# The Resume Parse Path, Field by Field, From Upload to Recruiter

*You will be able to point at any resume element and state what parsing does to it, what that proves, and whether it is worth changing.*

- Canonical URL: https://www.refolk.ai/candidates/guides/resume-parse-path-field-by-field
- Pillar: Positioning and materials
- Format: Reference
- Published: 2026-08-02
- Last reviewed: 2026-08-02
- Reading time: 16 min

This guide explains what hiring software does to each part of your resume between the moment you upload it and the moment a recruiter reads it. It is for job seekers who want a lookup, not a lecture: jump to the row for dates, or two-column layouts, or file type, and read what parsing does, what that proves, and whether it is worth changing. The aim is to fix the two things that break a parse and stop rewriting the eight that were fine.

The pages ranking on this topic are mostly fear tactics or scanner upsells that fuse two different claims: that nearly every employer uses an ATS, and that a bot rejects most resumes. The first is true and verifiable. The second is folklore. Once you separate them, the work gets small and concrete.

## What happens to a resume between upload and a recruiter

An applicant tracking system converts your file into a searchable structured record, then lets recruiters search and rank that record; it rarely rejects anything on its own. The document you designed becomes a database row, and everything in this guide is about which parts of your design survive that conversion intact.

The path runs in a fixed order. Each stage can pass your resume through cleanly or drop a field, and the stages where things break are not the ones the scare articles emphasize.

#### The resume parse path

1. **Text extraction** - The file is converted to a raw text stream; an image-based PDF yields nothing here.
2. **Reading-order reconstruction** - The parser decides the sequence of text; columns and tables scramble here.
3. **Segmentation** - Blocks are labelled Contact, Experience, Education, Skills using header keywords.
4. **Field parsing** - Blocks split into title, company, start date, end date, description.
5. **Index and rank** - Fields become a searchable record; the recruiter works a sorted list top-down.

*Every application moves through these stages in order, and the two where most candidates lose ground are reading-order and dates, not file type.*

The single most useful fact about this pipeline is that step five is a search engine, not a judge. As Jobscan puts it, an ATS "acts as a search engine for recruiters to filter and rank applicants based on keywords and knockout requirements." Nobody at the bottom of the list was rejected. A person simply never scrolled that far.

## The field-by-field parse table

Here is the lookup. Each row names a resume element, states what parsing does to it, and says whether it is worth your time to change.

| Element | What parsing does | Worth changing? |
| Single-column body text | Reads cleanly in order | No, leave it |
| Two-column layout | Flattens into one stream; text lands out of order | Yes, collapse to one column |
| Layout tables | Often scrambled or read row-then-column wrong | Yes, remove |
| Text boxes | Frequently treated as graphics and ignored | Yes, move content into the body |
| Headers and footers | Commonly skipped entirely | Yes, put contact details in the body |
| Images and icons | Not read as text at all | No effect if decorative; move any real text out |
| Fonts, bold, bullet glyphs | Parse fine regardless of style | No, leave them |
| Dates in MM/YYYY or Month Year | Parsed as start and end fields | Only if your format differs |

Two rows carry the weight: multi-column layouts and content trapped in text boxes, headers, or images. Everything a recruiter warns you about that is not on this "yes" list - the font choice, the bullet character, whether you bolded a company name - parses fine. The over-correction failure mode is real: people rewrite eight cosmetic things and miss the one that actually breaks.

> **Rule:** The paste test is the whole diagnostic
>
> Copy your resume from the exact file you will submit and paste it into Notepad or any plain text box. If the words come out readable and in the right order, the software can read them too. If they scramble, your columns or tables are the problem.

### Why columns break and text boxes vanish

The parser does not see your columns the way your eyes do. It grabs text in the order the file stores it, so a two-column layout can drop a skill into the middle of a job title. Text boxes are a separate failure: many systems treat them as graphical elements rather than text containers, so a phone number or a certification sitting in a sidebar box can be omitted with no error and no warning. That silent omission is worse than a scramble, because a scramble is visible in the paste test and a missing field is only visible if you are checking that every field is present, not just that the visible text is readable.

## How the software reads dates and titles

Dates and titles are the fields the ATS uses to compute your experience, so their format matters more than any other formatting choice. The ATS calculates years of experience by summing the gaps between job dates; if a date is formatted in a way the parser does not understand, the field comes back blank and you drop out of filtered searches.

Two date formats are consistently endorsed across engines: MM/YYYY (03/2022) and Month Year (January 2022 or Jan 2022). One tester reports that four formats parse in every engine tested - full month plus year, abbreviated month plus year, MM/YYYY, and year-only for older roles. The safe rule is to include both month and year and to use one style throughout the document.

Titles should be clean standalone strings. The safest pattern is title on one line, employer and location on the next, dates on the third, and bullets beneath. When you merge title, employer, and dates onto a single line to save space, you make the parser guess where one field ends and the next begins.

> **Watch out:** A date that parses but parses wrong
>
> "01/20" or "Jan 20" can read as a typo or an ambiguous year, and the parser may skip the whole entry, blanking your years-of-experience for that role. Always include the full year and pick one date style for the entire resume.

## PDF versus DOCX, and what actually breaks a parse

The file extension is a decoy. There is no rigorous public benchmark showing a precise reliability gap between clean PDF and clean DOCX on identical files, and I will not invent one. What practitioner sources agree on is that the PDF type is the variable, not the letters after the dot.

Every major modern ATS parses a clean, text-based PDF reliably. What breaks parsing is scanned images, tables, multi-column layouts, and content buried in headers. An image-based PDF - the kind you get from a scan or some design-tool exports - returns nothing at text extraction, because there is no selectable text to recover. That is why "PDF fails ATS" stories cluster on print-to-PDF and design-tool files, not on ordinary text PDFs.

| Channel | Best file choice | Why |
| Unknown or legacy portal | DOCX, single column | Parses most reliably across systems; safest for strict engines like Taleo |
| Direct email to a human | Clean text-based PDF | Preserves layout for the reader; parses fine if text-based |
| Portal that says "Word only" | DOCX | Follow the explicit instruction |
| Any channel | Under 2.5MB, not password-protected | Greenhouse documents a 2.5MB limit; password-protected PDFs cannot be opened at all |

DOCX is the safer default across unknown systems. A clean text-based PDF is a fine choice when the layout is simple and you know a human will open it. The two hard rules: keep the file small - Greenhouse documents a 2.5MB parsing limit and Bullhorn a 4MB failure threshold - and never password-protect it, because a locked PDF cannot be opened by a parser at all.

**4.6x - How much more often US recruiters list Workday than Greenhouse in Refolk's index**

Among US recruiter and talent-acquisition profiles, about 524 name Workday versus 113 naming Greenhouse, mirroring Workday's larger enterprise footprint.

Because enterprise hiring concentrates in a few systems, you are effectively optimizing for a handful of parsers. Workday powers hiring for roughly 39% of Fortune 500 companies, and in Refolk's index of professional profiles US recruiters name Workday about 4.6 times more than Greenhouse. Single-column, parseable formatting that targets the dominant systems covers most large-employer applications. You do not need a different resume per vendor; you need one that survives the strictest common denominator.

## The procedure: check every field in order

Run this once on the exact file you will submit. It takes about ten minutes and surfaces every failure that matters. Each step has a clear "done" state, so you know when to stop.

#### Verify your resume field by field

1. **Run text extraction on your real file** - Open the exact PDF or DOCX you will submit and press Ctrl+F to find your job title. If you cannot highlight or find the text, it is an image and the parser recovers nothing; re-export from the source document as real text.
2. **Check reading order with a plain-text paste** - Copy the whole resume from the submission file and paste it into Notepad or any plain text box. Done looks like readable words in top-to-bottom, left-to-right order; scrambled text means your columns or tables are flattening.
3. **Confirm every field survived segmentation** - Scan the pasted text for standard section headers - Contact, Experience, Education, Skills - and confirm each is present. If a phone number or skills sidebar is missing, it was likely in a text box or header the parser skipped.
4. **Verify dates and titles parse as fields** - Confirm each role reads as title on its own line, employer and location next, dates on a third line, in MM/YYYY or Month Year format used consistently. This is what lets the ATS compute years of experience and index each role.
5. **Match the file type to the channel** - Use DOCX for unknown portals and older strict systems, a clean text-based PDF for direct human email, and follow any explicit "Word only" instruction. Keep the file under 2.5MB and never password-protect it.
6. **Answer knockout questions accurately** - Fill in eligibility fields - work authorization, licenses, location - truthfully and completely, since these are the one place true automatic rejection happens. Done means no required gate is left blank or wrong.
7. **Stop after the two fixes that matter** - Once the paste test is clean and dates parse, stop. Fonts, bullet glyphs, and bold text parse fine; rewriting them is wasted effort that does not change your ranking.

The Ctrl+F test in step one and the paste test in step two are separate checks. Ctrl+F tells you whether text exists as text. The paste tells you whether that text comes out in the right order and with nothing missing. You need both, because a file can pass one and fail the other.

Tailoring the fielded parts of each application to the posting is the slow part of this work. [Refolk](/candidates) writes your resume from your own history, tailors it to each posting, and scores how well you actually fit, so the parseable structure is handled before you ever run the paste test.

## Where this goes wrong: false positives and myth-driven panic

The most common mistakes here are not doing too little; they are checking the wrong thing or believing the wrong story. Each failure below has a specific check that catches it.

| Failure mode | What it looks like | The check |
| Paste-test false positive | Text pastes cleanly, but you copied from Word or Google Docs, not the export | Paste from the actual PDF or DOCX you will submit |
| "Looks fine in preview" trap | Two-column PDF looks perfect on screen, parses to gibberish | Only the raw text stream counts, not the picture |
| Silent text-box omission | A skills sidebar or phone number simply vanishes, no error | Confirm every field is present in the pasted text |
| Date parses but wrong | "01/20" reads as a typo; the entry is skipped | Use MM/YYYY or Month Year, month and year, one style |
| Over-correction | Rewriting fonts, glyphs, and bold that parse fine | Fix only what the paste test actually breaks |
| Myth-driven panic | Keyword-stuffing because you think a bot deletes resumes | Answer knockout questions; get titles and dates parseable |
| Wrong file for the channel | Design-tool PDF into a legacy portal | DOCX for unknown portals, text-based PDF for human email |

Two of these deserve extra weight. The "looks fine in preview" trap is why the paste test exists at all: a two-column PDF can render beautifully and parse into a scramble where skills land inside job titles. The visual is not the evidence. Only the raw stream is.

The bigger trap is myth-driven panic. The belief that a robot deletes most resumes before a human sees them pushes people into keyword-stuffing, which does nothing useful, instead of the two things that move the needle: answering knockout questions accurately and making titles and dates parseable. The "75% rejected by bots" line is folklore. It traces to a 2012 sales pitch from a company that closed in August 2013, with no study behind it. The resume-improvement market is estimated at $268 million, which is a strong incentive to keep the myth alive.

> The verifiable adoption number and the debunked rejection number are different claims from different origins, welded together to sell you a fix.

## What the software does not do

An ATS almost never throws your resume away on its own. This is the fact the scanner-upsell pages work hardest to obscure, and it changes what you should spend your energy on. A 2025 study of 25 US recruiters found that 92% of ATS platforms do not auto-reject resumes for formatting or design, and recruiter Jan Tegze reports that 90 to 95% or more of applications are reviewed by a human.

The one place true automatic rejection happens is a knockout question: a hard eligibility check like work authorization, location, or a required license. If you answer one of those wrong, or leave it blank, the system removes you. That is the real automated gate, and it has nothing to do with your fonts.

#### Where applicants actually drop out

| Stage | Figure | Note |
| --- | --- | --- |
| Fortune 500 using a detectable ATS | 98.4% | Nearly universal at large employers |
| Applications reviewed by a human | 90-95% | Most resumes reach a person |
| Platforms that do NOT auto-reject on format | 92% | Formatting rarely deletes you |
| True auto-rejection point | Knockout only | Work authorization, license, location |

*Adoption is near-total, but the drop-out points are knockout questions and sort order, not a formatting delete button.*

So the mechanism that costs you interviews is burial, not rejection. The software scores every application, sorts them, and a recruiter works the list top-down. An unparsed date blanks your years-of-experience, dropping you below the filter line, and a qualified candidate loses to sort order rather than a delete button. Getting your fields parseable is not about escaping a bot. It is about not sorting yourself to the bottom by accident.

## Grounding the numbers: adoption is real, rejection is not

Two statistics run through every article on this topic, and they come from different places with different reliability. Keep them separate.

| Year | Fortune 500 with a detectable ATS |
| 2023 | 97.4% |
| 2024 | 98.4% (492 of 500) |

That adoption data is load-bearing and well-sourced: Jobscan reviewed the job-listing pages of all 500 Fortune 500 companies and detected the ATS on 492 of them. Broader figures from over a million scans across 12,820 companies put usage at roughly 70% of large companies and 20% of small and mid-sized businesses. The takeaway is simple: if you are applying to large employers, assume your resume will be parsed. That is why the paste test is worth ten minutes.

The rejection number is the opposite. The "75%" figure has no study behind it and originates from a defunct company. Do not let a page fuse the true 98% adoption fact to the false 75% rejection fact and sell you a scanner on the strength of the pairing.

**92% - Share of ATS platforms that do not auto-reject on formatting or design**

From a 2025 study of 25 US recruiters, this is the number that dismantles the "robot deleted my resume" story.

If you want to see who is on the other side of the parse - the recruiters who search these indexes and work the sorted list - you can look them up directly.

Ask me this: `Technical recruiters in the United States who use Workday.` - [run the search](https://www.refolk.ai/start?q=Technical%20recruiters%20in%20the%20United%20States%20who%20use%20Workday.).

*Returns US technical recruiters who list the dominant enterprise ATS, the exact people whose keyword and Boolean searches your parsed fields need to match.*

Refolk's index lists roughly 131,117 recruiter and talent-acquisition profiles in the United States against about 9,175 in the United Kingdom, a 14.3-to-1 ratio. The point of naming that is scale: the person reading your surfaced resume is real, plentiful, and searching by fielded skills, not admiring your layout.

## Keeping this current and calling it done

Run the paste test on the actual file every time you change the layout or export from a new tool, because the export is where parsing breaks, not the design view. The mechanism in this guide is stable, but individual systems tighten over time, so re-check whenever you switch templates or file types rather than trusting a check you ran on a different version.

#### Before you submit

- [ ] I ran Ctrl+F on the submission file and could find my job title as selectable text.
- [ ] I pasted the exact file into a plain text box and the words came out in reading order.
- [ ] Every field - contact, each role, education, skills - is present in the pasted text.
- [ ] Dates use MM/YYYY or Month Year, include month and year, and are one consistent style.
- [ ] Each role reads as title, then employer and location, then dates, on separate lines.
- [ ] The file is single-column, under 2.5MB, and not password-protected.
- [ ] I chose DOCX for an unknown portal or a text-based PDF for a human, per the channel.
- [ ] I answered every knockout eligibility question accurately and left none blank.
- [ ] I stopped after fixing what the paste test broke and did not rewrite cosmetic formatting.

When the checklist is clean, the parse is done. The remaining work - matching the posting, ranking your fit, deciding where to apply - lives in other guides in this library. But the file itself is now a clean structured record that a recruiter can find, which is the entire job of the parse path.

## Frequently asked questions

### Will my resume parse correctly if it looks fine on screen?

Not necessarily. A two-column PDF can look perfect visually and still parse into scrambled gibberish, because the parser reads a linear text stream in file-storage order, not the columns your eyes see. The only reliable check is the plain-text paste: copy from the exact file you will submit and paste it into a plain text box. If the words come out readable and in order, the software can read them too.

### Does an ATS read a two-column resume?

It reads the text but often in the wrong order. Parsers flatten a two-column layout into one stream, so a skill can land in the middle of a job title. Layout tables, text boxes, and headers or footers cause the same scramble or get skipped entirely. If your paste test comes out jumbled or missing content, collapse to a single column before doing anything else.

### Is PDF or DOCX better for resume parsing?

The extension is not the real variable; the PDF type is. Every major modern ATS reads a clean, text-based PDF reliably, while an image or scanned PDF returns nothing. DOCX is the safer default across unknown systems and older strict ones like Taleo. Use a text-based PDF for direct human email, DOCX for unfamiliar portals, and always follow an explicit "Word only" instruction.

### Does an ATS auto-reject my resume for formatting?

Almost never. A 2025 study of 25 US recruiters found 92% of platforms do not auto-reject on formatting or design, and one recruiter reports 90 to 95% of applications are seen by a human. The one place true automatic rejection happens is a knockout question - a hard eligibility check like work authorization or a required license. The bigger risk is burial low in a sorted list.

### How should I format dates and job titles so they parse?

Use MM/YYYY (03/2022) or Month Year (January 2022), pick one style, and keep it consistent. Put the title on its own line, employer and location on the next, and dates on a third. This lets the ATS compute years of experience by summing the gaps between dates. An ambiguous date like "01/20" can be read as a typo and skip the whole entry, blanking that field.

### Is the free paste test as good as a paid resume scanner?

For the failure that matters most, yes. The clipboard paste surfaces column-scramble and missing-content omissions directly, which is exactly what breaks parsing. Paid accessibility checkers often miss layout, font, and date problems. Paste the file you will actually submit, confirm every field is present and in order, and you have run the diagnostic that counts.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/candidates/guides/resume-parse-path-field-by-field*