17 resume mistakes, sorted by who catches them: the software, the recruiter, or no one
We broke one resume 17 ways and ran each through our ATS matcher: what each resume mistake costs, which ones the software never sees, and what the law says.
Every list of resume mistakes gives you the same seventeen items with the same weight. Typos sit next to buzzwords, a photo next to a missing phone number, as if each one cost you the same thing. None of the six ranking pages we read puts a number on any of them, or says who actually notices.
So we measured. We took a real posting, wrote a clean resume for it, then broke that resume one mistake at a time and ran every version through the keyword matcher behind our own ATS checker. Some mistakes cut the match from 86% to 10%. Some cost nothing with the software and everything with the person reading. A few, nobody catches until the offer is already gone. This article sorts the seventeen by who catches them, with the numbers.
What each mistake cost in our test
The posting was Brex's Data Analyst II in New York, pulled from the Greenhouse job board on 4 October 2026. We fed its Responsibilities, Requirements and Bonus points sections to our free keyword extractor, which returned 12 terms (SQL, Python, Looker, Tableau, reporting, data analysis...). We kept 11 of them, dropping the one it labelled with the French word for "development", and added 10 more by hand from the requirements (dashboards, stakeholders, data quality, dbt, Snowflake...). 21 keywords in total.
Then we wrote Daniel Ortiz, a fictional data analyst whose 203-word resume honestly covers 18 of the 21. That is the baseline: an 86% match. Each variant below changes one thing. Where a mistake only touches structure (bullets, length, contact line), we also report what our formatting check deducts. That check runs on resumes built in the HiredFolio editor, where the structure is known; it starts at 100 and its penalties are listed in our formatting test.
| Mistake | Keywords found | Match rate | Formatting | Who catches it |
|---|---|---|---|---|
| Baseline, tailored and clean | 18/21 | 86% | 100 | |
| 1. Same generic resume for every job | 3/21 | 14% | 100 | Software |
| 2. Buzzwords instead of the posting's words | 2/21 | 10% | 100 | Software |
| 3. Typos in five keywords | 14/21 | 67% | 100 | Software |
| 4. Skills drawn as bars or icons | 12/21 | 57% | 100 | Software |
| 5. Too short (45 words) | 6/21 | 29% | 62 | Software |
| 6. No email, no phone | 18/21 | 86% | 72 | Software, then nobody |
| 7. Two columns, header text, spaced letters | see below | Software | ||
| 8. Duties instead of results | 18/21 | 86% | 100 | Recruiter |
| 9. Irrelevant skills added | 18/21 | 86% | 100 | Recruiter |
| 10. Every job you ever held (1,269 words) | 18/21 | 86% | 92 | Recruiter |
| 11. Paragraphs instead of bullets | 18/21 | 86% | 90 | Recruiter |
| 12. A photo | 18/21 | 86% | 100 | Recruiter |
| 13. Age, marital status, religion | 18/21 | 86% | 100 | Recruiter |
| 14. References and salary on the resume | 18/21 | 86% | 100 | Recruiter |
| 15. A wrong phone number | 18/21 | 86% | 100 | No one, until too late |
| 16. Inflated claims | 18/21 | 86% | 100 | No one, until too late |
| 17. Fonts and margins | 18/21 | 86% | 100 | Not a software mistake |
| Mistakes 1, 2, 3, 4, 6 and 11 combined | 2/21 | 10% | 42 | Everyone |
Mistakes 12 to 17 leave the text unchanged, so the matcher returns the baseline. That is the finding, not a gap in the test: a matcher reads words. A photo, a font, a lie and a wrong digit in a phone number are not words it can judge.
86% to 10%
same person, same jobs
of the posting's keywords found, before and after the bullets were rewritten in buzzwords. "Spearheaded cross-functional synergies" contains none of the words the employer asked for.
Source: HiredFolio, 12 variants of one resume against one posting, October 2026
Two caveats. This is our matcher, the one you can test for free; employers run many others. And in production our score uses a keyword list extracted by a language model, with the matcher's forgiveness for plurals, accents and a few hundred synonyms. The free extractor plus a hand-picked list is the reproducible version of that. The full inputs, every resume variant and the script are in the pull request that published this article.
The seven mistakes the software catches
These are the ones that change what a parser reads or what a matcher finds. They are also the only ones a tool can warn you about before you send.
1. Sending the same resume to every posting
Our generic version was not bad. It was Daniel's real jobs, described the way he would describe them to a friend: "prepared weekly reports and presentations for management", "ran queries and spreadsheets". It matched 3 keywords out of 21. The posting says SQL, Looker, stakeholders and data quality; the generic resume says queries, reports and departments.
The fix takes twenty minutes per posting and is mostly vocabulary. Our guide to resume keywords counts, across 93 real postings, which words recur for five common jobs.
2. Buzzwords where the posting's words should be
This was the single most expensive mistake we measured: 2 of 21 keywords, a 10% match. The buzzword resume still names the same two employers and the same dates. It just says "leveraged cutting-edge analytics to drive data-driven decision making" where the baseline says "wrote SQL (CTEs, window functions) over 40M-row transaction tables".
Example the same bullet, two ways
0 of the posting's keywords
Spearheaded cross-functional synergies to unlock actionable insights for key stakeholders
4 of the posting's keywords
Built 14 Looker dashboards for Sales, Operations and Finance stakeholders; weekly reporting pack replaced 6 manual spreadsheets
Buzzwords are not neutral filler. Every one of them occupies the place where a tool name or a skill from the posting could have been.
3. Typos in your keywords
We misspelled five keywords, once each, everywhere they appeared: Pyhton, SLQ, Lookr, Tabelau, dashbaords. Four of the five dropped out of the match, 86% down to 67%. The fifth, dashboards, survived only because our lexicon treats it as a synonym of "reporting", which was spelled correctly elsewhere.
A typo in a sentence costs you with the reader. A typo in a keyword costs you with both, because the search is literal: "Tabelau" is not a misspelling to a matcher, it is a different word. We tested 26 wording variants (plurals, acronyms, word order) in our keywords article; the misses were all differences a person would not notice.
4. Skills drawn as bars, icons or an image
Designer templates like to show skills as progress bars, star ratings or a graphic. We simulated that by removing the skills line, because an image contains no text for a parser to read. Six keywords went with it: Hex, dbt, data visualization, communication, analytical skills, GitHub. The match fell to 57%.
If a skill matters, it has to be a word on the page. "Excel: advanced" is read; a four-star Excel is not. France Travail, the French public employment service, gives the same instruction on its page about recruitment software: "Pas de graphiques de compétences (barres de progression, étoiles)", no skill graphics, progress bars or stars (France Travail).
5. Too short
Our 45-word version kept the job titles, the employers and a line of tools per job. It matched 6 keywords out of 21. There was simply not enough text to carry the posting's vocabulary, and on an editor-built resume our formatting check also deducts 20 points under 150 words, plus 8 for the missing summary and 10 for the missing bullets. We measured how many words fit on a page if you are unsure where you stand.
6. No email, no phone
The keywords did not move. The formatting check did: 20 points for the missing email, 8 for the phone. A resume with no way to reach you has scored itself out of the process, however good the match. Put both in the body of the first page, not in the page header; in our formatting test, contact details placed in a Word header vanished from the extracted text entirely.
7. A layout that scrambles the text
We did not re-run this one, because we ran 27 versions of it already. The short version from that test: a PDF sidebar interleaved with the experience line by line; a name and headings set with 0.15 em letter-spacing or in small caps came out as "D A N I E L O R T I Z"; a Word header disappeared. Margins, font size and line spacing changed zero bytes of extracted text. The full results, including the ones that did no harm (icons, tables, text boxes), are in Resume formatting: what software reads.
The seven mistakes only the recruiter catches
In our test, these left the keyword match at 86%. They are still mistakes. They are just mistakes a tool cannot warn you about, which is a reason to take them more seriously, not less.
8. Duties instead of results
"Responsible for Looker dashboards for Sales, Operations and Finance stakeholders" matches exactly as well as "Built 14 Looker dashboards; weekly reporting pack replaced 6 manual spreadsheets". Same words, same 86%. The matcher does not count numbers.
The recruiter does. So does the part of our ATS score that is not a word count: a language model judges experience fit, 25% of the final score, and it is told to look at level, industry and progression. Our guide to writing an effective resume covers the verb + number + result pattern.
9. Irrelevant skills
We added ten: Photoshop, forklift certified, barista, Final Cut Pro, typing 80 wpm. Match unchanged at 86%, formatting unchanged. Extra words cost nothing with our matcher, which only looks for the posting's terms. They cost the recruiter's attention, and they suggest you did not read the posting either. Our skills article, counted on 212 postings, is the shortlist.
10. Every job you ever held
We padded Daniel's resume with 28 earlier roles, up to 1,269 words. Match rate unchanged. Our formatting check deducts 8 points above 1,200 words, one of its mildest penalties. The recruiter is the one paying, in time, and the relevant bullets are now on page three.
11. Paragraphs instead of bullets
Same words rearranged into prose. Match 86%, formatting 90 on an editor-built resume (10 points for experience with no bullets). The parser read the paragraph fine. The person scanning the page for your last role's scope did not.
12. A photo
A photo changes nothing for the software; it has no text. In the United States it is a problem for the reader. The Equal Employment Opportunity Commission is explicit in its guidance on pre-employment inquiries: "employers should not ask for a photograph of an applicant" (EEOC). Many employers will not want to be seen with one on file. If you are applying in France, the convention is different and we cover it in the French CV format.
13. Age, marital status, religion
Same EEOC page: inquiries that may indicate age or religion "should generally be avoided" because they "may be used as evidence of an employer's intent to discriminate". Volunteering that information puts the recruiter in a position they did not ask to be in. Your date of birth, marital status and religion do not belong on a US resume.
14. References and salary requirements
"References available upon request" says nothing a recruiter did not assume. Named references cost a line and your referees' privacy before anyone needs them. A salary figure on the resume anchors the conversation before you know the range; if the application form requires one, how to answer salary expectations covers the form and the call.
Three mistakes nobody catches in time, and one that is not a mistake
15. A wrong phone number or a dead email
Our formatting check verifies that an email and a phone number are present. It cannot know the number is your old one. Neither can the recruiter, until the call does not connect. Read your contact line digit by digit, out loud, before every send.
16. Inflated claims
Nothing in a resume scan detects an invented skill or a stretched date. It surfaces at the reference check or in the first month on the job. Our test scored "Snowflake" as missing from Daniel's resume because he has not used it. Adding it would have raised his match to 90% and his risk to a level the match does not measure.
17. Fonts and margins: the mistake that is not one
Most lists tell you a 9 point font or a half-inch margin will "fail the ATS". In our 27-file test, 9 point with half-inch margins and 12 point with one-inch margins produced byte-identical text. Font and margin choices are for the reader's eyes. Pick them for a laptop screen and stop worrying about the machine.
Dos and don'ts, in one place
Resume dos and don'ts paired with what we measured
Don't
Send the resume you already have (14% match)
Describe yourself in buzzwords (10% match)
Draw skills as bars or stars (57% match)
Put your contact line in the page header
List duties without a number
Do
Reuse the posting's own words where they are true (86%)
Name the tool, the scope and the result
Write skills as text: "SQL · Python · Looker"
Put name, email and phone in the body, first lines
Give each bullet one verb, one number, one outcome
One rule behind the others: if the word is not on the page as text, in the employer's spelling, no software will find it.
The worst resume we could build, scored
For the bad-example section, we did not invent a Comic Sans resume. We combined the software-facing mistakes above into one file: generic wording, buzzwords, misspelled tools, skills in a graphic, no contact line, and prose instead of bullets. It is 115 words long and still about the same analyst with the same two jobs.
It matched 2 keywords out of 21. The formatting check scored it 42 out of 100: 20 for the missing email, 8 for the phone, 10 for the absent bullets, 20 for the length. Here is its first job, next to the baseline's.
The same role, twice worst version and baseline
Data Analyst, Northwind Payments, New York, 2023 to 2026
Spearheaded cross-functional synergies to unlock actionable insights for key stakeholders. Leveraged cutting-edge analytics to drive data-driven decision making at scale, using Pyhton, SLQ and Lookr dashbaords. Championed a culture of excellence across the organization.
Data Analyst, Northwind Payments, New York, 2023 to 2026
- Built 14 Looker dashboards for Sales, Operations and Finance stakeholders; weekly reporting pack replaced 6 manual spreadsheets
- Wrote SQL (CTEs, window functions) over 40M-row transaction tables; ad-hoc analyses turned into 3 self-service data tools
- Automated recurring reporting in Python, saving about 10 hours a week across the team
- Set up data quality checks on the revenue pipeline with the data engineers; reconciliation errors down 70%
Same person, same employer, same dates. One version contains the posting's words and the other contains adjectives.
The worst resume is rarely the ugly one. It is the one that reads smoothly and says nothing a matcher or a recruiter can hold on to.
Questions people ask
How many mistakes are acceptable on a resume?
Depends which. In our test, one misspelled keyword is one missing keyword, every time. One typo in a sentence the posting does not care about costs nothing with the software and a little with the reader. Zero is the target for the contact line and the keywords; elsewhere, a second reader catches what you no longer see.
What are the red flags on a resume?
For software: fewer than 150 words, no email, text the parser could not read or read scrambled (an image, a scan, a two-column PDF). For a recruiter: a date range that does not add up, a skill the bullets never show in use, and vocabulary that fits any job and therefore none. For both: a resume written for a different posting.
How much is too much?
Our scorer flags a resume above 1,200 words, which is roughly three dense pages, and deducts 8 points. We found no page limit in 6,383 private-sector postings when we counted them. The limit is the reader's patience, and the relevant content's position on the page.
Check it against the posting before you send
Most of the seventeen are invisible from inside your own file. You know what you meant by "analytics"; the matcher only knows whether "SQL" is on the page. Our ATS checker takes your resume as a PDF and the posting's text, no account needed, and returns your score, the first of the keywords it found and missed, and the first problems it detected; the full lists open when you sign in. It is the same matcher that produced the table above.
It will not tell you that your phone number is wrong, that your photo is a problem in Ohio, or that your Snowflake experience is a lie. Those three are yours.
Key takeaways before your next application
Rewrite every bullet that contains an adjective and no tool name; buzzwords took our match from 86% to 10%.
Spell every tool the way the posting spells it, and write skills as text, never as bars or icons.
Read your email and phone number aloud before sending; no software checks that they are right.
Two minutes, no account: check your resume against the posting.
Read next
- How to write an effective resume in 2026: structure, examples and mistakes to avoid
Build a resume that is clear, ATS-friendly and tailored to each posting: a fill-in outline, a full example, what screening software forgives and what it does not, edge cases and a FAQ.
- Nursing skills for your resume: 30 to list, and how 60 hospital postings word them
30 nursing skills for a resume, grouped the way the 2026 NCLEX-RN test plan groups nursing, plus how 60 real RN postings write BLS, ACLS and the license.
- Customer service skills for your resume: what 44 support job ads ask for
The customer service skills to put on a resume, counted from 44 live support job ads, with the wording an ATS keyword match misses and how to prove each skill.