12 min read
You open the application preview and your name looks fine. Then the experience block fills with •, empty squares, and random letters where bullets used to sit. Job titles land under education. Dates float alone on a line. You're not unqualified. The file broke on export, and the parser read what was left.
When a resume turns into gibberish after upload, the fix isn't keyword stuffing. It's matching the scrambled pattern to one of four causes: character encoding, tables, a Canva PDF export, or the wrong font. Before you rebuild bullet three for the third time, check your resume for free with the posting pasted so you're not tuning words while the parser still can't read the file.
Job searching's already draining when portals feel opaque. This page walks the symptom, the four causes, how to tell which one hit you, and the fix for each. You should know what to test in the next ten minutes instead of guessing whether recruiters ignored you on purpose.
Quick Wins
- Paste your file into Notepad. If symbols appear there, the source is broken before the ATS runs.
- Delete every table and text box. Rebuild as single-column plain paragraphs.
- Switch Wingdings bullets to standard round bullets or hyphens.
- Screenshot the portal preview beside your original so you can compare field by field.
What gibberish looks like inside Workday and Greenhouse
The symptom: Preview panes show odd characters, merged lines, or blank sections where your jobs should sit. Recruiters who open the attachment might see a clean PDF while the parsed profile behind the screen is nonsense. Screening runs on the parsed layer first.
Gibberish is not the same as a low match score. Gibberish means the system couldn't map text to fields. Your employer list might read as one long string: Acme Corp 2021 Project Lead Beta LLC 2019 Analyst. Or bullets become • • • with no verbs attached.
A composite marketing coordinator uploaded a polished Canva PDF. The portal preview showed her headline twice and left experience empty. She assumed the ATS hated her title. The export had placed job history inside grouped shapes the parser skipped entirely.
A composite data analyst saved a Word file with a skills table: tools in the left cell, years in the right. Greenhouse merged each row into one token. Tableau 4 years sat next to SQL 6 years on a single line with no spaces. Keyword search couldn't find either skill cleanly.
A composite nurse with an accented surname saw José in the preview while Word on her laptop still showed José correctly. That's encoding, not ranking. The portal read a non-UTF-8 byte stream and displayed mojibake.
Read Workday resume parsing error when the portal is Workday-specific and fields stay blank after a clean-looking PDF upload.
When your resume turns into gibberish after upload: four causes
Archetype D means we work by cause, not a generic polish checklist. Walk the four failures below. Match what you see in preview to one primary cause before you change wording.
Cause 1: Character encoding (UTF-8 vs legacy saves)
Accented characters, curly quotes, and emojis need UTF-8. Save as older Windows ANSI and the bytes shift when the ATS ingests the file. José becomes José. Smart quotes become “ and â€. En dashes in date ranges turn into question marks or empty boxes.
{``}
Before: Rewriting bullet verbs because the preview showed random symbols in the name line.
After: Opening Save As in Word, choosing UTF-8, retyping accented characters once, and seeing José survive in the portal preview.
Fix for encoding: In Word, use Save As and pick UTF-8 if available. Strip smart punctuation: replace curly quotes with straight quotes, replace special dashes with hyphens. Remove emojis from headings. Re-export and paste into Notepad. If Notepad shows clean text, encoding is likely fixed.
Cause 2: Tables, columns, and text boxes
Tables feel efficient on paper. Parsers read row by row, left cell then right cell, not visually. A two-column header with name on the left and contact on the right becomes one string. Experience tables with dates in column two attach dates to the wrong employer on the next row.
Side-by-side layouts in Word or Google Docs behave like soft tables. The ATS linearizes them. Skills on the left may merge with metrics on the right: Python SQL 12 deployments 98% uptime with no separation.
Before: Keeping a sleek two-column template because it fits on one page.
After: Rebuilding as single-column with contact on its own lines, each job as Company / Title / Month Year to Month Year, bullets underneath in plain paragraphs.
Fix for tables: Select the table, convert to text, insert line breaks manually. Delete text boxes. Put phone and email on separate lines under your name. See how ATS handles tables and columns when you're unsure whether a layout element counts as a table to the parser.
Cause 3: Canva export (designed PDFs and hidden layers)
Canva resumes look sharp on Instagram. The export often stores headings as positioned text boxes and icons as vector paths. Parsers read text layers. They skip paths. Result: blank experience, duplicated name blocks, or icons rendered as empty squares where section labels should be.
Export to PDF from Canva does not create a Word-style text flow. Print to PDF from a browser preview still carries the same box structure. Some portals OCR image layers poorly and produce worse gibberish than a plain DOCX would.
Before: Re-exporting the same Canva template at higher resolution hoping clarity fixes parsing.
After: Copying plain text into Word, applying Calibri 11, single column, saving DOCX, and treating Canva as a portfolio piece instead of an application file.
Fix for Canva exports: Select all text in Canva, paste into Word without formatting, strip colors and icons, rebuild headings as Word Heading 2 styles. Save DOCX. If the posting demands PDF, print to PDF from Word only after the plain version parses clean in preview.
Cause 4: Wrong fonts (Wingdings bullets and custom typefaces)
Fancy bullets are the quietest gibberish trigger. Word displays Wingdings or icon fonts as neat dots on your screen. The ATS substitutes another font and outputs •, ¤, or nothing. Custom script fonts for your name sometimes map to placeholder glyphs in Greenhouse preview.
PDFs that embed rare display fonts fail the same way when the server doesn't carry that font file. The parser substitutes Arial and drops characters it can't map.
Before: A designer resume with icon bullets for each skill line.
After: Standard round bullets or hyphens, Calibri or Arial at 11 point, employer names spelled in plain Latin characters the parser already knows.
Fix for fonts: Select all, set font to Calibri or Arial. Replace bullet styles: Home → Bullets → round solid. Avoid ALL CAPS section headers in decorative fonts. Re-save and scan preview for any remaining square boxes.
How to tell which cause is yours
Run the tests in order. Each takes about two minutes. Stop when one test matches your preview pattern.
| What you see in preview | Likely cause | First move |
|---|---|---|
| Accents or quotes show as à or †| Encoding | Re-save UTF-8, strip smart punctuation |
| Jobs and dates merge on one line | Tables or columns | Convert to single-column plain text |
| Blank experience, duplicated header art | Canva export | Rebuild in Word, abandon designed PDF |
| Bullets are squares or • chains | Wrong font | Switch to Calibri and standard bullets |
Second diagnostic for a composite project manager: preview showed perfect summary text but employment empty. Canva had tucked jobs inside a grouped element behind a color band. Ungrouping wasn't possible after export. Rebuilding in Word filled the fields on re-upload.
Third diagnostic for a composite customer support lead: only bullets broke, body text was fine. Wingdings in an old template was the culprit. One font change fixed every garbled line without touching metrics.
Copy-paste plain-text rebuild skeleton
Copy-paste this into Word after you strip tables and Canva boxes. Replace bracket lines with your content. Keep one blank line between jobs.
[Full Name]
[City, ST] · [phone] · [email] · [LinkedIn URL plain text]
SUMMARY
[Two lines, no icons, standard fonts only]
EXPERIENCE
[Company Name] · [Job Title] · [Month Year] to [Month Year]
- [Metric verb bullet, standard hyphen bullet]
- [Second bullet]
[Next Company] · [Job Title] · [Month Year] to [Month Year]
- [Bullet]
SKILLS
[Comma-separated list, no table]
Before and after: operations coordinator
Before (parsed preview): Skills row read Scheduling • Inventory • KPI • Vendor Mgmt with icon bullets lost.
After: Skills line read Scheduling, inventory control, KPI tracking, vendor management with commas instead of icon bullets and Calibri 11 throughout.
Before (parsed preview): Header table merged name and phone into one field.
After: Name on line one, phone on line two, email on line three. Workday autofill populated contact separately.
Where candidates blame keywords when the file broke
Chasing keyword density on a file the parser can't read. Adding SQL twelve times won't help if the skills section merged into one token. Fix structure first, then tune wording.
Trusting the attachment view. Recruiters may download a pretty PDF while screening uses parsed fields. Always check the portal preview pane or a plain-text paste before you assume the upload worked.
Re-uploading the same Canva export. Higher DPI doesn't turn shapes into text. Switch source files instead of exporting again from the same template.
Keeping hidden tables. Word sometimes leaves invisible grid lines. Turn on show formatting marks and delete any cell borders you see around contact info or skills.
Mixing font fixes with encoding fixes at random. Change one layer at a time. Font swap first if only bullets break. Encoding save first if names break. Table flatten first if dates attach to wrong jobs.
I've screened stacks of these in Workday and Greenhouse, and the file that wins is boring and specific. Gibberish almost always traces to export choices, not hidden keyword rules.
Preview before you trust the upload
Run tools after you pick a cause, not before. Flatten tables or fix fonts first, then validate. Otherwise the checker flags missing keywords on text the parser never extracted.
Run a free ATS check with the job description pasted. Confirm employers and Month Year dates survive as plain text. A readable file won't fix a designed Canva export you haven't replaced yet, but it confirms the plain rebuild worked.
Build your resume in a single-column layout when you're rebuilding from a scrambled Canva file. Starting from a plain template beats untangling grouped shapes by hand.
When the posting also asks for a letter, use the cover letter generator after the resume preview shows clean text. Match one metric from bullet one so human readers see consistency if they open both files.
Stop reformatting the wrong layer
A resume turns into gibberish after upload when encoding, tables, Canva exports, or wrong fonts break the text layer the ATS reads. Match the preview pattern to one cause, apply that fix, and re-test in plain text before you touch keywords again.
This won't land interviews for roles you're not qualified for. It does stop a qualified file from dying because bullets were Wingdings or jobs lived inside a Canva text box. Screenshot the preview. Flatten layout. Save UTF-8. Upload once the plain-text paste looks human.
When preview text is clean and silence continues for weeks, parsing isn't your blocker anymore. Read PDF vs Word resume for ATS to pick the safer format for the next posting, then run the free checker before you submit.
Run the free resume check on the rebuilt file so you're solving structure, not guessing.
Read more
Frequently asked questions
Your screen renders fonts and layout the ATS cannot read. Word shows Wingdings bullets as dots. Workday reads them as • or blank squares. Canva exports text as shapes. Tables read left-to-right across rows, merging job titles with dates. The preview pane shows what the parser extracted, not what you designed.
Often yes when the export is a designed PDF. Canva places headings inside text boxes and icons as vector paths. Parsers skip paths and read boxes out of order. Export a plain DOCX from Word or Google Docs for applications, and keep Canva for networking PDFs only when the employer accepts attachments by email.
Not always, but misaligned tables are a top gibberish trigger. Side-by-side skill columns merge into one string. Date cells in the right column attach to the wrong employer. A single-column layout with Month Year on the same line as the company name parses more reliably in Workday, Greenhouse, Lever, and Taleo.
Use Calibri, Arial, or Times New Roman at 10 to 12 point. Replace Wingdings or custom icon fonts with standard round bullets or hyphens. Accented names need UTF-8 saving in Word. If José becomes José in preview, re-save as UTF-8 DOCX before you upload again.
Try the plainer format first. A single-column DOCX with standard fonts fixes most encoding and table issues. If the posting requires PDF, print to PDF from Word after stripping tables, not from Canva. Paste the file into Notepad on Windows or TextEdit plain mode on Mac. If Notepad shows garbage, the source file is broken before the ATS sees it.