What an ATS-friendly resume actually means in 2026

"ATS-friendly" means one thing: a piece of software can read your resume and fill in the right boxes with what it finds. That is the whole test. It is not a keyword score, it is not a hidden ranking, and there is no robot deciding you are unqualified before a person sees you.

The confusion is worth clearing up, because the advice that follows from the myth is bad advice. If you believe a machine is scoring you, you stuff keywords. If you understand that a parser is copying text into fields, you make the text easy to copy — and those are different resumes.

What the parser actually takes

When you upload a file, an applicant tracking system runs it through a parser that looks for a specific, short list of things. Greenhouse's standard parsing pulls out your name, email address, phone number, mailing address, current title, and current company. That is six fields from a two-page document.

Its Talent Matching feature — the part that lets a recruiter search a database of past applicants — extracts a second set: skills, job titles, years of experience, employment start and end dates, and company names. Greenhouse is explicit that it does not take more than that. The design is deliberate data minimisation, not a limitation they are apologising for.

So when someone tells you an ATS "reads your whole resume," it does not. It takes a handful of structured facts and stores the file. Everything else on the page exists for the human who opens it later.

What the parser stores, and what it does not

Greenhouse's two extraction paths. Anything outside these lists lives only in the attached file.

	Standard parsing	Talent Matching
Name	Yes	—
Email and phone	Yes	—
Mailing address	Yes	—
Current title and company	Yes	Yes
Full employment history	—	Company names, start and end dates
Skills	—	Yes
Years of experience	—	Yes
Your bullet points	Stored in the file, not as data	Stored in the file, not as data

This is why a Senior Accountant who writes "reconciled a $40M portfolio" in a bullet gets no credit for it in a recruiter's search. The sentence is in the file. It is not in the index. If reconciliation is a skill you want to be found for, it belongs somewhere the skills extractor will see it — a skills line, or the title-adjacent text near a role — not buried in the fourth bullet of a job from 2019.

The formatting choices that break a parse

Greenhouse publishes the list of things that cause what it calls an unsuccessful parse. It is short, specific, and almost entirely about layout:

A resume with spaces between the letters. Resumes that include graphics, photos, or word art. Resumes uploaded as an image rather than a document. Complex resumes with tables, headers, and footers. Resumes with the name and contact information in the header, footer, or text box. Resumes with a columned layout. Resumes without clear sections and differing formats throughout each section. And any file over 2.5MB, which the parser will not attempt at all.

Read that list again and notice what is missing. Nothing about keywords. Nothing about length. Nothing about fonts, or the word "utilised", or whether you used a serif. Every single item is a structural decision about where text sits on the page.

The two-column resume template is the most expensive design decision in a job search, and it is the one most commonly recommended. It looks organised to a person and reads as interleaved nonsense to a parser, which takes the page in reading order and produces a line that says "Senior Accountant Skills: Excel March 2021" — your job title welded to a skills heading. The header-and-footer problem is worse, because it fails silently in the one field that matters: your name and phone number end up somewhere the parser never looks, and the recruiter gets a candidate record with no way to contact them.

Here is what the recruiter sees

Take a Marketing Coordinator applying with a nicely designed two-column template, name in a header, a small photo top-left. Before:

Name: (blank) Email: (blank) Current title: Marketing Coordinator Education B.A. Communications Company: (blank)

The same person, same words, in a single-column document with the contact details in the body:

Name: Priya Raman Email: priya.raman@example.com Current title: Marketing Coordinator Company: Brightline Media

Nothing about the candidate changed. The second one is simply legible to the thing doing the reading.

Does a human ever open it?

Yes, and it is worth being clear about this, because "the ATS rejected me" has become a way to explain silence that is usually wrong.

The people on the other side of the form

939,700HR specialists employed in 2025

$75,940Median pay, May 2025

6%Projected growth, 2025–35

The Bureau of Labor Statistics counts 939,700 human resources specialists, with a median wage of $75,940 a year as of May 2025, and describes their work plainly: they "recruit, screen, and interview job applicants and place newly hired workers in jobs." That is a lot of people whose job is to read applications.

What the parser controls is not whether you are seen, but what you look like in the list they are scrolling. A record with a blank name and a mangled title is the one that gets skipped — not because software rejected it, but because a person with sixty of these to get through picked the ones they could read.

So what about keywords?

I said at the top that no machine scores you, and that is true. But "tailor your resume to the job description" survives as advice because there is a real mechanism underneath it — just not the one people describe.

The mechanism is search. A recruiter with a req to fill types into a box: senior accountant, revenue recognition, NetSuite. What comes back is candidates whose extracted fields match. Greenhouse's Talent Matching indexes skills, job titles, years of experience, employment dates and company names — so the words that decide whether you appear are the words in those fields, not the words anywhere on the page.

That distinction changes what you do. Stuffing a job description's phrases into a bullet at the bottom of page two does nothing, because bullets are not indexed as data. Making sure the tool you actually use appears as a skill, spelled the way the industry spells it, does something. So does using the title employers post rather than the one your company invented: if your business card says "Growth Ninja" and the market says "Demand Generation Manager", the parser will faithfully record Growth Ninja and you will not be in the results for the job you want.

There is a second-order version of this that matters for anyone with a long history. Talent Matching extracts years of experience from employment start and end dates. If a role on your resume has no end date, or a date format the parser cannot read — "Spring 2021", "'19–'22" — that role may contribute nothing to the total. A Senior Accountant with eleven years of experience can show up in a search for five-plus and be invisible in a search for ten-plus, entirely because of how three dates were typed. Use unambiguous months and years, and write "Present" for the current one.

None of this is a trick, and it is worth saying that plainly: none of it makes you a better candidate for a job you cannot do. What it does is stop a filing system from misfiling you.

The seven checks, in order

1. Open your resume in a plain text editor. Copy the whole document, paste it into Notepad or TextEdit. If your name is missing, it is in a header. If the lines interleave, you have columns. This takes thirty seconds and finds most problems.
2. Move your contact details into the body. First line of the document, plain text, no text box.
3. Delete the photo. It is not customary on a US resume, and it is the usual reason a file passes 2.5MB.
4. Collapse to one column. Keep the design in the type and the spacing, not the grid.
5. Use real headings. "Experience", "Education", "Skills" — consistently formatted, in that structure. Greenhouse names "resumes without clear sections" as a failure cause.
6. Put skills where the skills extractor looks. A short, plain skills line beats a skill implied in a bullet.
7. Export from the document, never from a scan. A PDF from Word or Google Docs is text. A photographed page is a picture of text.

What to do this week

Run the plain-text test on the resume you are sending right now. Whatever it produces is roughly what the recruiter's candidate record will look like — if that record has a blank field, fix that field before you send another application.

Then send the same file to one more job and stop redesigning it. The single-column, clearly-sectioned, no-header resume is not a style. It is a format that survives being read by software, and once yours does, the interesting question moves back to where it belongs: whether the work you describe matches the job.

If you would rather not retype the same six fields into every form, HeroApply reads your resume once and fills them in for you — including on the systems that would otherwise have got them wrong.