10 min read
You had an 84. You changed one bullet. You ran the checker again and got a 71. Your stomach drops because the edit felt like progress. You're not crazy to wonder if the tool is broken or if you broke something real. Most drops aren't random. They're the parser telling you a small move shifted where keywords live or how the file reads.
Run the free ATS checker on both versions if you still have the old PDF. Side-by-side beats guessing. Why your ATS score drops after small changes usually traces to placement shift, layout break, or density spike. Below is a diagnostic: the symptom, three causes, how to tell which one hit you, fixes, before/after pairs, and a copy-paste triage block for tonight.
This won't fix applying to roles you're not qualified for. It stops a qualified file from losing points because a template tweak scrambled your Experience block.
Quick Wins
- Export plain text from both PDFs. If section order scrambled, fix layout first.
- Check whether required terms moved from Experience bullets into Skills only.
- Undo density stuffing: one natural use per term beats twelve repeats.
- Re-run the checker on the same posting text, not a paraphrased job ad.
The symptom: your ATS score drops after a small edit
The symptom shows up the same way every time. You tailor for a posting. Score goes up. You bold a header, swap a synonym, or paste three keywords from the req. Score falls. You tweak again. Score wobbles. Anxiety tells you the file is worse when the only thing that changed was formatting or placement.
A drop after a small change is usually a signal about parser read order, not about your career. Checkers compare extracted text to the posting. They don't grade your worth. When extraction changes, the score changes even if the human-readable PDF looks identical.
A composite operations analyst ran 81 on version A. She merged two bullets and moved SAP S/4HANA into a Skills chip row for visibility. Version B scored 68 on the same posting. Plain text export showed SAP under Skills but missing from the Experience block the matcher weighted highest. One edit. Eleven points. Fix was moving the term back into bullet one, not reverting the whole resume.
Edge case: you pasted a different job description into the checker by mistake. Always confirm the posting text matches the req you are targeting before you diagnose a drop.
Edge case: the checker updated its weighting or keyword list between runs. Less common, but if nothing changed on your file and the posting is identical, re-run once more before you rewrite bullets.
Why your ATS score drops: three causes
Most score drops after small edits fall into three buckets. Placement shift means required terms moved out of Experience or out of the first eight words of a bullet. Layout break means the PDF still looks fine but plain text scrambles sections. Density spike means you added so many repeats the matcher flags stuffing. Below: how each looks, how to tell it's yours, and the fix.
Cause 1: Keyword placement shift
Checkers and parsers both lean on Experience before Skills. When you shorten a bullet, swap synonyms, or move a tool name into a summary line, the term may disappear from the zone that gets weighted. Moving content from bullet to Skills often lowers experience match even when the word still appears somewhere on page one.
How to tell this is yours: plain text still reads in order, but required terms show only under Skills or Summary after the edit. Another tell: you replaced a posting phrase with a synonym the checker does not map to the req.
Before: Bullet one opened with Salesforce CPQ admin for a $12M renewal book; score 79.
After: Rewrote bullet to lead with improved quote workflow; CPQ moved to Skills only; score 66.
Fix: Put the posting's top three terms back into bullet one under your current role. Use the exact phrase from the req once, naturally. Keep Skills as support. Re-run on the same posting text.
Cause 2: Layout and parser break
Small template edits cause outsized score drops when they break extraction. Adding a column, icon, text box, or table for Skills can reorder plain text so dates detach from employers or bullets merge into one line. Workday and Greenhouse still accept the upload. The checker reads what the parser would see, not what your eyes see.
How to tell this is yours: the drop followed a design change, not a wording change. Plain text export shows scrambled headers, missing employer names, or bullets run together. Special characters or bullets replaced with icons are another tell.
Before: Single-column Calibri PDF scored 82; readable export.
After: Added two-column Canva section for Skills; plain text showed employer line after Skills block; score 59.
Fix: Rebuild as single-column DOCX or PDF at 11-point Calibri or Arial. No text boxes. Standard round bullets. Export plain text and confirm Experience precedes Skills and dates sit next to employers. Then re-run the checker.
Cause 3: Keyword density spike
You added posting language everywhere to be safe. The matcher sees the same term twelve times in six lines and scores down for stuffing even though you were trying to help. Density spikes often follow a panic edit right after a first rejection email.
How to tell this is yours: score dropped after you pasted req language into Summary, Skills, and three bullets. Plain text reads fine but repeats jump out when you skim. Another tell: you added a keyword block footer or white text (never do this; some parsers still catch it).
Before: Three natural uses of stakeholder management across two bullets; score 77.
After: Pasted stakeholder management into every bullet plus Summary; score 64.
Fix: Keep three to four total uses of each hard requirement across the file. One in bullet one, one in a second bullet, optional once in Skills. Cut repeats before you cut proof. Read resume rejected by ATS when the portal rejected you and you need to separate filter failure from stretch reqs.
How to tell which cause is yours
Run a five-minute triage. Export plain text from the new PDF. If order is broken, start with Cause 2. If order is fine but terms live only in Skills, Cause 1. If order is fine and terms repeat in every paragraph, Cause 3. Compare against the old PDF if you still have it. The diff tells you more than staring at the score number.
When the score is high but callbacks still don't come, the problem may not be the number at all. Read why your resume scores high but you still get rejected for queue and match issues the checker cannot see.
Copy-paste score-drop triage block
Posting: [title + company] · Checker posting text pasted: [Y/N]
Version A score: [ ] · Version B score: [ ] · Delta: [ ]
Edit made: [wording / design / keyword paste]
Plain text order OK: [Y/N] · Required terms in Experience: [Y/N]
Repeat spam visible: [Y/N]
Diagnosis: [placement / layout / density]
Fix applied: [ ] · Re-run score: [ ]
Fill the block before you run version C. Naming the cause stops the loop where you undo good bullets because a layout tweak caused the drop.
If the posting asks for a letter, use the cover letter generator for structure after the resume is stable. Do not chase score on the letter while the PDF is still scrambling in plain text.
Chasing the score instead of the file
Reverting good bullets because the number dropped. If Cause 2 broke layout, fix the template. Don't go back to vague language that scored lower for the wrong reasons on version A.
Micro-editing ten times in one night. Each pass oscillates between placement and density. Diagnose once, fix once, re-run twice max.
Trusting the visual PDF over plain text. Accepted upload does not mean readable extraction. Export text every time the score moves more than a few points.
Stuffing Skills to recover points. You can inflate Skills match while Experience match stays weak. Recruiters ctrl-f under your last job title, not your chip row.
Comparing scores across different postings. An 85 on one req and a 70 on another is not a drop from your edit. It's a different keyword set.
Re-check before you reapply
Paste the same job description into the free resume checker after each layout fix. If Experience terms reappear in plain text and the score stabilizes within a few points, stop tweaking and submit.
Use build your resume when Cause 2 keeps coming back from template exports. A single-column builder output beats fighting a design tool that breaks extraction on every save.
Tools won't tell you which cause fired if you paste the wrong posting. Pair the checker with the triage block so version B gets a diagnosis, not another random synonym swap.
Stop the tweak loop tonight
Why your ATS score drops after small changes is almost never a mystery once you name placement, layout, or density. Fix the cause that matches your plain text export. Don't undo proof lines because a template broke extraction. Don't paste keywords into every line because the number wobbled once.
Open version A and B. Fill the triage block. Export text. Apply one fix. Re-run the checker on the same posting. If the score stabilizes and Experience carries the req terms, submit and move on to the next row in your tracker.
I've screened stacks of these files in Workday and Greenhouse, and the candidates who stop spiraling are the ones who diagnose the drop instead of chasing a perfect score on version F.
Read more
Frequently asked questions
Yes when the edit moves a required term out of Experience, breaks the PDF layout, or stuffs keywords into a block the parser reads as a table. Checkers score against the posting you pasted. A small wording change that looked fine to you may have shifted where the matcher finds terms. Re-export plain text and compare before you assume the tool is wrong.
Not automatically. Undo only if the drop came from a layout break or keyword removed from the wrong section. If you replaced vague language with posting terms and the score fell because density spiked, trim duplicates instead of reverting to weak bullets. The goal is a readable file that still carries required terms under Experience, not the highest number on a third tweak.
Many checkers weight terms under Experience higher than a Skills list. Moving keywords from bullets into Skills can raise a skills count while lowering match on the experience scan. Parsers in Workday and Greenhouse often read Skills after Experience or merge them oddly. Keep required terms in bullet one under your current role and treat Skills as support, not the main warehouse.
Sometimes. Heavy formatting, text boxes, icons, and column breaks can scramble export order so keywords land in the wrong section or disappear from plain text. The portal may accept the upload while the parser reads garbage. If your score dropped right after a Canva or template tweak, export to plain text first. If terms vanish, fix layout before you chase keyword density.
Submit when plain text export reads in order, required terms appear under Experience at least once, and a second checker pass on the same posting stays within a small band. Chasing a perfect score across ten micro-edits usually means you're oscillating between placement and density causes. Run the file once, fix the diagnosed cause, run again, then apply.