Every application form you meet now offers to fill itself in. Upload your CV and the fields populate. Click "Apply with LinkedIn" and your work history appears. Let your browser drop in your name, email and phone. Install an extension that reads the page and answers everything for you. The time saving is real, and so is the damage: autofill introduces small, specific errors that a recruiter reads as carelessness and a filter reads as a mismatch.
This guide covers the four kinds of autofill, exactly how each one breaks, and the review routine that takes ninety seconds and catches almost all of it.
The four kinds of autofill, and why they fail differently
The "parse my CV" button on the ATS itself
Workday, Greenhouse, Lever, SuccessFactors and Taleo all offer to read your uploaded file and populate the employment and education sections. This is a parser making guesses about your document structure. It is guessing where one job ends and the next begins, which line is a job title and which is a company, and what a date range means.
It fails on layout. Two-column CVs, dates placed in a sidebar, job titles styled the same as company names, and role descriptions written as one paragraph all confuse it. Greenhouse and Lever parsers tend to do a clean job with a single-column file and then quietly drop your third and fourth roles. Workday is the most aggressive: it will create employment entries with dates it invented from context.
Apply with LinkedIn, one-click and easy-apply flows
These pull from your LinkedIn profile, not from your CV. That is the core problem. Your profile is almost certainly less current, less precise and differently worded than the CV you tailored for this role. If your profile still lists a role you left in March, the form now says you are still there.
LinkedIn also stores a small library of CV files you have used before. Easy-apply flows will pick one automatically, usually the most recent upload, and it is often not the version you want for this employer. Candidates regularly discover they sent a CV tailored to a completely different sector.
Browser and password-manager profile autofill
Chrome, Safari and password managers fill contact fields from a saved profile you built years ago. This is where old phone numbers and dead email addresses come from. If you moved, the address field arrives with your previous city, and address is a field some employers use to filter on commuting distance or work location.
These tools also match fields by their internal name, not their visible label. A field labelled "Current employer" whose internal name is "organization" will get your own postal address's organization value. A field asking for a portfolio URL can receive your personal website because both are typed as a link.
Third-party AI form-filling extensions
The newest category, and the least predictable. An AI extension reads the rendered form and writes an answer into every input it finds, including the ones you would have thought about. It produces confident answers to free-text questions, fills in a salary figure, picks a notice period, and answers yes or no screening questions from whatever it inferred about you.
The specific danger is that these extensions answer the knockout questions. A question like "Do you require sponsorship to work in this country?" has one answer that ends your application. An extension guessing wrong on that field costs you the role before a human sees anything.
The breakages worth looking for
Employment dates that shifted or mangled. The single most common failure. A role that ran from June 2021 to September 2023 becomes 2021 to 2023 with January defaults, or the end date of one job lands on the start of the next, or a month is transposed. Any parser reading "06/2021" has to guess whether that is June or a week in 2021.
A current role filled in as a past one. The parser sees an end date where you wrote "Present" and closes the entry. The form now shows every role as finished, and to a recruiter scanning the list you look unemployed.
The wrong current employer. When roles get merged or reordered, the employer name from an older job attaches to your most recent dates. Recruiters compare this against your LinkedIn in seconds, and a mismatch on your current employer is the worst possible first impression.
Bullets merged into an unreadable block. Your carefully structured achievements arrive as one 900-character paragraph with no separators. Nobody reads it.
Accented characters turned into symbols. If your surname carries an accent, a cedilla or a tilde, some parsers and older ATS versions convert it into a question mark, a box or a string of stray characters. Your name is the first thing on the form.
Old contact details. A phone number you stopped using. An email from a previous employer. These are silent failures because nothing on screen looks wrong.
Fields truncated at a character limit. Many summary and description fields cap at 250, 500 or 1000 characters, and most cut without warning. Your sentence stops mid-word. Autofill will not tell you it happened.
The wrong CV version. The stored library problem. Check the filename shown next to the attachment, every time.
Salary and notice period prefilled with a default. Some forms remember a figure from a previous application. Some extensions invent one. Committing to a number you did not choose is a real cost.
Screening questions answered by default. Work authorisation, years of experience with a named tool, willingness to relocate, driving licence, security clearance. These are frequently the actual filter. A default "no" on a question you would have answered yes eliminates you invisibly.
Address and right-to-work fields from stale data. Visa status, residency and location all change, and saved profiles do not.
Duplicate applications. When an autofill flow breaks halfway and you restart, the ATS may already hold a partial record. Two submissions from the same person on the same requisition reads as either a system error or as someone who is not paying attention.
Why these errors matter more than they look
Two reasons, and the second is the one people underestimate.
The first is perception. A recruiter opening your application sees a form with your current role marked as ended and a paragraph of run-together text. They do not know a parser did it. They see a candidate who submitted without reading. If you are one of eighty applicants, that is enough.
The second is structural. The form data is what gets indexed and searched, not always your PDF. When a recruiter runs a search inside the ATS for candidates with four years in a discipline, or filters on current job title, the system queries the structured fields you submitted. Your beautifully formatted PDF sits in an attachment tab that nobody has opened yet. If autofill put the wrong title in the title field, you will not appear in that search regardless of what your CV says.
That also creates a credibility problem. When a recruiter finally opens the PDF and the dates disagree with the form, the discrepancy looks deliberate. Nobody assumes a parsing bug. They assume you gave two versions of your history, and questions about honesty are much harder to recover from than questions about fit.
When autofill is safe, and when to type
Autofill is fine when you are populating a long Workday application you would otherwise abandon, when the fields are structural rather than persuasive, and when you have the time to check the output. It works well for education entries, which are short and rarely ambiguous.
Fill by hand when the application is short enough that autofill saves you two minutes rather than fifteen. Fill by hand for any screening or knockout question, without exception. Fill by hand for salary, notice period, work authorisation and location. Fill by hand for every free-text answer, because those are the fields the recruiter actually reads and generic text is obvious.
Turn off AI form-filling extensions on applications you care about. Their whole value is speed on volume applications, and volume applications are not how good roles get won.
The order of operations that works
Do it in this sequence and you keep the speed without the risk.
- Autofill first, as a rough draft. Let the parser do the typing. Treat every field it produced as unverified.
- Open your CV in a second window. Not from memory. Side by side.
- Correct every field against the CV, top to bottom. Names, dates, titles, employers, descriptions.
- Answer the screening questions manually. Clear anything prefilled and answer it yourself.
- Attach the file last, after the text fields match it, and confirm the filename is the version you intended.
Attaching last matters because some ATS platforms re-run the parser when a new file is uploaded, overwriting corrections you already made. Upload, then check, or check after every upload.
Pre-submit review checklist
Run this on the review screen before you confirm. It takes about ninety seconds.
- Your name is spelled correctly, with no stray symbols where an accented letter should be
- The email address is the one you check, and the phone number is the one you answer
- Your current role shows as current, not ended
- The employer on your most recent entry is correct
- Every start and end month is right, not just the year
- No role is missing and none is duplicated
- Job titles match your CV word for word
- Description fields are readable, with line breaks intact, and no sentence stops mid-word
- Location and address reflect where you actually live now
- Work authorisation and visa answers are current
- Salary and notice period are figures you chose
- Every screening question was answered by you
- The attached file is the right version, with the right filename
- You have not already applied to this requisition
When you find an error after submitting
Most ATS platforms let you withdraw and reapply, and some let you edit until a recruiter opens the application. Log back into the careers portal first and look for an edit or withdraw option on the application record.
If you cannot edit, judge the size of the error. A wrong phone number or a wrong current employer is worth a short, plain email to the recruiter or the address on the job posting: state which role you applied for, name the field that is wrong, give the correct value, and stop there. No apology paragraph. A truncated description is not worth an email.
If the error is a knockout question answered wrongly by an extension, contact them. That answer may have already routed you to rejection, and a one-line correction is your only route back in.
Do not submit a second application to fix the first. That creates the duplicate problem on top of the original error.
Keep a master text version that parses reliably
The reason autofill mangles your CV is usually your CV, not the parser. Keep a plain, boring version specifically for machines.
- One column. No text boxes, no sidebars, no tables, no headers or footers holding your contact details
- Contact details as normal body text at the top
- One employment entry per block, in reverse chronological order
- Dates written the same way every time, with the month spelled out: March 2022 - Present
- The company name on its own line, the job title on its own, never combined with a separator character
- Achievements as short separate bullets, not one paragraph
- Standard section headings: Experience, Education, Skills. Not "Where I Have Made An Impact"
- Save as a text-layer PDF exported from your editor, never a scan or an image
Keep a .txt or .md copy of the same content. When a form gives you a free-text box for your history, you can paste from that instead of relying on a parser at all.
Then run the test that matters: upload it to a real application form, let it parse, and look at what comes out. If the fields are clean on a Workday or Greenhouse form, they will be clean almost everywhere. If they are not, the file needs fixing before your next fifty applications inherit the same problem.