9 min read
The cruel part of resume screening is that the file which looks best on your screen is frequently the one that arrives as nonsense.
These are the five formatting mistakes that kill your resume with AI before anyone assesses your experience. They're not stylistic preferences and they're not close calls. Each one produces a specific, predictable failure, and each one takes a few minutes to find and fix.
Do the diagnostic first. Upload your current file to the free ATS resume checker and read what it extracts. If a job is missing its dates or your skills have landed under the wrong employer, one of the five below is why.
Quick Wins
- Paste your resume into a plain text editor right now and read the order.
- Search the PDF for your phone number. No result means the text isn't real.
- Turn on gridlines in Word and look for a grid around your dates.
Why formatting breaks AI screening
Before any matching or ranking happens, something has to turn your document into data: employer, title, start date, end date, degree, skills. That step is called parsing and it is much dumber than the AI language around it suggests.
A parser reads in a linear order and assigns what it finds to fields. It has no idea that the box on the left is a sidebar and the text on the right is your career. When your layout is ambiguous, it doesn't stop and ask. It guesses, commits, and passes the result downstream.
The consequence is worse than rejection. A file that fails to parse doesn't bounce back with an error. It becomes a profile with two jobs missing and a graduation date where a start date should be, and that's what gets scored, ranked, and summarised.
You are never told this happened. Which is exactly why people send the same broken file to forty employers.
1. Two columns
The single most damaging layout choice, and the most popular one in modern templates. Skills down the left, experience on the right, contact details in a coloured strip.
A parser reading that page has to decide whether to read across or down. Read across and your job title collides with a skill from the sidebar. Read down and your entire sidebar arrives before any of your experience, so the first thing the system learns about you is a list of software names with no context.
How to spot it in 30 seconds: select all, paste into a plain text editor. If your career reads out of order, that is precisely what was submitted.
The fix: one column, top to bottom, for the whole document. Summary, experience, education, skills. It looks plainer and it performs better at every stage, including with the human at the end.
2. Tables you can't see
This one is genuinely sneaky, because the problem is invisible. Most templates that align dates neatly to the right margin do it with a borderless table. You see clean spacing. The parser sees a grid.
Tables scramble reading order in exactly the way columns do, and dates are the usual casualty. A job with no dates attached can be dropped from experience calculations entirely, which then fails the years-of-experience filter on the application form.
How to spot it: open the file in Word, go to the table layout options and turn on gridlines. Or simply click into one of your dates and look for the small cross-shaped handle that appears at the corner of a table.
The fix: rebuild those sections as plain stacked text. Put the dates on the same line as the job title, separated by a comma or a pipe, and let them sit left rather than fighting for the right margin.
Instead of a table row: Senior Analyst | Acme Ltd | [right-aligned cell] Mar 2022 – Present
Write: Senior Analyst, Acme Ltd — March 2022 to Present
3. Contact details only in the header
It looks tidy and it's a genuine risk. Some systems do not scan header and footer regions of a document at all, so a name, email, and phone number placed only there can be entirely absent from the imported profile.
The failure mode here is the most frustrating of all: a strong candidate who ranks well and cannot be contacted, or whose record has a blank name field and gets skipped.
How to spot it: in Word, double-click at the very top of the page. If the region becomes separately editable and the rest of the document greys out, your details are in a header.
The fix: put your name, email, phone, city, and LinkedIn in normal body text at the top of page one. Write them as words, not icons. A small telephone graphic is not read as a phone number by anything.
4. Text that isn't actually text
If your resume was exported from a design tool, scanned, screenshotted, or flattened, there may be no extractable text in it whatsoever. It looks perfect and it contains nothing.
Partial versions are more common than the full disaster. A logo, a headshot, a skill bar chart, or a set of icons replacing labels. Each one is a region the parser cannot read, and skill rating graphics are the worst offender: they claim a proficiency level that no system can extract and no recruiter believes anyway.
How to spot it: open the PDF and search for your phone number with Ctrl or Cmd+F. If it isn't found, it isn't text. Try the same with a job title.
The fix: export from a word processor rather than a design tool, and write skills as a plain comma-separated list. If you want to show proficiency, say it: "Python (4 years, production)" beats four filled circles out of five.
5. Creative section headings
"My Journey." "Where I've Made an Impact." "The Toolkit." A parser looking for a section called Experience does not find one, and everything below that heading may never be mapped to your work history.
The same applies to job titles. Internal titles like "Analyst III" or "Member of Technical Staff" mean nothing outside the company that invented them, and they don't match anything in the posting either.
How to spot it: read your headings and ask whether each one is a word a form field would use. Experience, Education, Skills, Certifications, Summary. If it isn't on that kind of list, rename it.
The fix: boring headings, and market-standard job titles. Where your internal title is unrecognisable, write the standard one and add the internal one after it: "Data Analyst (internally: Analyst III)". There's more on that in handling internal titles .
The five-minute repair pass
Run these in order on your current file. Most people find two or three problems.
- Copy-paste test. Select all, paste into plain text, read the order.
- Search test. Ctrl+F for your phone number and one job title in the PDF.
- Gridline test. Open in Word, show gridlines, hunt for tables.
- Header test. Double-click the top of the page and see what becomes editable.
- Heading test. Read your section names out loud. Are any of them clever?
- Full check. Upload to the checker and compare the extraction to the document, line by line.
Do this now: the copy-paste test takes twenty seconds and catches the most damaging mistake on the list. Do that one before you apply to anything else today.
Edge cases
You're a designer and the portfolio matters
Keep two files. A plain single-column resume for applications, and a designed portfolio PDF you link to. The application system never needed to parse your visual work, and your layout skills are demonstrated in the portfolio rather than in the file being scanned.
Academic CVs
Long publication lists are fine and normal in academia, where a human reads them. Applying to industry roles with the same document usually goes badly, so keep a two-page industry version with the publications summarised rather than listed.
You have to upload and also fill in the form
Check every auto-filled field after upload. If the import scrambled something, correct it in the form rather than assuming the resume speaks for itself. The form data is frequently what gets filtered on.
Non-US date formats
Write dates as Month Year, spelled out, when applying to US employers. Ambiguous numeric formats get misread, and a job that appears to have ended before it started tends to be dropped.
You already applied with the broken file
Some systems let you update your profile or reapply after a period. Fix the file first, then check whether the employer allows a resubmission, and don't assume the original import will be refreshed automatically.
Check it, then rebuild if you have to
These tests tell you something is broken. A full check tells you what the system actually received, which is the part you can't see any other way.
Upload the file to the free ATS resume checker and compare the extracted output line by line against your document. Missing jobs, missing dates, and misplaced skills are the three things to look for.
If the template itself is the problem, patching it is usually slower than starting from a clean base with the free resume builder . And once the file parses, the cover letter generator handles the part a parser doesn't score but a human does.
The short version
- The formatting mistakes that kill your resume with AI are columns, hidden tables, header-only contact details, non-text text, and creative headings.
- Each has a 30-second detection test. Run all five today.
- A file that fails to parse never tells you. It just gets scored on whatever survived.
Do this today: run the copy-paste test. If your career comes out in the wrong order, rebuild the layout before you send another application.
Then confirm the fix. Check your resume for free and read the extraction rather than the score.
Read more
- How to beat AI resume screeners — all five gates, not just formatting.
- The most common ATS formatting mistakes — the wider list.
- Why you're not hearing back — working out whether format is really your problem.
Frequently asked questions
A two-column layout. Parsers read a page in a linear order, and columns force them to guess where that order runs. When the guess is wrong, skills from a sidebar get attached to the wrong job and your work history arrives interleaved and unreadable.
Open the file in Word and turn on table gridlines, or click inside a date and look for the small move handle in the corner. Many templates use borderless tables for alignment, so the structure is invisible on screen but very much present in the file.
No. Some systems do not scan header and footer regions at all, which means your name, email, and phone number can be missing from the imported profile even though they are clearly visible on the page. Always repeat them in the body text.
Either works if the text is genuinely selectable. A clean DOCX is the safer default for older enterprise systems, and a text-based PDF is fine almost everywhere. What fails is an exported image, a scan, or a design file flattened into a picture.
Icons used in place of labels do. A small envelope graphic next to your email is not read as the word email, so the field may not map. Skill rating bars are worse, because they convey a level that no system can extract and no recruiter trusts.