ATS article skips, and it is where a surprising number of resumes quietly break. Dates, titles and contact lines are the fields the system actually indexes.
Dates
Parsers derive your tenure, your gaps and your total years of experience from date ranges. Give them something unambiguous.
- Write
Mar 2021 - Presentor03/2021 - Present. Both parse reliably. - Use a plain hyphen between dates. En dashes and em dashes are handled inconsistently by some older parsers.
- Include the month.
2021 - 2023could be four months or thirty-five, and the system will usually assume the smaller number. - Keep one format across the whole document. Mixing
2021-2023andMarch 2021 to June 2023makes the pattern matching worse, not more interesting.
Job titles
Many systems normalise your title against an internal taxonomy so recruiters can search by role family. Internal titles that mean nothing outside your last company defeat this entirely.
If your business card said "Growth Ninja," write the market title first and keep the internal one in parentheses:
Digital Marketing Manager (internal title: Growth Ninja)
Same principle for company names. Acme Corp (formerly Beta Systems) is worth the extra words if the acquisition is recent enough that recruiters still search the old name.
The contact line
Keep it in one place, in body text, one item per type:
- One email address. Two confuses the field mapping and the parser may pick the wrong one.
- One phone number, with a country code if you are applying internationally.
- LinkedIn as visible text (
linkedin.com/in/yourname), not a hyperlinked word like "LinkedIn." Many parsers read only the text, not the link target. - City and country. A full street address is unnecessary in 2026 and some employers strip it anyway.
- Month and year on every role.
- Plain hyphens in date ranges.
- Market-standard job titles.
- Visible URLs rather than hyperlinked words.
- One email, one phone, in normal body text.
Make keywords work without sounding robotic
Recruiters find candidates by searching the parsed database, usually with boolean queries built from the job description. Keyword matching is not a score you are trying to maximise. It is whether you appear in a list at all.
Pulling the right keywords
- Paste the full job description into a separate document.
- Mark repeated skills, tools and qualifications:
incident management,Python,stakeholder communication. - Note exact phrases, not just isolated words. Recruiters search phrases.
- Note the hard requirements separately. Certifications, clearances, licences and specific years of experience are often the actual filters.
Then map them to work you genuinely did:
- If the posting says
search engine optimization, use both the full phrase andSEO, each somewhere it makes sense. - If
incident resolutionappears four times in the posting, use that exact phrase in your incident-related bullets instead of a looser synonym. - Put the most important terms where they will be seen twice: once in the skills section, once inside a bullet that proves it.
- Mirror the posting's job title and key phrases where they are true for you.
- Spell out an acronym once, then use the short form.
- Place each important term in context, attached to something you did.
- Two or three well-placed mentions is plenty.
- No comma-separated keyword block at the bottom of the page.
- No white text, no zero-point font, no keywords hidden behind an image. Recruiters see the parsed text, where the trick is glaringly visible, and some systems flag it outright.
- No tools you cannot discuss for five minutes in an interview.
- No stuffing the same term ten times. It reads badly to a human and buys you nothing with the machine.
A note on where matching is heading. Newer platforms such as Eightfold and Beamery use embedding-based matching rather than literal string search, so a resume that says "Kubernetes" can surface for a query about "container orchestration." That trend genuinely reduces the value of exact-match gymnastics. It does not remove it, because the recruiter typing the search is still typing literal words, and the boolean search box has not gone anywhere.
Write bullets that both ATS and humans respect
Structure and keywords get you into the search results. Bullets are what happens next, when a person is actually looking.
A pattern that holds up:
Action verb + scope + method or tool + outcome
Reduced[action]mean incident resolution time by 35%[outcome]by introducing an on-call rotation and documenting runbooks in Confluence[method].
Weak:
- Responsible for managing incidents for our SaaS platform.
Stronger:
- Led incident response for a multi-region SaaS platform, cutting average resolution time from 90 to 55 minutes by introducing runbooks and tightening on-call escalation paths.The second version gives the parser concrete nouns to index and gives the recruiter a number to react to.
Three to six bullets per recent role. Fewer for older ones. If you cannot attach a number to a bullet, attach a scale instead: team size, region count, ticket volume, budget, or the name of the system you worked on. "Improved the deployment process" is invisible. "Improved deployment for a 40-engineer platform team" at least has edges.
One honest caveat about metrics. Do not manufacture percentages you cannot defend. An interviewer who asks "how did you measure that 35%?" and gets a blank stare has learned more about you than the bullet ever said.
Export the file correctly
A perfectly written resume can still parse badly because of how it was saved. This step takes two minutes and it is the one people skip.
Choose the format the posting asks for. If it specifies, that is the answer. If it does not, a text-based PDF is fine for essentially every current system, and .docx remains the safest choice if you have reason to think the employer is running an older platform.
Never send a scanned or image-based PDF. Open your file, try to select a line of text with the cursor, and copy it. If nothing selects, the entire document is a picture and the parser will extract nothing at all.
Watch out for ligatures. Some exporters, LaTeX and certain design tools in particular, render letter pairs like "fi" and "fl" as single glyphs. Copy your text out into a plain editor and check that "efficient" and "workflow" come out as real words rather than ef cient and work ow. If they do not, turn ligatures off before exporting.
Embed your fonts. An unembedded font can extract as substituted or garbled characters. Most modern export dialogs embed by default, but design tools sometimes do not.
Name the file properly. alex-rivera-backend-engineer.pdf beats resume_final_v7(2).pdf. Some systems display the filename in the recruiter's list, and it is free signal.
Open your resume, select all, copy, and paste into a plain text editor.
If the reading order is scrambled, sections are missing, words are broken, or your phone number has vanished, the ATS will see the same thing. Fix the document, not the wording.
The application form matters more than you think
Here is the part that surprises people who spend a weekend polishing a PDF. When an application is genuinely auto-rejected, it is almost always because of an answer on the form, not a score on the resume.
Knockout questions. "Are you legally authorised to work in this country?" "Do you have five or more years of experience with Java?" "Are you willing to work on site three days a week?" These have configurable disqualifying answers, and the rejection can be immediate and automatic. Read them carefully. Answering a years-of-experience question conservatively because you feel modest is a real and common way to be filtered out.
The parsed profile. Workday and several others turn your uploaded resume into an editable profile, then ask you to confirm it. That profile, not your PDF, becomes the record recruiters search. If the autofill has mangled your last three job titles and you click through without looking, you have just submitted the mangled version. Fix every field before you submit. It is tedious and it is the single highest-leverage five minutes in the whole process.
Screening questionnaires and assessments. Increasingly common at high-volume employers, and often weighted more heavily than the resume itself.
Duplicate applications. Most systems key on your email address. Applying to eight roles at one company with slightly different resumes creates one candidate record with a visible history, and recruiters can see all of it.
Adjust for your career stage
The parsing rules do not change. What you put in front of them does.
Early career
Move Skills and Education above Work Experience, and treat coursework, internships and personal projects as real experience with real bullets. A university capstone project written as Built a Django inventory tool used by 3 campus societies, cutting manual stock counts from weekly to none reads far better than a line in a "Relevant coursework" list.
Include graduation dates. Some early-career pipelines filter on them.
Career changers
Use the target field's vocabulary while telling the truth about where the work happened. Your bullets will come from adjacent roles, so lead each one with the transferable part.
Two lines is enough:
Operations lead moving into project management, bringing 6+ years coordinating cross-functional teams and improving on-time delivery.
Be aware that a title-based search will not find you. This is the one situation where getting a human referral is worth more than any amount of resume optimisation, because the search you need to appear in is one nobody is running.
Senior professionals
Screening still applies, but the read is different. Recruiters look for scope: team size, budget, headcount, regions, revenue, product surface. Put those numbers in.
Two practical points at this level. Compress roles older than fifteen years into a short "Earlier experience" list with titles and companies only. And check that your titles match what the market currently calls the job, because internal senior titles drift furthest from the taxonomy.
Sample ATS-friendly resume
Backend engineer with 7+ years building APIs and distributed services on AWS. Designed and scaled authentication, billing, and analytics systems for high-traffic SaaS products.
Python, FastAPI, PostgreSQL, Redis, AWS (Lambda, ECS, RDS), Docker, CI/CD, Observability (Prometheus, Grafana), Incident Management
- Designed and implemented a token-based authentication service on AWS, reducing login latency by 40% and cutting auth incidents by 50%.
- Led migration of a legacy monolith to containerized microservices on ECS, improving deployment frequency from monthly to multiple times per week.
- Partnered with product and support teams to build rate-limiting and abuse detection features, reducing API abuse-related downtime.
- Built RESTful APIs in Python and FastAPI to expose analytics features to B2B customers.
- Optimized PostgreSQL queries and indexing strategy, lowering average report generation time by 60%.
- Introduced structured logging and dashboards, improving mean time to detect incidents.
BSc, Computer Science - University of Porto, 2018
AWS Certified Developer - Associate, 2022
Notice what this sample does beyond the obvious. Every date has a month. Every date range uses a plain hyphen. The LinkedIn URL is visible text. The skills line spells out acronyms alongside product names. Certifications carry years. None of that is decorative, and all of it lands in a parsed field.
Common myths worth dropping
"The ATS scores my resume and rejects me below 75%." Match scores exist in some systems, and recruiters can sort by them, but automatic rejection is normally driven by application-form answers rather than a resume score. The scores you see quoted by resume-checking tools are those tools' own inventions, not numbers pulled from any employer's system.
"75% of resumes are never seen by a human." This statistic has circulated for over a decade with no traceable source behind it. The real problem is not resumes being deleted unseen, it is resumes that parse into records nobody searches for.
"You must use one page." Length is a recruiter preference, not a parser constraint. One page suits under ten years of experience. Two is normal above that, and academic CVs run longer by convention.
"Submit a plain .txt file to be safe." Overcorrection. It parses fine and reads terribly, and a human still opens it eventually.
"Keyword stuffing in white text works." It has not worked for years. The parsed text a recruiter sees includes it, plainly, and it reads exactly as dishonest as it is.
Final check before you apply
- Layout: single column, no tables, text boxes, or images.
- Contact: name, email, phone in body text at the top, not in a header.
- Sections: standard headings, normal letter spacing.
- Dates: month and year, plain hyphens, one consistent format.
- Titles: market-standard, with internal titles in parentheses if needed.
- Keywords: mirrored from the posting, placed in context, never hidden.
- Bullets: action, scope, method, outcome.
- File: text selectable, fonts embedded, no broken ligatures, sensible filename.
- Form: knockout questions answered accurately, parsed profile corrected before submitting.
Then read the whole thing out loud. If a sentence sounds like something you would never say to a colleague, rewrite it plainly. A resume that survives parsing and then bores the recruiter has only solved half the problem.
FAQ
What is an ATS-friendly resume in 2026?
A single-column document with standard section headings, simple embedded fonts, unambiguous dates, and keywords from the job description used in context. It parses cleanly into an employer's database and still reads well to the recruiter who opens it.
Does an ATS automatically reject my resume?
Usually not. Automatic rejection normally comes from disqualifying answers to application-form questions, such as work authorisation or a minimum years-of-experience threshold. The more common failure is invisible: your resume parses into an incomplete record, so you never appear in the recruiter's search results.
What file format is best for ATS resumes in 2026?
Use whatever the posting specifies. If it says nothing, a text-based PDF works with current systems, and .docx is the safer bet if you suspect an older platform. Never submit a scanned or image-based PDF, because it contains no extractable text.
Do graphics, icons, or tables break an ATS resume?
They can. Icons and skill bars carry no text. Tables and multi-column layouts can scramble reading order during extraction. Content inside headers, footers, and text boxes is frequently ignored altogether. Plain body text avoids all three risks at no cost.
How long should an ATS-friendly resume be?
Parsers do not penalise length. Recruiters do. One page for under ten years of experience, two pages for senior or deeply technical roles, longer for academic CVs. Every line should earn its place.
How do I test whether my resume will parse correctly?
Three checks, in increasing order of effort. Copy all the text into a plain text editor and confirm the order still makes sense and nothing is missing or broken. Upload the file to a free parser preview and read the fields it extracted. Then start an application on a real careers page that uses Workday and inspect the profile it autofills, which shows you exactly what one major system made of your document.
Should I tailor my resume for every application?
Tailor the summary, the skills section, and the top two or three bullets of your most recent role. That covers most of what a search will match on. Rewriting the entire document for every posting is a poor return on the time.
Do the newer AI matching systems make keywords irrelevant?
No, though they reduce the value of exact-match tricks. Platforms using embedding-based matching can connect related terms, so "Kubernetes" may surface for a "container orchestration" query. But recruiters still run literal boolean searches every day, and those still match strings. Write naturally, use the posting's actual vocabulary where it is true, and you satisfy both.
