Make your resume ATS-friendly without chasing a mystery score

What ATS-friendly resume format actually means

Almost every list of format rules on the internet is unsourced, and some of it is invented. So each rule below is labelled with where it comes from: what a named vendor documents in public, what we recommend without a vendor behind it, and what nobody has actually published.

A rule that applies to one system is a fact about that system, not a law of nature. We have kept those apart deliberately, including in the cases where admitting we do not know makes the list less satisfying.

Vendor documentation read and verified 2026-08-23.

• Columns

  Vendor-documented: Greenhouse

  Greenhouse lists "resumes that have a columned layout" among the causes of an unsuccessful resume parse.

  Use one column. A sidebar looks organised to a human and flattens into interleaved nonsense when a parser reads the page as a single stream.

  Greenhouse Support — Unsuccessful resume parse

• Tables

  Vendor-documented: Greenhouse

  Greenhouse names "complex resumes with tables, headers, and footers" as a parsing failure cause.

  Do not use a table to lay anything out, including the contact block and skills grids, which are where tables usually hide.

  Greenhouse Support — Unsuccessful resume parse

• Headers and footers

  Vendor-documented: Greenhouse

  Greenhouse specifically calls out "resumes with the name and contact information in the header, footer, or text box".

  Keep your name, email, phone and links in the body of the document. This is the single highest-consequence item on the list: if it fails, the employer has your resume and no way to contact you.

  Greenhouse Support — Unsuccessful resume parse

• Text boxes

  Vendor-documented: Greenhouse

  Greenhouse names text boxes alongside headers and footers as a place contact details get lost.

  Delete every text box. Text inside one is not part of the normal reading order and can be skipped entirely.

  Greenhouse Support — Unsuccessful resume parse

• Images, photos, icons and word art

  Vendor-documented: Greenhouse and Workday

  Greenhouse lists "resumes that include graphics, photos, or word art". Workday separately advises that for best results you should "use resumes that don't have images or image-based styles".

  Two vendors say the same thing, which is about as close to a universal rule as this subject gets. No headshot, no logos, and never an icon standing in for a word: an envelope glyph in place of the word Email carries nothing once the text is extracted.

  Greenhouse Support — Unsuccessful resume parse

• A resume saved as an image

  Vendor-documented: Greenhouse

  Greenhouse lists "resumes that are uploaded as an image, rather than a document (such as a .docx or .pdf)".

  Never send a scan, a screenshot, or an image-only PDF. There is no text in the file to recover, so extraction returns nothing at all. This is the one failure that is total rather than partial.

  Greenhouse Support — Unsuccessful resume parse

• File size

  Vendor-documented: Greenhouse

  Greenhouse states plainly: "Greenhouse Recruiting can't parse resumes larger than 2.5MB," and gives high-quality images as a common cause of exceeding it.

  Stay well under 2.5 MB. A text-only resume is normally well under 200 KB, so a large file is usually a symptom of embedded images you should not have anyway. Note this is Greenhouse's documented number and not a universal limit.

  Greenhouse Support — Unsuccessful resume parse

• Section headings

  Vendor-documented: Greenhouse

  Greenhouse names "resumes without clear sections and differing formats throughout" as a parsing failure cause.

  Use conventional headings and keep them consistent: Professional Summary, Experience, Skills, Education. Inventive headings such as Where I Have Made A Dent are a real cost for no benefit.

  Greenhouse Support — Unsuccessful resume parse

• Job titles

  Vendor-documented: Greenhouse

  Greenhouse lists "resumes with incomplete job titles".

  Write the full title. If your internal title was unhelpful, use the recognisable equivalent rather than an abbreviation or an internal code name.

  Greenhouse Support — Unsuccessful resume parse

• Company names

  Vendor-documented: Greenhouse

  Greenhouse names "company names that don't include identifying words such as Inc., Co., LTD, or LLC" as a cause of employers being skipped.

  Include the identifying suffix where the company has one. This is the most obscure item on the list and costs nothing to fix.

  Greenhouse Support — Unsuccessful resume parse

• Letter spacing

  Vendor-documented: Greenhouse

  Greenhouse lists "a resume with spaces between the letters".

  Never space out letters to stretch a heading. It turns one word into several tokens and destroys the word for anything reading it.

  Greenhouse Support — Unsuccessful resume parse

• Word order and general formatting

  Vendor-documented: Workday

  Workday states that resume parsing results "can vary based on resume format and order of words".

  Worth quoting whenever someone tells you formatting does not matter. It is the vendor saying, in its own admin documentation, that the same content in a different arrangement produces a different extraction.

  Workday Documentation — Concept: Resume Parsing

• Fields that will not be auto-filled

  Vendor-documented: Workday

  Workday states: "Workday doesn't auto-fill fields you configure as hidden or these fields: Languages, Skills."

  Do not assume a skills list carries your application by itself in Workday. Evidence of the skill inside your experience bullets does work that a keyword row cannot.

  Workday Documentation — Concept: Resume Parsing

• Dates

  RoleSprint recommendation

  We have found no first-party vendor documentation prescribing a date format. What our own extraction pass repeatedly shows is that dates sitting in a separate column, or written unusually, stop being attached to the role above them.

  Put the date range on its own line directly under the role and company, and write it plainly, for example 2022 – Present. Then check the timeline actually came out attached to the right jobs.

• Fonts

  Unknown / platform-dependent

  The specific fonts named endlessly in resume advice are not, as far as we can find, specified in Workday's or Greenhouse's public documentation. What the vendors document is images, layout and structure, not typeface choice.

  Use an ordinary system font because it is a safe default, not because a vendor requires it. Anyone presenting a font list as an ATS rule is repeating folklore, and the effort is better spent on the documented items above.

• Resume length

  Unknown / platform-dependent

  No first-party parsing documentation we have found sets a page limit. Length is a human preference and varies by field, seniority and country.

  Decide length on whether every line earns its place. Do not shrink a resume to satisfy a rule nobody has published.

• Whether any of this gets you rejected

  Unknown / platform-dependent

  Whether an employer auto-rejects, and on what, is configured per employer and per job. It is invisible from outside, including to us.

  Fix what is documented and testable, then stop. The remaining anxiety has no object, and the strongest filters are usually knockout questions on the application form rather than anything about your resume's layout.

PDF or DOCX? The honest answer

There is no universal rule that PDF always wins or that DOCX always wins across every hiring system, and we are not going to invent one to sound decisive.

What the documentation actually shows is that both are treated as ordinary documents. Greenhouse's list of parsing failures does not name PDF or DOCX as a problem: the entry is about resumes uploaded as an image "rather than a document (such as a .docx or .pdf)", which puts the two formats on the same side of the line. The real division is not PDF against DOCX. It is a file with recoverable text against a file without one.

• A PDF exported from a word processor normally contains real selectable text and extracts fine.
• A PDF produced by scanning or photographing a page contains an image of text and extracts nothing at all. This is the failure that actually matters, and it is invisible until someone tries.
• A DOCX built out of tables, text boxes or a header block can parse worse than a clean PDF, because the documented failure causes are structural rather than about the container.
• If a posting explicitly asks for one format, send that one. An instruction from the employer beats any general rule, including ours.

So the useful question is not which extension is safer in the abstract. It is whether the specific file you are about to send survives extraction, which takes about fifteen seconds to check.

What an ATS-friendly template should look like

We built one to the documented rules above and then ran it through our own diagnostic. It is free, there is no signup, and there is no email box in front of it.

It looks plain on purpose. Every visual flourish that makes a resume template look impressive in a gallery is on somebody's documented list of parsing failure causes, which is a bad trade for a document whose only job is to survive being read by a machine and then by a tired human.

• One column, and no table anywhere in the file, including the contact block where tables usually hide.
• No header, no footer, no text box. Your name and contact details sit in the body of the document, one per line, because that is the failure Greenhouse calls out by name.
• No image, no icon, no logo, no word art. Nothing carries meaning that would vanish when the text is extracted.
• Conventional section headings in the wording parsers look for: Professional Summary, Experience, Skills, Education, Certifications.
• Real years in the example date ranges rather than bracketed placeholders, because a placeholder teaches you nothing about the format that actually matters.
• Company placeholders written as "[Company Name, Inc.]", because Greenhouse documents that a missing Inc. or LLC can cause an employer to be skipped.
• Bullet patterns that force evidence: what you did, at what scope, and what changed. The template will not write your achievements for you and does not pretend to.
• About 10 KB, which is roughly one 250th of the 2.5 MB limit Greenhouse documents.

Why an ATS score is not an employer decision

The category collapses four different questions into one percentage, and the averaging is where the information dies.

Readable is not the same as relevant

Passing a parser test only tells you the document can be read. It does not tell you whether the experience inside it makes you competitive for this job.

This is where we think the category has it backwards, and where we will argue plainly for our own method. Keyword overlap can be raised by anyone in an afternoon, which is exactly why it predicts so little — a resume can contain every phrase in the posting and still be a weak application, and a resume with lower overlap can carry far stronger evidence. RoleSprint separates machine readability from job evidence instead of hiding both inside one opaque score, and then goes further than any parser can: it reads the real posting against the real evidence in your resume and tells you whether the role is worth your effort before you spend it. That last decision is the economically significant one, and it is the one an ATS percentage is standing in for, badly.

What nobody can tell you from outside

We would rather be useful about the limits than confident about things we cannot see.

• Whether a specific employer auto-rejects anyone. Screening rules are configured per employer and per job. The software makes it possible; the employer decides whether to use it.
• What knockout questions a posting uses. Those are application-form questions the employer writes — work authorisation, location, licences — and they filter far more decisively than resume wording. They are also invisible until you reach the form.
• Which parser an employer runs, or how it is tuned. Different systems and settings extract the same file differently, which is why we describe our result as one honest sample rather than a verdict.
• Whether a human will read it. That depends on volume, recruiter workload, and process, none of which is a property of your document.
• Whether a particular employer uses AI matching at all. Greenhouse, for instance, documents its Talent Matching as a paid add-on the employer enables, calibrated with criteria and weights the hiring team sets themselves — and states plainly that it does not automatically advance or reject candidates. None of that configuration is visible from outside, which is exactly why an external match percentage cannot mean anything.
• What a status label means at a specific company. Statuses are employer-configured, so the widely-published tables of what each one supposedly means are describing something that is not fixed. We write up what the vendors themselves document instead, for Workday and for Greenhouse.

How this test works

Everything on this page is mechanically derived from your file, with no model involved, so you can check any of it yourself.

• Your file is uploaded to our server, its text is extracted in memory with the same libraries the product uses elsewhere (unpdf for PDF, mammoth for DOCX), and the file is discarded. It is never written to storage and never logged.
• Field detection is pattern matching over the extracted text: an email, a phone-shaped string, a link, and date ranges in the common formats. The name line is explicitly a guess from position and shape, and is labelled as one.
• Format risks are fingerprints, not layout inspection. We do not have your document's internal structure, so we report what a layout does to extracted text — heavy line fragmentation, column-like spacing, characters that failed to decode — and say what each pattern usually means.
• The two readings we show, machine readability and structure risk, are OUR coarse bands over those findings. They are not an employer's ATS score and are never combined into one number.
• With a job description supplied, the requirement language check is literal word presence and nothing more. Finding a phrase proves the phrase is present, not that you meet the requirement. It is shown unrolled on purpose.
• This is RoleSprint's own diagnostic. It is not Workday's parser, Greenhouse's parser, Lever's parser, or whatever your employer has configured, and we do not have access to any of those. That is not a weakness to apologise for — it is the reason we only claim what this diagnostic can actually observe, while a percentage implies knowledge of a configuration nobody outside the company has.
• PDF extraction often returns text with no line breaks at all. Where that happens we say so and skip the checks that read line shape, rather than reporting them as passed.

For what the vendors themselves publish, we have written up Workday and Greenhouse from their own documentation, with sources cited.

Questions about ATS and resumes

What makes a resume ATS-friendly?

Text that can actually be extracted, and a structure that survives extraction. In practice: real selectable text rather than an image or a scan, a layout that does not scramble when flattened into a single stream, standard section headings, and date ranges written so they stay attached to the right role. Those are testable properties of your file. Everything beyond that is the employer's configuration, not your document.

Does an ATS automatically reject resumes?

It can, but whether it does is the employer's decision, not the software's default. The filters that reject most decisively are usually knockout questions on the application form — work authorisation, location, licences — rather than anything about resume wording. Nobody outside the company can tell you how a given posting is configured, and any tool that claims to is guessing.

Is an ATS score worth anything?

Not much on its own, and it is not an employer's decision. A score generally blends how readable your document is with how many words it shares with the posting, and those are different things with different levels of certainty. Keyword overlap in particular is easy to raise and predicts little: a resume can echo every phrase in the posting and still be a weak application. We show the readability and the literal word overlap separately, labelled, rather than averaging them into a number that looks precise and is not.

Should I add keywords from the job description?

Adding words you cannot back up is a bad trade: it may raise a keyword count, and it makes the interview harder because you now have to substantiate claims your experience does not support. If a requirement is genuinely part of your history and simply is not written down, adding it is worth doing. If it is not part of your history, the useful question is whether this role is the right target at all.

Does this tool store my resume?

No. The file is read in memory to extract its text and then discarded. It is not written to storage, not logged, and nothing from it is sent to analytics — the events we record contain only coarse bands such as whether readability came out strong, mixed or weak.

Why does this not give me a single ATS score?

Because the number would be false precision. Different employers run different systems with different settings, and the strongest filters are configured per job and invisible from outside. A single percentage implies a universal answer to a question that has no universal answer. Showing the separate, checkable findings is less satisfying and considerably more useful.