Parse-truth CV rebuild
Job parsers delete your CV before a human sees it.
Upload once. See exactly which fields real parsers drop from your file — the name, the dates, the years of experience — then download the same CV rebuilt to survive them. Your words are never rewritten.
Free verdict. You only pay if the rebuild is worth it.
- No account
- Files deleted in 30 days
- No AI rewriting your words
- Name
- not found
- not found
- Roles
- 4 found · titles reassigned
- Tenure
- 0 months
- Skills
- 0 of 14
- Languages
- 2 of 2
- resumes parsed in a 17-day audit
- 3,835
resumes parsed in a 17-day audit
- carried at least one parse-damaging defect
- 81%
carried at least one parse-damaging defect
- of those defects are invisible on screen
- 45%
of those defects are invisible on screen
- lost their name entirely
- 26%
lost their name entirely
Measured by a practitioner who parsed 3,835 real resumes in 17 days (public write-up, August 2026) — not a survey we commissioned. Character-level defects — em dashes, smart quotes, icon bullets — are what nobody sees until they are screened out.
The damage
Same file. Two readings.
A parser does not see your layout. It sees lines of text in the order they were drawn — so a designed CV arrives as two columns interleaved, with every job title next to the wrong employer.
- 1Resume← name was a logo
- 2
- 3SKILLS Design Figma Systems
- 4MY JOURNEY← unknown section
- 5Senior Product Designer Northwind Ltd
- 62019 — 2021← date fails → tenure 0
- 7Led the design system Acme Corp← title married to the wrong employer
- 82016 — 2019
- 9Chief of Staff Globex
- 102014 — 2016
- 11
- 12e?cient — built 12 systems← ligature + em dash
The parser does not read columns. It reads line by line, across both — so each job title lands next to the wrong employer, the dates never match, and tenure comes back as zero. Nothing tells the candidate it happened.
What we fix
Twelve ways a CV gets damaged — and the rule that removes each one.
Every rule is deterministic and testable. Percentages come from the 3,835-resume audit where it measured them; blanks are defects it did not count, not defects that do not happen.
- 01Image-only PDF — a scan, a Canva export, a print of a graphic~10%Zero text. Nothing to read at all.→OCR, show you the recovered text, require your confirmation
- 02Two columns, or one side bar11%Columns interleaved line by line — job titles married to the wrong employers.→Single column, one reading order, no side bars
- 03Contact block in a page header or footer8%Header zone skipped — “no email found”.→Contact in the body flow, immediately under the name
- 04Tables, text boxes, shapes, SmartArtCells merged, or the whole block dumped at the end.→Real paragraphs. No tables, no containers, no floats
- 05Your name as a logo; ✉ ☎ as icon rows26% name · 18% iconsName = “Resume”. Contact fields missing entirely.→Name and labels as plain text
- 06Sections called “My Journey”, “Where I've Been”Section not recognised — the entire block is dropped.→Canonical labels only: Summary · Experience · Education · Skills · Certifications · Languages
- 07Dates like “Summer '19”, “2019–20”, “03/2019–present”9%Date parser fails → tenure = 0 → filtered out by “5+ years”.→Mon YYYY – Mon YYYY, present = Present
- 08Skills as a tag cloud, an icon row, or “Python ●●●●○”41%Skills list empty.→One plain, comma-separated list
- 09Fonts with no character map, ligatures, Symbol bulletsMojibake — “efficient” becomes “e?cient”.→Embedded standard font, standard bullets
- 10Em dashes — the ones Word autocorrect inserts19%Read as a stray glyph, or the text either side is dropped.→Plain hyphen only
- 11Smart quotes8%Mangled by older parsers.→Straight quotes, or no quotes
- 12Name merged with the next line — MUHAMMAD TANZEELHAYDER8%The name cannot be split from the headline.→Name alone on the first line of the body, headline on the next
Three of these — the em dash, the smart quote, the icon bullet — are invisible on screen and account for 45% of the damage on their own. That is the whole case for buying a check instead of reading advice.
How it works
Upload once. Ship the file that reads clean.
Four steps, no builder, no template gallery, no job-description matcher. You are not being sold a new CV — you are being sold the same one, readable.
Upload once
PDF or DOCX. Nothing else, no builder, no template gallery. Your file is read by five real extractors: pdftotext -layout, PyMuPDF, pdfminer.six, python-docx and Apache Tika.
See the verdict
Per field: what was recovered, what was dropped, re-ordered or mis-dated. We report the worst extractor, never the best — so you see the reading a screening system would actually get.
Get the rebuild
The record is normalised by fixed rules and rendered through a template we own. Single column, canonical sections, normalised dates, plain text, embedded font. No model in the critical path.
It ships verified, or not at all
Before delivery the rebuilt file is parsed again and diffed field-by-field against the record. One hundred percent of fields, on every extractor — otherwise it does not ship, and you see the diff.
What lands in your inbox
The primary parse target. Applicant tracking systems parse DOCX more reliably than PDF.
Generated from that DOCX. Tagged, with a real text layer and embedded fonts.
The lowest common denominator. Accepted by every portal ever built.
Your original design file, untouched, for the human who gets to see it.
Word-level. What changed, and the words that never did.
The guarantee
A claim you can check, not a promise you have to believe.
If our own verifier drops a single field, or any of the five test parsers does, we rebuild it free. If it still fails, you get your money back.
This is checkable rather than a slogan: no model sits in the critical path, marginal cost per order is under five cents, and the output is reproducible byte-for-byte from the record. That is what makes the guarantee affordable.
Pricing
One-off. No subscription.
The verdict is free and stays free. You pay for the file that actually gets submitted — once, not monthly, with no trial that quietly renews.
One CV
one-offOne file in, three files out. For the CV you are sending today.
- DOCX + PDF + TXT, rebuilt and verified
- Your original design file back, untouched
- Field-by-field verdict — worst of five extractors
- Word-level diff: what changed, what never did
- Free rebuild if our verifier drops one field
Search pack
one-offUnlimited re-runs until you are hired. For the whole search, not one application.
- Everything in One CV
- Re-check your rebuilt file every time you edit it
- New versions of your CV, same rebuild engine
- Priority rebuild queue
- Same guarantee, for as long as you are looking
Payment is one-off through Stripe — no subscription, no trial that renews. You only pay once the free verdict tells you something is wrong. For reference: Jobscan is $89.95 per quarter and Resume Worded $99 per quarter. Both check your CV; neither sends you the rebuilt file.
Questions
What people ask before trusting us with a CV.
Does it rewrite my CV?
No. Every one of the twelve fixes is mechanical: columns flattened, dates normalised, section labels replaced, em dashes turned into hyphens, contact moved out of the header. Your sentences are never rewritten. You get a word-level diff so you can see exactly what changed — and the words that did not.
Is there an AI in this?
Not in the rebuild. The record is extracted, normalised by fixed rules, and rendered through a template we own. That makes the output reproducible byte-for-byte — which is what makes the guarantee affordable. The only place a model can appear is OCR on a scanned file, and there it shows you the recovered text and waits for your confirmation.
I like how my CV looks. Do I lose it?
No. Your original file comes back untouched — that is the twin, for humans. The rebuilt file is the one you submit to portals. Nobody has to choose between looking good and parsing clean.
My CV is a Canva design with two columns. Will it survive?
That is exactly the case this exists for. A designed file is fine as a PDF for a human and a disaster for a parser: the columns interleave, titles land next to the wrong employers, and tenure comes back as zero. We rebuild it as a single column and keep your design file for the human copy.
Can you promise I get an interview?
No, and anyone who does is lying. What we check is narrower and verifiable: every field in your record must survive all five extractors in our harness before we ship the file. We call that parse-clean, verified by our harness, reproducible. We never call it ATS-proof.
What if my file is a photo or a scan?
We run OCR and show you what came back, then ask you to confirm before anything is rebuilt. We never guess at someone's employment history. If recovery fails, you are refunded and nothing is charged.
What happens to my file?
It is used to produce your verdict and your rebuild, nothing else. No training, no sharing, no public page. Deleted 30 days after upload, or immediately on request. You can delete your account and its files from your account page.
Which file do job portals actually take?
DOCX is the primary parse target — applicant tracking systems parse it more reliably than PDF. You also get a tagged PDF with a real text layer and embedded fonts, and a plain TXT as the lowest common denominator. Submit the DOCX unless a portal refuses it.
See what the parser drops from your CV.
Free verdict in seconds. No account, no card. If nothing is wrong, you owe nothing and you will know it.
Free verdict. You only pay if the rebuild is worth it.