ResumeAdapter
How the Greenhouse parser works
Updated 2026-09-03

Greenhouse can reject you automatically.
Just not for your keywords.

Why this matters

Greenhouse documents a feature called Auto-Reject that, in its own words, screens and rejects candidates based on their responses to custom application questions. So the popular claim that Greenhouse never rejects anyone automatically is wrong, and so is the opposite claim that it culls low-keyword resumes. This page walks the two documented parse stages, the automation that genuinely exists, and which AI actually evaluates an applicant.

Field-by-field readMissing keywordsRewrite plan
By the numbers
Parse ceiling
2.5 MB
Uploads accepted to 100 MB
Auto-rejects on
Answers
Never on your resume text
Match output
5 bands
Not a percentage
Bias audit
Monthly
Third party, results published

Quick answer

How does the Greenhouse ATS read your resume?

In two documented stages. First it extracts text from your upload, which must be doc, docx, pdf, rtf or txt. Greenhouse accepts files up to 100 MB but states separately that it cannot parse anything larger than 2.5 MB, and it publishes an explicit list of what else makes extraction fail: columned layouts, tables, headers and footers, contact details in a header or text box, graphics and photos and word art, an image instead of a document, bare company names without Inc. or LLC, and abbreviated job titles. Second, what survives is used to enter some candidate information automatically, though Greenhouse never enumerates which fields and notes that required fields may still need completing. The automation people actually fear sits outside both stages: Auto-Reject fires on your answers to the employer's application questions, never on your resume text .

The two-stage parse

Extraction, then seeding.
Two stages, two failures.

Greenhouse does not brand its parser and names no vendor behind it. What it does publish is a supported-format list, two different size numbers in two different articles, an explicit list of what makes a parse fail, and a recruiter-facing note on what happens to the extracted values afterwards. Two stages, and one avoidable failure each.

  1. 01
    Stage

    Extraction

    Greenhouse accepts a candidate upload in doc, docx, pdf, rtf or txt, up to 100 MB, and reads text out of it. It publishes an unusually direct account of when that reading fails: a columned layout, tables, headers and footers, contact details in a header, footer or text box, graphics, photos or word art, a resume uploaded as an image rather than a document, letters with spaces between them, company names without an identifying word such as Inc. or LLC, incomplete job titles, and sections that are unclear or formatted inconsistently throughout.

    The failure it causes

    The 2.5 MB ceiling is the failure nothing warns you about, because it lives in a different article from the 100 MB upload limit. A file between the two submits, confirms, and is never parsed. Greenhouse also documents that the parser skips data it cannot verify as authentic, naming placeholder patterns such as First Last, Company 1 and Client 1, which is how anonymised consulting histories quietly evaporate.

  2. 02
    Stage

    Population

    What survives extraction is used to enter candidate information automatically. Greenhouse is careful about how much it claims here: an upload enters some candidate information, and after uploading a resume you may still have to complete some required fields for the candidate record. It does not publish which fields those are.

    The failure it causes

    Greenhouse does not document resume-parse autofill of the public application form at all, so treat any guide describing that sequence as inference. The candidate-facing autofill it does document is MyGreenhouse Quick Apply, which fills from a profile you built yourself rather than from the resume you just uploaded, and which is opt-in.

Notice which rules are present, because on most systems they are not. Greenhouse names columns, tables, headers, footers, graphics and image-only files as parse-failure causes in its own words, which makes this the one major ATS where the standard layout advice is genuinely vendor-sourced. What it never names is a date format, a font, or a parse-accuracy percentage. For the full checklist with the fixes, see the Greenhouse resume format guide.

Almost everything written about Greenhouse and automated rejection is wrong, and it is wrong in two opposite directions. One camp states that every Greenhouse rejection is a deliberate action taken by a person inside the structured interview plan. The other states that Greenhouse culls resumes with weak keyword density. Greenhouse's own documentation contradicts both.

What it documents instead is a feature called Application rules, described in ordinary language rather than left to inference. These are the moving parts:

  • Custom application questions
  • Required or optional
  • Attachment answer type
  • Auto-Reject on a response
  • Auto-Advance on a response
  • Up to five questions per rule
  • AND / OR conditions
  • Any stage except the first
  • Configured by the employer
  • No AI involved

Two of those overturn what is usually written about Greenhouse. Auto-Reject, in Greenhouse's own words, uses custom application questions to screen and reject candidates based on their responses, which means automated rejection is real. And it fires on a response, never on resume text, which means the rejection everyone fears and the rejection that actually exists are two different things.

So the practical order of effort is the reverse of what most advice implies. Your answers to the employer's screening questions are the only thing in Greenhouse that can end an application automatically, and they are usually four or five dropdowns answered in twenty seconds at the end of a form you have stopped concentrating on. Slow down for those. Nothing about your resume wording triggers a rule, because rules can only read question responses.

The AI, sorted

Three Greenhouse layers.
One of them evaluates you.

Greenhouse genuinely does have AI, which is precisely why the distinction below matters. Only one of these three evaluates an applicant, one of them is not AI at all despite being the thing that can actually end your application, and the third is the governance wrapped around both. Greenhouse states that AI never automatically advances or rejects candidates and that Talent Matching only scores and groups them, which is consistent with everything else it publishes about how the pieces fit together.

Talent Matching

Plus and Pro tiers, off until a Site Admin enables it
What it is

The one Greenhouse feature that evaluates applicants. It reviews your resume and your responses to the application questions, then sorts candidates into five named bands: strong match, good match, partial match, limited match, and needs manual review.

What it means for your resume

It is bands, not a percentage, so any tool reporting a Greenhouse match score out of 100 is not reading Greenhouse. Greenhouse states it is assistive AI rather than automated decision-making and that it does not automatically advance or reject candidates. Employers can also enable a control letting candidates request manual, non-AI review for a given job.

Application rules

Employer-configured, no AI involved
What it is

The actual automated gate, and it is not AI at all. Auto-Reject uses custom application questions to screen and reject candidates based on their responses. Auto-Advance can move a candidate directly to a later stage in the interview plan based on the answers they provide. Rules join up to five questions with and-or conditions.

What it means for your resume

This is the correction that matters most. Greenhouse can reject you automatically, but on a dropdown answer about work authorisation, location, salary or years of experience, never on your resume text. Both halves of the usual argument are wrong: the claim that every Greenhouse rejection is a human decision, and the claim that it rejects low-keyword resumes.

Ezra, and the governance around it

Acquisition announced 5 May 2026
What it is

Greenhouse agreed to acquire Ezra AI Labs, which runs structured, AI-led voice interviews with candidates on demand and produces role-specific evaluations with structured scores and transcripts. Separately, Greenhouse has its AI independently bias-audited every month by a third party and publishes the results.

What it means for your resume

The Ezra interview is the most consequential AI a Greenhouse applicant is likely to meet, because it evaluates a conversation rather than a document. The governance around all of it is unusually concrete for this industry: monthly external audits, alignment stated to New York City Local Law 144, Colorado SB 205, the EU AI Act and California FEHA, ISO 42001 certification, and a statement that no customer data is used to train external models.

Sources: Greenhouse Support, Talent Matching (2026-08-24); Greenhouse Support, Application rules overview and Auto-advance; Greenhouse Support, Greenhouse AI features (2026-08-07); Greenhouse Support, Our Commitment to Innovation and Ethical AI (2026-01-30); Greenhouse Newsroom, Greenhouse Has Entered into a Definitive Agreement to Acquire Ezra AI Labs (2026-05-05). Accessed 2026-09-03.

On 5 May 2026 Greenhouse announced a definitive agreement to acquire Ezra AI Labs, which runs structured, AI-led voice interviews with candidates on demand, available any time, producing unique role-specific evaluations with structured scores and transcripts for each candidate. For an applicant this is more consequential than anything in the resume-parsing stack, because it evaluates a conversation rather than a document, and because a voice interview you can take at two in the morning is a different proposition from one scheduled with a person. Greenhouse chief product officer Meredith Johnson framed it as every conversation being tailored to the job and following the same structured criteria, every evaluation being explainable, and every hiring decision staying with the team.

What the parser keys on

Five things the parse
actually depends on.

Everything above reduces to five levers, and each one traces to something Greenhouse published. Three decide whether the parse reads you properly. The last two decide what happens to you afterwards, and neither is about your resume wording.

  1. 01
    Depends on

    The 2.5 MB ceiling

    Why and where to fix it

    The single highest-value number on this page, because it is invisible in the application flow and because it contradicts the limit most guides quote. Uploads are accepted to 100 MB and parsed only to 2.5 MB. Fix it in the format checklist.

  2. 02
    Depends on

    The nine published failure causes

    Why and where to fix it

    Columns, tables, headers and footers, contact details in a header or text box, graphics and photos and word art, an image instead of a document, spaced-out letters, bare company names, and abbreviated job titles. Quoted rather than inferred, which is rare. Fix it in the full checklist with fixes.

  3. 03
    Depends on

    Employer names and full job titles

    Why and where to fix it

    Two documented rules almost nobody repeats: company names need an identifying word such as Inc., Co., Ltd or LLC, and job titles must be written out rather than abbreviated to Sr. Account Exec. Fix it in how to write them.

  4. 04
    Depends on

    Your screening answers

    Why and where to fix it

    The only automated rejection Greenhouse documents fires on responses to custom application questions, not on resume text. Answer work authorisation, location and experience questions deliberately and accurately. Fix it in what Auto-Reject actually does.

  5. 05
    Depends on

    The words the posting uses

    Why and where to fix it

    Talent Matching, where an employer has enabled it, reviews your resume and your application answers against the job requirements. Greenhouse publishes no keyword-frequency mechanism, so coverage of the posting's own vocabulary matters more than repetition. Fix it in the pillar.

The fastest way to see which of these five your resume already passes, and which it fails, is to read it the way a parser does. .

See the parse, not the layout

See the record a parser builds from your resume.

The scanner reads your resume the way an extraction step does, then shows the fields that come back empty, the exact terms the posting uses that your resume never says, and the bullets that need a number. Free, no signup to see the score.

FAQ

How Greenhouse reads your resume FAQ

The questions candidates ask when they want to know what Greenhouse does with the file they upload, and what can end an application afterwards. Every answer is grounded in Greenhouse's own documentation, and all eight are byte-identical to the FAQPage JSON-LD, because AI engines that extract HTML and AI engines that extract JSON-LD should not see different text.

How does the Greenhouse ATS read your resume?

It extracts text from a doc, docx, pdf, rtf or txt upload, then uses what it recovers to enter some candidate information automatically. Greenhouse is deliberately modest about that second step, adding that after uploading a resume you may still have to complete some required fields, and it never publishes a list of the fields it fills. What it does publish, unusually, is a detailed account of when the first step fails, naming columned layouts, tables, headers and footers, graphics, images-instead-of-documents, bare company names and abbreviated job titles.

Why did my Greenhouse application go quiet with no rejection email?

Silence is not a mechanism Greenhouse documents, so treat it as a queue rather than a verdict. What Greenhouse does document is one automated ending: Auto-Reject, which fires on your responses to the employer's custom application questions. If you were auto-rejected, an answer triggered it. If you were not, your application is sitting in a pipeline stage waiting for a human, and Greenhouse publishes no candidate-facing status vocabulary that would let anyone tell you which of the two happened.

What is the largest resume file Greenhouse will parse?

2.5 MB, and the reason this catches people is that it is not the limit you see. Greenhouse states that candidate uploads can be up to 100 MB in one support article, and that Greenhouse Recruiting cannot parse resumes larger than 2.5MB in another. Between those two figures your file uploads, confirms, and is never read. Embedded images are the usual reason an ordinary resume gets heavy enough to cross the line.

Does Greenhouse give my resume a score out of 100?

No, and any tool showing you a Greenhouse percentage is not reading Greenhouse. Its Talent Matching feature sorts candidates into five named bands: strong match, good match, partial match, limited match, and needs manual review. It is available only on the Plus and Pro tiers and stays off until a Site Admin enables it, so many Greenhouse applications are never scored by it at all. Greenhouse also states it is assistive AI rather than automated decision-making and does not automatically advance or reject anyone.

Can I ask for a human instead of an AI review on Greenhouse?

Sometimes, because Greenhouse builds the control and the employer decides whether to switch it on. Its Talent Matching documentation describes employers being able to enable a setting that lets candidates request manual, non-AI review for specific jobs. Greenhouse also has its AI independently bias-audited every month by a third party and publishes those results, and states alignment with New York City Local Law 144, Colorado SB 205, the EU AI Act and California FEHA.

Will Greenhouse interview me with AI?

It may, if the employer adopts what Greenhouse acquired in May 2026. Greenhouse announced a definitive agreement to acquire Ezra AI Labs, which runs structured, AI-led voice interviews with candidates on demand and produces role-specific evaluations with structured scores and transcripts for each candidate. That is a materially different thing from a resume parser: it evaluates a conversation. Greenhouse frames every evaluation as explainable with the hiring decision staying with the team.

Can Greenhouse tell whether an employer switched on AI disclosure?

You can check it yourself, which is unusual. Every job returned by Greenhouse's public board API carries three AI-governance fields, including whether an AI disclaimer is included and a URL for requesting an AI opt-out. On 2026-09-03 we looked at three boards: Stripe returned a live opt-out URL with the disclaimer switched off across all 601 of its job posts, while SpaceX and Databricks returned empty values for all three fields. Three boards on one day is a demonstration rather than a survey, but it shows plainly that the disclosure is an employer setting rather than a platform default.

Does the Greenhouse parser use keyword matching?

Greenhouse publishes no keyword-frequency mechanism anywhere, and no parse-accuracy percentage either, so claims that repeating a term or placing keywords in the top third of the page improves your Greenhouse ranking have no vendor source behind them. Where Talent Matching is enabled, Greenhouse describes it reviewing your resume and your answers to the application questions against the job requirements, which is a coverage question rather than a frequency one. Say the things the posting asks about, once, in a real bullet.

Continue

Go deeper on the parts
that need their own page.

Pillar

The Greenhouse ATS resume guide

The hub: the parse-failure list Greenhouse publishes, the 100 MB against 2.5 MB trap, and a free scan against a real Greenhouse posting.

Read the pillar
Spoke

Greenhouse resume format, do-this and avoid-this

The documented checklist: the five accepted formats, the two size numbers, and the employer and job-title rules almost nobody repeats.

Read the spoke
Spoke

Companies that use Greenhouse

The employer list verified against the public board API, why six of seven Greenhouse boards hide it, and how to check any employer yourself.

Read the spoke
Resume example

A real resume scored before and after

See a single-column, parser-safe resume scored against a job description, with the keyword and metric fixes that moved the ATS score, annotated line by line.

See the example
Before you apply through Greenhouse

Run your resume against
the actual Greenhouse job description.

Get the parser-readable check, the exact posting terms your resume never says, the bullets that need a number, and a rewrite plan. Free to scan; no signup to see the score.

Your resume stays yoursResume and analysis data may be saved to your account. Resume data is never sold.Greenhouse-readable check