# ATS Platform Playbook What each mainstream hiring system actually does to your resume, and what to do about it. Researched August 2026 against vendor documentation, live application forms, and federal rules. **How to read this file.** Every claim here is either traceable to a vendor's own documentation, a live form inspection, or a federal rule, or it is marked *[inferred]* or *[unverified]*. Numbers that circulate widely with no method behind them are named and buried in the last section rather than repeated. If a claim here disagrees with something you read elsewhere, the disagreement is usually because the other source is repeating 2018 advice. --- ## 1. Find out which system you are submitting into Look at the URL behind the Apply button. It names the vendor. This takes five seconds and almost nobody does it. | Host pattern | Platform | |---|---| | `*.wd1.myworkdayjobs.com`, `*.wd5.myworkdayjobs.com` | Workday | | `taleo.net` | Oracle Taleo **Enterprise** | | `tbe.taleo.net` | Taleo **Business Edition**, a different product | | `*.oraclecloud.com`, `fa-*.fa.*.oraclecloud.com` | Oracle Recruiting Cloud | | `job-boards.greenhouse.io`, `boards.greenhouse.io` | Greenhouse | | `jobs.lever.co` | Lever | | `jobs.ashbyhq.com` | Ashby | | `*.icims.com`, often inside an iframe on the employer's own domain | iCIMS | | `jobs.smartrecruiters.com` | SmartRecruiters | | `jobs.jobvite.com` | Jobvite | | `career4.successfactors.com`, `career5.successfactors.eu` | SAP SuccessFactors | | `*.avature.net` | Avature | | `recruiting.ultipro.com`, `recruiting2.ultipro.com` | UKG Pro | | `*.dayforcehcm.com/CandidatePortal` | Dayforce | | `workforcenow.adp.com/mascsr/...` | ADP classic career center | | `myjobs.adp.com//` | ADP newer branded site | | `paycomonline.net/...?clientkey=` | Paycom | | `recruiting.paylocity.com/recruiting/jobs/...` | Paylocity | | `applytojob.com`, `app.jazz.co` | JazzHR | | `.bamboohr.com/careers/` | BambooHR | | `apply.workable.com` | Workable | | `governmentjobs.com` | NEOGOV, US state and local government | | `sjobs.brassring.com` | IBM Kenexa BrassRing | | `usajobs.gov` | US federal, entirely different rules | If the host is none of these, scroll to the bottom of the application page. Most platforms put their branding there. **Two facts that compress everything else.** Only a handful of parsing engines sit under most of the market. **Textkernel**, owned by Bullhorn since June 2024, parses Oracle Recruiting Cloud, SmartRecruiters, SAP SuccessFactors and Bullhorn. **DaXtra** parses iCIMS, Jobvite and Zoho Recruit. **Workable runs a large language model** (Google Gemini). Everything else is undisclosed. So the real question is not "which of forty systems is this" but "which parser generation am I feeding," and there are three. --- ## 2. What actually rejects you Ranked by how many real candidacies each one ends. **1. A structured answer to a knockout question.** This is the only thing in the entire stack that is documented, unambiguous, automated rejection. It is confirmed in vendor documentation for Greenhouse, Ashby, Workable, Recruitee, Breezy, SmartRecruiters, Jobvite, SAP SuccessFactors, Oracle Taleo and Oracle Recruiting Cloud. The questions are always the same handful: sponsorship, work authorization, location and relocation, licensure, minimum years of experience, shift availability, salary expectation, willingness to travel. There is no partial credit and no human reviewing the borderline case. At Greenhouse, Jobvite and Workable there is no notification either. Greenhouse's own documentation states that people set up to receive new-application alerts "will not be notified of auto-rejected applicants," so following up about your application can reference a record no human ever saw. **2. The application you never actually submitted.** Structurally invisible and probably enormous. Handshake counts you as an applicant when you click through to an external system, whether or not you finished there. Paycom puts a mandatory electronic signature at the very end, so "saved" is not "submitted." USAJOBS hands you off to the agency's own system, and stopping at the hand-off means you did not apply. A failed LinkedIn Apply Connect webhook retries six times and then silently converts the job to off-site apply. **3. Queue position.** Systems that score your resume mostly do not reject anyone. They order a list. Under real volume, sorting by score and working top down produces the same outcome as rejection without a rejection event ever occurring. **4. Not being findable in the search a recruiter runs.** This is where formatting damage actually lands. An empty Workday Skills section makes you unfilterable. A resume with no text layer removes you from iCIMS boolean search across that employer's entire candidate database. Indeed's profile default is "Employers can't find you," which keeps you out of the largest resume database in the US until you flip one switch. **5. A human reading it and passing.** The largest bucket by volume, and the one everything above exists to get you to. **The single accurate sentence:** your resume is not scored against a rejection threshold. It is parsed into structured fields that determine whether you are findable and legible, while a separate set of structured questions is the only thing that eliminates anyone automatically. Bad formatting does not get you rejected. It gets you silently omitted from the search a recruiter runs. --- ## 3. The universal safe format Every rule here has a documented mechanism behind it. 1. **A real text layer.** Not a scan, not a photo, not a flattened design export. SAP documents that scanned PDFs "will not parse as expected." Breezy's screening cannot parse scanned resumes and returns nothing rather than a low score. Ashby marks an unreadable file `skipped`, which means no criteria evaluated and no signal to you. Some systems have OCR, but it is usually a separately licensed add on, so you cannot know from outside whether the employer bought it. 2. **Single column, linear reading order.** Extraction flattens the page top to bottom, left to right, and a sidebar interleaves into the main column. The only real measurement anyone publishes is Textkernel's own engineering data: 62% to 90% well-rendered after they built explicit column handling. Best in class still mishandles roughly one column resume in ten, and most systems are not best in class. Single column costs you nothing. 3. **Contact details in the document body, never in a Word header or footer.** In the underlying Word file format, headers and footers are separate parts from the main document body. An extractor that reads the body never sees them. Your name, phone and email vanish and the resume otherwise parses perfectly, which is what makes it so expensive. 4. **One date format, month and year, held consistently.** Every enterprise system coerces parsed dates into structured From and To controls. Workday's HiredScore layer computes employment gaps and tenure from parsed dates and shows them to the recruiter as computed fields, so a misparse manufactures a gap you never had. 5. **Conventional section headings.** Parsers pattern match on heading strings. Creative headings fail both the structural test and the keyword fallback at the same time. 6. **Skills stated inside role bullets, not only in a standalone block.** SmartRecruiters extracts skills from employment history, education and body text. A detached keyword block contributes less than the same terms sitting inside evidence. 7. **Every hard requirement evidenced explicitly and separately.** Ashby's application review returns a binary Meets or Does Not Meet per criterion with citations back to your text. An implied qualification scores as Does Not Meet. Implication is not evidence to a citation based evaluator. 8. **Education, skills and experience legible as three distinct blocks.** ADP computes three separate sub scores and weights them per sector. A resume that buries education is under scoring a component it could have won outright. 9. **Do not bundle resume, cover letter and references into one file.** Recruitee names this specifically as a parse killer. Attachment slots exist for a reason. 10. **No password protection, encryption, or digital signature.** On Recruitee's published failure list, and explicitly forbidden by USAJOBS. 11. **Keep the file small. 2 MB is the safe ceiling, 1 MB is safer.** SmartRecruiters documents a 2 MB API limit. Taleo defaults to 1 MB. iCIMS extraction times out at 90 seconds with two attempts. A resume over 1 MB almost always contains images, which is the real problem. 12. **Standard bullet glyphs and standard punctuation.** Jobvite, verbatim: "Avoid special characters as bullets, or risk having resume content deleted, including diamond shapes, scrollwork, color-coordinated square boxes, checkmark boxes, or arrowheads." Note the mechanism. Content gets deleted from the extraction, not rejected. 13. **A mainstream, properly embedded font.** The failure mode is a missing character map producing garbled text, not aesthetics. Related and worth knowing: standard ligature glyphs mean words like "certified", "qualified" and "efficient" can extract as characters that never match the keyword a recruiter searched. 14. **Fill the structured fields by hand after any autofill.** Not a formatting rule, but the highest leverage action in the whole process. Workday's admin documentation states parsing does not fill Languages or Skills, ever. SAP fills only standard fields and only empty ones. Whatever the parse missed is what the recruiter's filter cannot find. 15. **Put a version marker in the filename.** UKG stacks re-uploaded resumes with the newest at the bottom, so a recruiter scanning top down sees your oldest file first. 16. **Avoid `.htm` and `.html` even where the uploader accepts it.** Workday's uploader takes it; the HiredScore grading layer's supported list does not include it, and an unsupported format can leave you ungraded, which drops you out of the sorted queue recruiters work first *[the intersection reasoning is inferred from two official lists]*. ### Cargo cult, killed by verification | What you were told | What is actually true | |---|---| | "One page maximum" | **No ATS counts pages.** Every documented limit is a byte size. The only real page cap runs the other way: USAJOBS caps federal resumes at two | | "Never use a PDF" | Greenhouse's live form lists PDF first. USAJOBS affirmatively recommends PDF. Workable's parser handles image based CVs. Every DOCX versus PDF fidelity percentage in circulation is fabricated and they contradict each other | | "Use exactly this font or be rejected" | Font guidance is real and narrow. Font caused *rejection* has no source anywhere | | "ATS can't read bullets, bold or italics" | No vendor states that bold or italics impair parsing. The claim has no source at all. Standard bullets are fine. Only decorative glyphs are documented as harmful | | "Match 80% of the job description's keywords" | Nothing rejects on keyword density. Vocabulary alignment moves a rank, not a gate. SmartRecruiters processes only the Title, Description and Qualifications fields, deliberately excluding Company Description, so mirroring the "About Us" section is wasted effort | | "Name drop prestigious employers and schools" | SmartRecruiters strips employer and university names from ranking generation | | "Recruiters only read the parsed fields" | False, and so is the opposite. The original file is retained and opened. The parsed fields drive the grid, facets and filters. Both audiences are real | --- ## 4. File format decision rule Build three artifacts once, then choose per target. - `Lastname_Firstname_Resume_2026-08.docx`: single column, conventional headings, no tables, no text boxes, nothing in the header or footer, under 1 MB - `Lastname_Firstname_Resume_2026-08.pdf`: exported from that DOCX, text layer verified - `Lastname_Firstname_Resume_2026-08.txt`: the emergency fallback, one column by definition **Send DOCX when** the parser is classical: iCIMS, Jobvite, Zoho Recruit, Taleo, JazzHR. Also for Workday, where both work and DOCX carries no downside. Also to any staffing agency, because they reformat your resume onto their own letterhead before submitting you and DOCX saves them the work. **Send a text layer PDF when** the parser is modern: Oracle Recruiting Cloud, SmartRecruiters, SAP SuccessFactors, Workable, Greenhouse, Lever, Ashby. Do not degrade a well built document out of generic fear. Also send PDF to USAJOBS, which recommends it, and to BambooHR, where a human reads it on a phone. **Send `.txt` when** you have evidence the parse failed, or when a recruiter asks you to paste into an email body. Paste from the `.txt`, never from Word. **Never send** a file with no selectable text layer, anywhere, ever. Never `.pages`. Never a file whose extension you changed by renaming, because Jobvite validates on detected file type rather than extension. Never anything over 5 MB. One more, from Textkernel's own integration guidance: send the original file. Not a copy paste, not a conversion by some other tool, not a scan. Exporting your DOCX to PDF is fine. Rebuilding it somewhere else is what hurts. --- ## 5. Layout killers, ranked by real damage 1. **No text layer.** Total loss. Uploads fine, displays perfectly to a recruiter, and you are absent from every search. 2. **Multi column or sidebar.** The highest damage item that still returns plausible looking output. Titles land next to skills, dates detach from employers. 3. **Contact details in a header or footer.** Catastrophic when it fires, because the parse produces a record with no way to contact you. 4. **Tables used for layout.** Cell text extracts in row order, so a two column skills table flattens into nonsense. A table used as an invisible layout grid is indistinguishable from a real column layout. 5. **Text boxes.** Separate content streams in both Word and PDF. Commonly skipped entirely rather than mis-ordered, so the loss is silent and total. 6. **Several documents merged into one file.** A parser looking for one resume finds three documents' worth of contradictory dates and headings. Almost nobody warns about this. 7. **Graphics, icons, logos and skill rating bars used instead of text.** A five star Python rating is zero characters of extractable text. 8. **Decorative bullet glyphs.** Documented content deletion, not rejection. 9. **Letter spacing used for visual effect.** Breaks the rule parsers use to join characters into words, so "EXPERIENCE" extracts as "E X P E R I E N C E" and matches no heading. 10. **Exotic or badly embedded fonts.** Garbled extraction from a missing character map. 11. **Password protection or digital signatures.** Total, and trivially avoidable. 12. **Oversized files.** Either a visible upload failure, which is recoverable, or a silent empty parse, which is not. 13. **Creative section headings.** Zero upside to the risk. 14. **Tracked changes and comments.** Unaccepted changes can extract as both the inserted and the deleted text, so your resume arrives with duplicated or contradictory content. Accept all changes and delete all comments before you send. One thing commonly misdiagnosed as a parser failure: on Workday, each role has a **single Role Description free text field**, so a five bullet block collapses or truncates no matter how you format it. Lead every role with its strongest, most self contained line. --- ## 6. Headings and dates **Use these strings.** `Professional Experience` (or `Work Experience` or `Experience`), `Education`, `Skills` (or `Technical Skills`), `Certifications`, `Professional Summary`, `Languages`, `Projects`, `Publications`, `Security Clearance`, `Military Service`. **Never use** narrative headings. `Where I've Made an Impact`, `My Journey`, `What I Bring`, `Toolbox`, `Superpowers`, `Academic Journey`. Jobvite names `Career Experience` specifically as worse than `Professional Experience`, which is how narrow the tolerance is presented as being. **Dates. Pick one format and never deviate.** `January 2020 to March 2023`, or `Jan 2020 to Mar 2023`, or `01/2020 to 03/2023`. - **Month and year on every role.** Not optional. Year-only ranges degrade tenure and gap computation everywhere and are a leading cause of a federal resume being rated ineligible. - Write `Present` as a word, consistently, on every current role. - One separator throughout. If you use a dash, use the same one every time. A plain hyphen is the lowest risk character because it is plain ASCII and cannot be mangled by a character map problem. Writing `to` avoids the question entirely. - Put dates on the same line as employer and title, in the same order on every entry. - Separate rows for separate roles at the same employer. Promotions collapse into one entry on Workday routinely, and the promotion is exactly the signal a recruiter is looking for. - **Never** use seasons, bare years for jobs, two digit years, or mixed formats in one document. Every attested date failure traces to inconsistency rather than to any particular format. Honest caveat: **not one vendor in this document publishes its accepted date formats or normalization behavior.** The symptoms are well attested. The mechanism is not. These rules are consistency hedges, not documented requirements. The exception is USAJOBS, where month and year and hours per week are stated requirements. --- ## 7. Per-platform playbooks ### Workday 1. **Fill the Skills section by hand. Always.** Workday's own admin documentation states parsing does not fill Languages or Skills. Skills facets are how a recruiter filters the pool. 2. **Treat autofill as a draft and audit every row before Submit.** Check for employer and title inversion, blank dates, and promotions collapsed into one entry. Applications are immutable after submission. 3. **Attach a resume file even if you typed everything.** If the employer runs HiredScore, no file means no grade, and an ungraded candidate is absent from the sorted queue. 4. **Upload everything on the My Experience page.** It is the only page where attachments can be added, and missing documents make the application incomplete. 5. Expect a fresh account per employer. Same email is fine, the accounts are separate, and recruiters cannot reset your password. ### Oracle Taleo 1. **Check the host.** `tbe.taleo.net` is Business Edition, a different product from the `taleo.net` Enterprise sites that almost all online Taleo advice describes. 2. **Answer prescreening as if it were the whole application,** because for pass or fail purposes it is. A disqualification answer exits you instantly and your resume is never consulted. 3. **Claim the optional criteria deliberately.** Meeting only the required ones puts you in the minimally qualified bucket, which is invisible when a recruiter filters for top candidates. 4. **Spell out acronyms.** Taleo's recruiter search does not expand abbreviations. Write `Project Management Professional (PMP)`, not `PMP certified`. 5. **Go back into your profile after submitting and fix what it captured.** Parse errors land in fields the apply flow never showed you. 6. Under 1 MB, DOCX, single column. And do not read tea leaves in the status string. ### Oracle Recruiting Cloud 1. **Use the preview step.** It lets you include or exclude parsed skills before commit. This is the single biggest candidate side improvement over Taleo and most people click straight past it. 2. **Verify Work Experience and Education specifically.** Oracle maintains a support article precisely because parsing intermittently fails to populate those two sections. 3. **Treat every choice question as scored.** Oracle's guidance to employers is to use negative scores for disqualifying answers, and the feedback showing which answer killed you is explicitly not visible to candidates. 4. A text layer PDF is fine here. The parser is materially better than Taleo's. ### Greenhouse 1. **Answer every knockout dropdown truthfully and deliberately.** Auto-Reject is a shipped feature that fires on your answers with no human involvement and no notification to anyone. 2. **Read the question list before you start.** `boards-api.greenhouse.io/v1/boards/{token}/jobs/{id}?questions=true` returns every field and whether it is required, without authentication. That is how you learn a cover letter is mandatory before you are twenty minutes into the form. 3. **Keep it simple and guarantee a text layer.** `.txt` is accepted if you suspect a parse problem. Talent Matching highlights the matching passage for the recruiter to read, so stuffing is self defeating. 4. **On export controlled defense roles, the clearance and US person questions are legal attestations, not preferences.** Anduril's forms carry roughly 23 questions including a required cover letter. Budget real time. 5. Greenhouse acquired a voice AI interviewing company in May 2026, so an AI screen may sit in front of a human. ### Lever 1. **Fill every optional URL field.** The form is deliberately short, so LinkedIn, GitHub and portfolio links are a disproportionate share of your entire structured footprint. 2. **Your resume does more work here than anywhere else.** The visible form collects no employment dates at all. Everything historical comes out of the file. 3. **Verify the upload actually succeeded.** Lever's file input accepts anything and validates server side afterward. Wait for the success state. 4. Applying once puts you in a permanent, re-searchable pool, so it is worth doing even for a stretch role. ### Ashby 1. **Do not spray.** Application limits are real and configurable: throttling across an employer, and blocking re-application to a job you were rejected from. 2. **Evidence every hard requirement explicitly and separately.** Review returns Meets or Does Not Meet per criterion with citations to your text. 3. **Guarantee a text layer.** An unreadable resume returns `skipped`, with no criteria evaluated and no signal to you. This is the worst outcome on the platform. 4. **Use the site's own autofill box, not a browser extension.** Ashby actively rejects the way mass-apply extensions submit files. 5. Auto-reject rules use crude substring matching, so keep free text answers plain and literal. ### iCIMS 1. **Upload the resume first, then correct the autofill.** The parsed text feeds boolean search across the employer's entire candidate database and persists long after the posting closes. 2. **Keep the file well under 1 MB and export from Word or Google Docs, never from a design tool.** Extraction has a hard 90 second timeout with two attempts. A heavy PDF uploads successfully and leaves you with an empty profile. 3. **Use the exact skill nouns from the posting.** The parser normalizes against a fixed taxonomy, so the canonical industry term beats a clever synonym. 4. **Expect browser autofill and password managers to fail.** Most iCIMS career sites are the employer's domain with iCIMS in an iframe, and that boundary breaks extensions and makes the back button destructive mid-application. ### SmartRecruiters 1. **Keep the file under 2 MB.** The tightest documented limit in the enterprise tier and the most common preventable failure here. 2. **Mirror language from the Title, Description and Qualifications sections only.** The model deliberately excludes Company Description. 3. **Put skills inside role bullets.** A thin resume yields no useful score at all. 4. **The screening questions auto-reject you, not the AI.** Configuration is at the question set level, so the same question can knock out on one job and not another at the same company. 5. A rejection arriving five days later can still have been an instant automated knockout. The delay is configurable. ### SAP SuccessFactors 1. **Audit the parsed dates specifically.** SAP has an acknowledged date parsing defect. Trusting the prefill means submitting SAP's bug as your own work history. 2. **If you have applied to this employer before, manually clear the stale fields.** The parser fills only empty fields and never overwrites, so last year's record silently ships with this year's application. 3. **Apply from a desktop browser.** SAP documents that parsing does not fully work on mobile. 4. **Two independent auto-disqualify paths exist**: a single wrong answer on a flagged question, and a weighted total below the requisition's required score. 5. Under 5 MB, never a scan. And check the host, because SAP acquired SmartRecruiters in September 2025 and a longtime SuccessFactors employer may now be a SmartRecruiters employer. ### UKG, Dayforce, ADP, Paycom, Paylocity - **UKG:** identify by `recruiting.ultipro.com`. If you re-upload a corrected resume, put the version in the filename, because resumes stack with the newest at the bottom. Submit PDF or DOCX only, never an image, even though images are in the accepted list for credential documents. - **Dayforce:** do not apply twice to the same requisition, because duplicates are automatically declined. Clean up your LinkedIn first, because Dayforce parses your resume **and** your LinkedIn into the same record and a recruiter sees both. Spell out acronyms, because parsed data is indexed into the recruiter search engine. - **ADP:** read the host. Anything other than `workforcenow.adp.com/mascsr/` or `myjobs.adp.com/` means the employer runs a different ATS behind ADP and none of this applies. The score ranks, it does not reject: ADP documents no cut-off scores. Make education, skills and experience legible as three separate blocks, because the model scores them separately. - **Paycom:** the mandatory electronic signature is at the very end. Saved is not submitted. - **Paylocity:** shipped integrated pre-hire assessments and one-way video in 2025, so expect the assessment stage earlier than you would elsewhere. ### Workable 1. **Do not sign your cover letter.** Workable's anonymized screening redacts name, photo, address, email and phone, but its documentation states signed cover letters are not anonymized. 2. **Answer every Yes or No question literally.** Auto-disqualify fires only on Yes/No types and only on a "No." You are not told, but it is a sort rather than a delete and lands in a reviewable tab. 3. **Budget real time.** Every custom question is mandatory. There is no skip. 4. **Save the confirmation email.** It contains a withdraw link and a full copy of your answers, which is the only retraction mechanism in this tier. 5. **A designed or even scanned PDF is genuinely lower risk here than anywhere else,** because the parser is a language model. Do not degrade a good layout out of generic fear. ### BambooHR and JazzHR - **BambooHR:** optimize for a phone screen, not a parser. Reviewers use a mobile app. Single column, generous type, short bullets. No AI ranking or knockout screening is advertised. BambooHR publishes no parsing documentation at all, so distrust any guide that states its behavior confidently, including this one. - **JazzHR:** fill the manual fields with real care, because the parse is the weakest documented in the SMB tier. Prefer the employer's own JazzHR page over a one-click board apply, which strips formatting. AI matching is a paid tier feature, so at many employers a human reads in submission order, which raises the value of applying early. ### LinkedIn 1. **The profile, not the resume, is the searchable artifact.** Recruiter filters read job titles, skills, companies, schools, industries, location and languages as structured fields. Anything living only in your PDF is invisible to search. 2. **Put the exact target job title in your headline and current experience title.** A creative headline removes you from the search recruiters actually run. 3. **Recruiters cannot sort search results at all,** only narrow them. So the levers that re-cut the list are LinkedIn's "spotlights": recently updated profile, resume shared with recruiters, following the company, and having a connection inside it. 4. **Prefer the off-site Apply button over Easy Apply for jobs you actually want.** Easy Apply reaches a real ATS only when the employer built the integration. Otherwise it lands in a LinkedIn inbox many employers never open. 5. **There is no undo.** You cannot edit or withdraw a LinkedIn application once submitted. ### Indeed 1. **Go to profile settings and switch to "Employers can find you."** The default is the opposite, and uploading a resume for an application does not change it. Highest value five seconds on the platform. 2. **No knockout blocks you.** Indeed's documentation: it "delivers every application to the employer, whether or not the applicant meets the qualification criteria." A near miss is worth submitting. 3. **Keep the resume file single column and text based** and put the designed version in the supporting documents slot. Indeed forwards several parallel renderings of you and employers filter on the text. 4. **Download a local copy before replacing your resume.** There is one file slot and a new upload overwrites the old with no backup. 5. Answer numeric screeners precisely. "About 5 years" typed into a numeric field can fail a "5 or more" test. ### Dice and ClearanceJobs - **Dice:** write skills as literal current stack nouns. Log in regularly, because "Last Active" is a filter that runs before skills are compared. Link GitHub and a portfolio. Set a realistic salary expectation, because a blank one gets you cut before anyone reads anything. - **ClearanceJobs:** fill all five clearance fields exactly: level, **status**, sponsoring agency, polygraph type, and polygraph issuing agency. Clearance details written in resume prose are invisible to the filters. **Know the difference between Active and Current:** Active means not terminated, Current means terminated but still eligible for reinstatement. If you have left the cleared position you are almost certainly Current. Recruiters verify against DISS and Scattered Castles, and the platform ships a one-click path to report a mismatch. This is a federal-adjacent misrepresentation, not a resume embellishment. ### Handshake 1. **On any job that sends you off-site, go all the way through and confirm you got a submission confirmation from that system.** Handshake's own documentation says its applicant list shows everyone who clicked to apply, whether or not they finished. You can look like an applicant to Handshake while the employer has no record of you. 2. **Complete the structured fields.** Major, GPA, graduation year, location, work authorization are the literal inputs to applicant ranking. Blanks are treated as non-matches, not as neutral. 3. You are ranked against criteria you cannot see, some of which are employer free text that appears nowhere in the posting. --- ## 8. USAJOBS, which is a different game Federal hiring genuinely does not work like the private sector, and it was re-plumbed twice inside twelve months. **Almost every federal resume guide online now describes a system that no longer exists.** **What changed** - **Merit Hiring Plan, 29 May 2025.** The federal resume is **capped at two pages**, and the USAJOBS resume builder enforces it. A stored longer resume will block you from applying. The same memo phased out self-assessment questionnaires for rating and ranking, except for seasonal work and GS-01 through GS-04, and required at least one technical or alternative assessment in its place. - **"Rule of Many" final rule, effective 7 November 2025.** Numerical scoring and ranking are restored as an option alongside category rating. **Read "How you will be evaluated" before writing a word.** Under category rating there is no numerical score, you are placed in a quality category, and veterans' preference changes your order within a category rather than adding points. Under the Rule of Many you get an actual score augmented for preference, and the agency certifies from the top using a mechanism it has to declare in the announcement. A "top ten percent" announcement is a different writing problem from a two-category one. **Three knockouts, all of which precede any ranking** 1. **Hiring path eligibility.** The "This job is open to" block. If the announcement is status only and you are not a current competitive service employee, nothing else matters. 2. **Minimum qualifications**, judged from your resume against OPM standards for the series and grade. This is where month and year dates and hours per week decide the outcome, because specialized experience is counted in months at a defined level. 3. **Supporting documents.** A missing DD-214, SF-15, SF-50 or transcript is a hard, unappealable knockout. Agencies will not chase you for it. **Writing the two pages** 1. **Build it around the announcement's Specialized Experience paragraph,** one block per requirement, in its exact vocabulary. USAJOBS states it plainly: if the qualifications section says you need experience with a named tool, you have to use those words. And the hiring agency will not make assumptions about what is in your resume. Spell out every acronym. 2. **Every job block carries five elements:** employer, job title, month and year start and end, hours per week, and for federal jobs the pay plan, series and grade. For example: `Program Analyst GS-343-11 / January 2009 to Present / 40 hours per week`. 3. **Cut ruthlessly.** Anything older than about ten years or off target comes out. 4. **Export as PDF and count the rendered pages before uploading.** Under 5 MB. No encrypted or digitally signed files, no photo, no Social Security number, no date of birth. 5. **Finish inside the agency system.** USAJOBS hands you off, and applications that stop at the hand-off do not exist. Go back and confirm receipt in your profile. 6. **Turn on "Make searchable."** It exposes your resume to agency recruiters for direct outreach, which is a genuinely different channel from the announcement queue. *The two-page cap and the questionnaire change are sourced to OPM's Merit Hiring Plan memorandum of 29 May 2025 and to USAJOBS' own help pages. Federal rules in this area moved twice in a year, so confirm the current requirement on USAJOBS before you submit.* --- ## 9. Test your parse before you send **Tier 1, thirty seconds, catches the most common failure in this whole document.** Open your PDF, select all, copy, paste into a plain text editor, and read what came out. - Nothing pastes: there is no text layer. Your file is an image. Fix this before anything else. - The reading order scrambles: every parser in the market sees the same scramble. - Your phone and email are missing: they are in a header, a footer, or a text box. - Bullets come out as `q` or blank boxes: you are using a decorative glyph from a symbol font. **Tier 2, round trip through a real parser.** Affinda offers a free sandbox that returns the raw extracted JSON. This is the closest you can get to seeing what an ATS sees. Check specifically that **every role has both a start and an end date attached to the correct employer**, because employer and title inversion and orphaned dates are the two highest risk failures. **Tier 3, the platform's own preview, which most people click past.** Oracle Recruiting Cloud lets you include or exclude parsed skills before commit. Workday's entire My Experience page after autofill *is* the parse, and it is your only chance because the application is immutable after submit. On iCIMS, SmartRecruiters, SuccessFactors and Taleo the autofilled form is the parsed record. **Tier 4, the human test.** Print it and read it at arm's length for ten seconds. Can someone see your main credentials from the top quarter of the first page? On BambooHR employers, read it on a phone, because that is where the hiring manager will. **What none of this tests** is whether your knockout answers will disqualify you, which is the only thing that automatically rejects anyone. There is no way to test that except reading the questions carefully. On Greenhouse you can read them in advance via the boards API above. --- ## 10. What we could not verify Stated honestly, because a short honest list beats a long confident wrong one. - **Parser engines are undisclosed for most of the market.** Confirmed: Textkernel for Oracle Recruiting Cloud, SmartRecruiters, SuccessFactors and Bullhorn. DaXtra for iCIMS, Jobvite and Zoho Recruit. Gemini for Workable. **Unknown**: Workday, Taleo, Greenhouse, Lever, Ashby, Dayforce, UKG, ADP, Paycom, Paylocity, BambooHR, JazzHR, Breezy, Recruitee, NEOGOV, Cornerstone. Any article naming one of those is making it up. Also note that iCIMS runs DaXtra, not Textkernel, contradicting a very widely repeated claim. - **Date format handling is undocumented by every single vendor here.** The symptoms are well attested, the mechanism is not. The date rules above are consistency hedges. - **Section heading recognition is undocumented by everyone except DaXtra.** Nobody publishes a heading whitelist. The heading table is a free hedge, not a proven rule. - **DOCX versus PDF parse quality has never been publicly benchmarked by anyone.** Every comparison percentage in circulation is fabricated. The decision rule above is built on documented format acceptance and known parser generation, not on measured fidelity. - **Accepted formats and size caps are undocumented** for Dayforce, ADP, Paycom, Paylocity, BambooHR, Avature, JazzHR, Teamtailor, Zoho Recruit, NEOGOV and Cornerstone. The "under 2 MB" default is a safety margin derived from the platforms that do publish limits. - **ATS market share figures are unusable.** For the same year and the same product, one vendor has been reported at 40%, 15%, and 0.47%. - **Numbers that are fabricated and should never be repeated:** "75% of resumes are rejected by an ATS" (traces to a 2012 sales pitch by a company that shut down in 2013 without publishing a study); "over 60% of parsing errors come from two-column layouts"; "Workday autofill has a 34% error rate"; every DOCX versus PDF fidelity percentage; and a widely cited "2023 Jobscan study" that does not exist. - **Not researched here and worth knowing about:** NEOGOV, which covers 13,000+ US public sector agencies and deserves the same treatment Workday got; PageUp, which is actively winning Taleo migrations; and Phenom, Beamery and Eightfold, which are matching layers that sit *in front of* an ATS and can screen you before you ever reach the systems described here. **One real gap worth naming.** There is no accessibility guidance in this document, and that is a genuine omission. Ashby's job board is a client-rendered app that is hostile to some assistive technology. iCIMS runs in an iframe that breaks password managers and the back button. A disabled applicant navigating these systems has no map here, and should. --- ## A note on hidden keywords White text, tiny text, off-page text, and keyword blocks in document metadata all still extract. That is exactly why they fail rather than why they work. Academic work presented at USENIX Security in 2026, built on roughly 200,000 real resumes, found that about 1% contain hidden injected text, that more than 90% of it is hidden skill lists rather than instructions to an AI, and that detection is cheap and deployed. The most effective detector renders the page to an image, reads it visually, and compares that against what the machine extracted. Anything present in the file but absent from the rendered page is flagged by construction. There is no version of this trick that survives that check. A separate 2026 paper found the tactic only helps when few candidates use it and collapses as adoption spreads, which is where it now is. Meanwhile Greenhouse highlights the matching passage for the recruiter and Ashby cites your source text, so an inflated score hands a human the exact sentence that inflated it. And select-all in any PDF viewer reveals hidden text instantly. Do not do it. Not because it never works, but because it works badly, is detected cheaply, and the downside is being blocked at the company and sometimes at the agency.