What changed in your ATS check
Current engine version: 2.21.6. These notes explain changes to parsing, scoring and simulated parser output.
Compare scores using the same version
A higher score after an engine update does not necessarily mean your resume improved. Recheck both documents with the same version, job description, analysis tier and market. There is no universal conversion between versions, and no score predicts an interview.
Category weights stayed the same across these releases, but extraction and individual checks changed. Dates below come from the engineering release record; they are not deployment timestamps. Validation notes summarize that record and are not a claim that its full historical test corpus was rerun today.
v2.21.6 ·
Skills, jobs, schools and certifications are read as they are printed
Skills are read without the category names printed beside them, such as "Analytics: SQL", a skill that wraps onto the next line is read as one, and a second skills list is read too. A job, school or skill continued on the next page under a "(continued)" heading is read. A job printed with its employer and city above its title and dates is read as one job, and a city printed after an employer, a school or a job title is read as the place, not as part of the name. A certification's issuer and date are read from their own column, their own line or a bracket, even when the bracket wraps onto the next line, and a licence or standard number is no longer taken for a date. Lines such as "CERTIFICATIONS: …" and "LANGUAGES: …" under one heading are each read as their own section.
Effect on your results
Scoring formulas and weights are unchanged. Where your skills, schools or work history were hidden this way, they are now read, so a missing-skills or missing-experience finding no longer fires by mistake. Where a work history or summary is now read in full, its content is checked, so a score can also go down.
Recorded checks and limits
Regression tests cover each printed form. A re-read of 2,483 real resumes changed 91 scores: 70 up, mostly by 5 points where a missing-skills finding stopped firing, and 21 down by 1 to 6 points where a work history, summary or job is now read in full. On 96 downloads of the templates marked as tested with hiring software, the checker and the test behind that mark read the same jobs, schools, certifications and skills, and the seven newest designs score 95 to 99 on their own samples. These checks do not measure acceptance by an employer’s live ATS.
v2.21.5 ·
"Profile Summary" is read as your summary
A resume that opens with a "Profile Summary" heading has that section read as its summary, the same as "Professional Summary". Before, the summary was read as part of the header.
Effect on your results
No scoring rule or weight changed. Missing summaries already carry no points, so scores stay the same.
Recorded checks and limits
Regression tests cover the heading. A re-read of 2,483 real resumes found none using it, so no score moved. These checks do not measure acceptance by an employer’s live ATS.
v2.21.4 ·
Job and school lines are read whole, and findings name what they saw
An employer written before dates in brackets, such as "Helix Analytics (2021 – Present)", is read without leftover brackets, and a highlight on the employer lights the employer, not the job title. "Degree — School, Year" and "Degree at School" lines are read as a degree and a school. When skills sit under a heading of your own, such as "Data Stack Mastery", the skills finding names that heading. The dash tip counts only dashes inside sentences, never the dashes between a job title and its employer.
Effect on your results
No scoring rule or weight changed. The skills finding keeps its weight; only its wording and the place it points to change. The dash tip carries no points.
Recorded checks and limits
Regression tests cover bracketed dates, degree and school lines, skills under other headings and dashes in headers. A re-read of 2,880 PDFs, including 2,483 real resumes, moved no score. These checks do not measure acceptance by an employer’s live ATS.
v2.21.3 ·
Readable text and unfamiliar headings receive different guidance
Recovered text is checked separately from whether section headings were recognized. Short readable resumes receive length guidance, and visual skill ratings receive advice without losing points. Names before contact separators remain tied to the source text, including accented and non-Latin names.
Effect on your results
Scoring weights and allowances are unchanged. Affected scores may change when text recovery or a candidate name is read correctly. Unfamiliar headings and visual skill ratings carry no deduction.
Recorded checks and limits
Regression tests cover readable short resumes, unfamiliar headings, visual ratings and names before contact separators. These checks do not measure acceptance by an employer’s live ATS.
v2.21.1 ·
Job matches preserve what your experience actually says
Explicitly denied experience no longer satisfies a job requirement. Employment notes, student projects and combined skills and language headings retain their meaning. Credential completion, issue and expiry dates remain distinct. Actual unpaid work, projects and volunteering count as practical experience.
Effect on your results
Scoring weights are unchanged. Affected scores may change where earlier results counted denied experience or misread employment, projects or credentials. Related experience may receive partial credit without claiming an absent skill.
Recorded checks and limits
Regression tests cover denied and qualified experience, employment notes, student projects and credential dates. These bounded checks do not measure acceptance by an employer’s live ATS.
v2.21.0 ·
Career breaks and different education paths are treated fairly
Career breaks, concurrent roles and job changes no longer reduce consistency scores. Continuing consultancy is considered alongside shorter engagements, so overlapping jobs do not create a false break. A school or training record does not need to name a degree.
Effect on your results
Affected resumes may score higher when an earlier deduction came only from their career path or the absence of a degree. Missing school names, reversed dates and clearly future dates still receive their existing checks. Break guidance remains optional.
Recorded checks and limits
Regression tests cover concurrent work, continuing consultancy, year-only dates and education without a degree. These tests do not measure acceptance by an employer’s live ATS.
v2.20.0 ·
Completeness checks follow the resume you actually show
Saved-document checks use the sections visible in the selected design. Blank fields do not count as evidence, and practical experience can come from projects, coursework or volunteering. Missing summaries and headlines are optional suggestions.
Effect on your results
Hidden or blank sections no longer satisfy completeness checks, so affected scores may fall. A single supported skill is sufficient; there is no arbitrary minimum number of skills. Missing summaries and headlines do not cause deductions.
Recorded checks and limits
Regression tests cover hidden sections, empty fields, practical evidence without employment and optional summaries. These tests do not measure acceptance by an employer’s live ATS.
v2.19.1 ·
Occupational words and software requirements receive more precise checks
Requirements such as teaching and fundraising match regardless of capitalization. Ambiguous software names need supporting context. The writing checks distinguish valid costume and animal-care terms from genuine typos and preserve quantity ranges and restrictions in supported English constructions.
Effect on your results
Removing a false spelling warning may raise a score. Correcting an unsupported software match may lower a job-match score. The existing numerical golden score fixtures remain unchanged.
Recorded checks and limits
Authored regression tests cover ordinary requirements, occupational wording, software evidence and quantitative claims. They are limited examples, not a measurement of universal language accuracy or hiring outcomes.
v2.19.0 ·
Writing checks better preserve quantities and distinguish tool names
Rewrite checks look for quantities attached to the wrong activity, dropped negation and stronger ownership claims. Spelling and job matching use surrounding words to distinguish occupational terms and software names from everyday language.
Effect on your results
Spelling findings and job-match scores may change where the surrounding evidence changes the interpretation. A proposed rewrite can be withheld when it does not preserve the supported claim.
Recorded checks and limits
Authored tests cover metric swaps, negation, occupational spelling and ambiguous tool names. These are bounded checks and do not prove that every semantic change can be detected.
v2.18.0 ·
Honest experience does not need compulsory numbers or inflated responsibility
Specific qualitative work counts without a quota of numerical achievements. Accurate “helped” and “assisted” wording is no longer treated as weak writing. The editor no longer replaces a vague duty with a stronger ownership claim without evidence.
Effect on your results
Affected resumes may score higher when earlier deductions came only from missing numbers or honest supporting responsibility. Career-break advice carries no deduction. Existing checks for vague duties, spelling, repetition, missing essentials and factual date errors remain.
Recorded checks and limits
A comparison of 25 authored fixtures across 13 occupations left 23 content scores unchanged and raised two from 90 to 100. All 25 structure scores were unchanged. Separate tests cover career-break deductions. These results do not predict hiring outcomes.
v2.17.2 ·
Every school, each certification once, and the next page are read
Two or more schools listed without a blank line between them are now read as separate schools, whether each entry opens with the school, the degree or the date, and the school is named from the line that names it. Degrees written as "Masters", "Bachelors" or "Associates", or as abbreviations such as BSN, MSW, A.A.S., J.D. and GED, are read as degrees. A certification whose name wraps over several lines, or whose issuer and date sit on the line below it, is read once, with its issuer and date. A section heading repeated at the top of the next page, with or without "(continued)", is no longer read as an entry, and the entries below it are read.
Effect on your results
Scoring formulas and weights are unchanged. A school that was hidden inside another is now checked on its own: if it names no degree, the existing missing-degree finding applies. A degree written in the forms above no longer draws that finding.
Recorded checks and limits
Regression checks cover schools that open with a name, a degree or a date, two degrees at one school, a school written over two lines, wrapped and two-line certifications, and repeated headings. A re-read of 2,483 real resumes found 3,783 schools where 2,648 were read before, with scores moving by one or two points in 142 of them, and the exports of every template read each sample certification once. These checks do not measure acceptance by an employer’s live ATS.
v2.17.1 ·
A summary without a heading, and clinical section headings, are read
A summary written as a paragraph directly under the name, title and contact details, with no "Summary" heading above it, is now read as the summary. A paragraph under another label such as "Highlights", or one that opens with a date or a list label, is not. "Clinical Skills" headings are read as skills, and "Licensure" or "Licensure & Certification" headings as certifications. Skills separated by a spaced middle dot (" · ") are read as separate skills.
Effect on your results
Scoring formulas and weights are unchanged. A resume whose summary has no heading no longer loses points for a missing summary, and its summary now receives the same content checks as one under a heading. A clinical resume whose skills or licences sit under those headings no longer draws a missing-skills finding.
Recorded checks and limits
Regression checks cover summaries under one-line and multi-line headers, labels, lists, dated openings and headed summaries. A re-read of 2,484 real resumes found 29 more summaries and lost or changed none. These checks do not measure acceptance by an employer’s live ATS.
v2.17.0 ·
Reading problems are charged only when the text shows them, and more real problems are named
The checker no longer takes points for a font or layout that every reader still reads correctly. Bullets drawn with a symbol font, dates printed at the right edge and two-column headings are now read with their jobs. It now names objective statements, referees’ phone numbers and e-mails, an e-mail address with extra words beside the name, a claimed count the resume contradicts, and more clichés, including sales and British phrasing. The editor’s own quick check no longer deducts for a missing summary and labels its score as a quick check of the copy.
Effect on your results
Resumes that lost points for problems no reader actually had may score higher. An objective statement or printed referee contact details each cost 2 points. The e-mail and claimed-count notes cost nothing. Category weights are unchanged.
Recorded checks and limits
Regression checks run seven real test resumes with planted mistakes through the full checker and pin each catch. A re-read of 2,483 real resumes found reading warnings on 10 files, down from 178, and samples of every new finding were read by hand. These checks do not measure acceptance by an employer’s live ATS.
v2.16.10 ·
Jobs written as "Title • Company • City" are read correctly
A job written on one line with bullets between its parts, such as "Nursing Assistant • Harbor Crest Rehabilitation Center • Fresno, CA", is now split into its title, employer and city. Before, the whole line was read as the title, so the checker reported a missing employer, or cut a title with a comma short and read the city as part of the employer. A city at the end of a job line is never read as the employer.
Effect on your results
Scoring formulas and weights are unchanged. Affected resumes lose false "missing company or job title" findings, so their scores may rise. A job line that names a city but no employer now draws the missing-employer finding, so that score may fall.
Recorded checks and limits
Regression checks cover bullet-separated, bar-separated, employer-first, undated and city-only job lines. A re-read of 1,031 jobs exported through 22 templates and of 2,483 real resumes found no job that lost its title or employer. These checks do not measure acceptance by an employer’s live ATS.
v2.16.9 ·
Free checks explain incomplete resumes and recognize more experience headings
The free Builder preview now applies the existing 45-point limit to resumes under 100 words that are missing practical evidence, education or training, or skills. The result names what is missing. A paid server check is not needed to see the limit. Uploads now recognize Career Experience, Employment Experience and Industry Experience headings. Advice for short resumes includes projects and coursework without prescribing a word target.
Effect on your results
Short incomplete local previews may score lower. Concise complete resumes, including graduates with projects or volunteering, retain their measured scores. Newly recognized jobs and bullets may raise an upload score. Category weights and length deductions are unchanged, and potential recovery counts removal of the limit once.
Recorded checks and limits
Regression checks compare the local and server completeness rules, cover incomplete and concise complete resumes, and exercise three-job text imports with phone numbers and financial outcomes. These checks do not measure acceptance by an employer’s live ATS.
v2.16.8 ·
Incomplete resume scores are explained and summary wrapping stays harmless
A short resume missing practical evidence, education or skills receives a score limit of 45 with the missing details named. Complete compact resumes are not capped just for being short, and projects or volunteering can supply practical evidence without paid employment. The report shows the category total and any completeness deduction separately. A concise summary is no longer called too short after it already meets the sentence guidance, and PDF line breaks or qualifications such as B.Sc. do not count as extra sentences.
Effect on your results
Short incomplete resumes may score lower. Complete short resumes retain their measured score. Correcting false summary findings may raise a score. Potential improvement includes removing a completeness limit only by adding the missing truthful evidence. Existing category weights are unchanged.
Recorded checks and limits
Regression checks cover sparse and complete short resumes, graduate projects and volunteering, wrapped summaries, qualifications, accented sentence starts and repeated scoring of identical normalized input. These checks do not measure acceptance by an employer’s live ATS.
v2.16.7 ·
Technical wording stays correct and motion design needs creative evidence
Valid technical uses of "preformed", such as medical splints and manufactured inserts, no longer draw a spelling finding or an automatic replacement with "performed". A bare "Motion" skill with an engineering design title or degree no longer counts as creative motion design, and a title such as "Designated Safety Officer" supplies no design evidence. Creative roles and explicit motion-design or motion-graphics skills still count. The browser job match and server checker use the same rule.
Effect on your results
Scoring formulas and category weights are unchanged. Removing a false spelling finding may raise an affected resume’s score. Removing an unsupported motion-design match may lower its job-match score. Cached assessments are refreshed so results use the corrected rules.
Recorded checks and limits
The engineering release record describes the technical-wording and creative-evidence corrections and reports unchanged golden scoring fixtures. These checks are not evidence of acceptance by an employer’s live ATS.
v2.16.6 ·
One-line job headers and skill rows are read correctly
A job written on one line as "Title at Company" or "Title @ Company" is split into its title and employer, and a title with a comma such as "Charge Nurse, Medical ICU" or a hyphen such as "Co-Founder" stays whole. Skill names such as PostgreSQL or SolidWorks that are separated only by spaces are no longer reported as words merging during extraction, bold text or a right-aligned date is no longer reported as PDF parsers disagreeing, a word such as "Co-op" in a dated job line is no longer reported as text out of order, the word "constructor" is no longer flagged as a misspelling, and a job titled with a role such as Nursing Assistant, Delivery Driver or Physical Therapist is no longer merged into the job above it.
Effect on your results
Scoring formulas are unchanged. Affected resumes lose false "missing company or job title", "words are merging", "parsers disagree", "out of order" and misspelling findings, so their scores may rise. Writing gRPC, OpenTelemetry or SolidWorks in lower case now draws the same capitalisation suggestion as other technology names.
Recorded checks and limits
The engineering release record lists the header formats and skill rows covered, and reports unchanged golden scoring fixtures. This is not evidence of acceptance by an employer’s live ATS.
v2.16.5 ·
Job match stops finding skills that are not there
Job match no longer counts a date such as "2020 – Present" as presentation skills, or short words such as "do", "AP" or "PM" as software, accounting or project-management skills, and an item under a "Preferred" heading in a job post is read as preferred. The checker reads a name written over two lines as one name, finds a location that sits behind an icon, no longer calls an icon between two contact details a collision, and no longer suggests adding a summary to a resume that has a short one. A Word text box drawn at the top of the page is read first.
Effect on your results
Scoring weights are unchanged. Without a job description, affected resumes can only rise: a short summary now counts, and a location behind an icon is found. With a job description, a requirement that was matched only by a date or a short word no longer counts, so those match scores can fall to what the resume shows.
Recorded checks and limits
The engineering release record lists the replayed tester files and the corpus measurements, and reports unchanged golden scoring fixtures. This is not evidence of acceptance by an employer’s live ATS.
v2.16.4 ·
Names in page headers are read, and job match shows its evidence
A name and contact line placed in a Word page header or text box is read correctly, so the checker no longer names you after your first heading or reports your contact details missing. A PDF whose header was drawn last is read top to bottom. A degree dated "2015 – 2019" graduates in 2019. Job match shows, for each requirement, the line of your resume that meets it, finds experience written in different words (such as Medical-Surgical for med-surg, or Mentored for mentoring), and says it could not check a requirement instead of calling it missing when a posting could not be read.
Effect on your results
Scoring formulas are unchanged. Affected resumes lose a false "missing contact information" finding, so their scores may rise. With a job description, requirements you meet in different words now count as matched, so match scores may rise.
Recorded checks and limits
The engineering release record lists the file layouts and job-posting wordings covered, and reports unchanged golden scoring fixtures. This is not evidence of acceptance by an employer’s live ATS.
v2.16.3 ·
Indian education and project headings are recognised
The parser recognises headings such as Academic Qualifications, Internships and Academic Projects. Personal Details, Declaration and other supporting sections stay separate from the section above. Indian degree names and education scores retain their meaning during extraction.
Effect on your results
Affected resumes may receive different findings when their existing education, projects and internships are read correctly. A CGPA scale or percentage is an education result, not a resume-quality score.
Recorded checks and limits
The engineering release record lists Indian degree names, CGPA scales and percentage formats covered by the parser update, and reports unchanged golden scoring fixtures. This is not evidence of acceptance by an employer’s live ATS.
v2.16.2 ·
Common job headers stay together during import
The parser keeps a job title with its company, location and dates across common one-line and multi-line header formats. This avoids splitting one job into incomplete entries or losing the next job title.
Effect on your results
Scoring formulas are unchanged. Affected uploads may receive different findings and scores because their job history is now read correctly. Previously cached results are invalidated.
Recorded checks and limits
The release record describes corrections for dated company lines, combined title and employer headers, and date lines placed between a title and company. Department qualifiers remain part of the job title.
v2.16.1 ·
Unavailable analysis no longer becomes a score
If a required check fails, a job description cannot be parsed, or a posting contains no extractable skill requirements, the checker returns an analysis-unavailable error instead of grading the applicant using a substitute value.
Effect on your results
Completed assessments keep their existing formulas unchanged. Previously cached results are invalidated because unavailable analysis is no longer presented as a successful score.
Recorded checks and limits
The release record retains the existing fixed-score expectations. Extracted text remains visible when scoring is unavailable; extraction confidence is not used as a resume-quality score.
v2.16.0 ·
More sections visible in parser simulations
Languages, awards, publications and other supported sections now appear in the simulated parser output. Structure that a simulation cannot confidently retain is disclosed as an advisory.
Effect on your results
Overall scores and parser compatibility scores are unchanged. These new advisories carry no scoring penalty.
Recorded checks and limits
The release record documents scoreless advisories and unchanged compatibility totals. These are simulations, not tests against employers’ live systems.
v2.15.1 ·
Wrapped sidebar headings stay together
A heading split across two lines in a narrow sidebar is recognized as one section instead of two incomplete sections.
Effect on your results
Scores may change for uploads with wrapped headings because the extracted structure changes.
Recorded checks and limits
The release record reports 24 of 24 strict typography cases passing. Flat-layout corpus results were unchanged.
v2.15.0 ·
Additional career sections count toward analysis
Supported sections such as volunteering, languages and professional development now contribute their real text to the analysis, including job-description matching.
Effect on your results
Scores may change when previously omitted sections contain relevant evidence. Category weights are unchanged.
Recorded checks and limits
Existing fixed-score samples remained unchanged, but did not contain these additional sections. That result alone does not establish coverage for them.
v2.14.0 ·
Word heading styles preserve custom sections
Word documents with explicit heading styles can retain unfamiliar section names instead of merging their content into the preceding section.
Effect on your results
Scores may change for Word uploads whose section boundaries are recovered.
Recorded checks and limits
The release record describes heading-style confirmation in Word extraction. Plain text has no heading styles and cannot use this signal.
v2.13.1 ·
Sidebar headings retain their typography
Two-column PDF headings no longer lose their size signal when text from both columns shares a horizontal line.
Effect on your results
Scores may change for PDF uploads with affected sidebar headings.
Recorded checks and limits
The release record reports 23 of 23 typography cases passing, including sidebar headings and rejection cases. Fixed-score samples were unchanged.
v2.13.0 ·
PDF heading size helps recognize custom sections
Larger text can confirm an unfamiliar section heading. Bold text alone is insufficient, avoiding false headings for ordinary job titles and body text.
Effect on your results
Scores may change for designed PDFs whose custom sections are now recognized.
Recorded checks and limits
The release record explicitly notes that the 2,484-PDF corpus is typography-flat and cannot measure this feature’s recall. Separate typography cases are needed.
v2.12.0 ·
Unknown headings stop creating false entries
Recognized custom section boundaries reduce cases where a heading became a fake degree, skill or experience bullet.
Effect on your results
Scores may change when a corrected extraction removes invented structure.
Recorded checks and limits
The release record documents a 2,484-PDF extraction corpus and negative cases for false headings. Fixed-score samples alone do not exercise upload extraction.
v2.11.0 ·
Custom section text reaches the scorer
Text in user-defined sections can now support keyword matches and other content checks instead of being omitted.
Effect on your results
Scores may change for resumes containing custom sections. Older scores did not see that text.
Recorded checks and limits
The release record identifies that the three existing fixed-score samples had no custom sections, so their unchanged scores do not prove this change has no effect.
v2.10.0 ·
International contact details are recognized
Phone and location extraction handles more national formats. Recognized international section labels also preserve more resume structure.
Effect on your results
Scores may change when valid contact details or headings were previously missed.
Recorded checks and limits
The release record gives UK, Indian and German phone formats as concrete cases. Recognition is broader, not a guarantee for every format.
v2.9.0 ·
Personal-detail guidance follows the selected market
Date-of-birth and marital-status guidance now follows the selected resume market, when those fields are present.
Effect on your results
Scores may change for resumes carrying these fields. This is document-convention guidance, not an assessment of the person.
Recorded checks and limits
The release record reports unchanged fixed US samples, which contain neither field. Those samples do not validate every market.
v2.8.0 ·
Photo and nationality guidance becomes market-specific
Resume-convention checks use the selected market rather than applying one set of norms everywhere.
Effect on your results
Scores may change for resumes with photos or nationality details under the selected market’s rules.
Recorded checks and limits
Fixed US samples without photos were unchanged according to the release record. Market conventions are guidance, not hiring predictions.
v2.7.1 ·
Common month-year dates stop triggering false warnings
Parser simulations accept standard English month names and abbreviations, including a trailing period, when paired with a valid year.
Effect on your results
Affected parser-family results may change. The main scoring weights and fixed-score samples are unchanged.
Recorded checks and limits
The release record includes Jan 2020 and Sep. 2024 as accepted examples. Unsupported labels such as Spring 2026 still receive a warning.
v2.7.0 ·
Writing suggestions and checks share a vocabulary
The writer and scorer use the same language guidance, reducing cases where a generated phrase was then criticized by the checker.
Effect on your results
Scores may change for documents containing the affected phrases or repeated writing patterns.
Recorded checks and limits
Three fixed-score samples retained scores of 69, 71 and 95 in the release record. They are short, clean examples and do not trigger every new check.
v2.6.0 ·
Extraction artifacts stop looking like resume defects
Checks distinguish text joined by PDF extraction from actual writing problems, accept normal date ranges, and avoid penalizing absent optional sections.
Effect on your results
Scores may change when earlier findings were caused by extraction artifacts or optional sections.
Recorded checks and limits
The release record describes reproduction using real PDF extractors. A changed score can reflect a corrected check rather than an edited resume.
v2.5.0 ·
Scoring recognizes evidence across professions
Job matching and content checks no longer rely only on software-industry vocabulary. Clinical, trades and other professional evidence can contribute.
Effect on your results
Scores may change for career evidence previously missed by narrow vocabulary lists.
Recorded checks and limits
The release record adds job-match and content tests and broadens skill-parity checks beyond software terms.