What application forms actually ask

Every job-application form asks you things that are not on your résumé. This page counts which ones, from a log of real forms rather than from memory.

The platforms are not alike, and the difference is the useful part

The single most practical finding here is not which questions get asked. It is how much the same questions repeat on each platform. The encounter log splits by source like this (the answers row is the pipeline's internal answer-map hits, not an Applicant Tracking System; the residual row keeps the table honest — the rows must sum to the log's own line count, 61,132):

Greenhouse questions are written per posting, so they are a long tail of rarely repeated labels. Workday questions are configured once per employer tenant and then reused across that employer's whole job board, so 10,708 Workday encounters collapse into just 293 distinct questions — about 37 reuses each. In practical terms: answering a Workday employer's question set once covers nearly every other job at that employer, and reading a Greenhouse posting's form tells you very little about the next one. The Workday set is small enough to list, so there is a separate page for it.

What the questions are about

Each of the 6,521 distinct labels in the gap queue (all platforms, weighted by how often it was met) was matched against 17 ordered categories, first match wins — the full rule list is printed at the bottom of this page:

Dated snapshot: this table was computed 2026-11-07 by the published rule list below, and its rows sum exactly to its own denominator of 61,717 encounters — the residual is a row, not a footnote. (The queue total above is a separate instrument read at a different instant, so the two counts may differ by the day's traffic.)

Read the last row first. 16.8% of encounters matched none of the 17 categories — mostly role-specific questions, "if other, please specify" follow-ups, and fields whose label the page never gave in readable form. A category scheme that explained everything would be a scheme that had been fitted to the answer. This one was not, so the residual sits in the table, where you can see it.

Twenty real questions, with how often each was asked

These are the form's own words, with the automation's internal field-type prefix removed and sentence capitalisation restored by hand (the queue stores labels lowercased, so this table is a hand-normalised derivative, not raw database bytes). Counts are the gap queue's current encounter count for that exact normalised label, measured 2026-11-07. A count of 0 means the label has since left the queue under that exact wording — it was really asked, but today's instrument no longer holds it. No employer is named on this page, deliberately — a large share of real labels contain the company's name, and publishing those would disclose who was applied to.

213 of them are meant to stay blank

Of the 6,521 distinct labels, the disposition breaks down as: 6,256 open (no stored answer yet), 213 deliberately blank, 38 answered per-application rather than from a stored fact, and 14 with a stored answer.

The 213 are the ones worth understanding. Citizenship, sponsorship, security clearance, criminal history, age, disability and veteran status are questions where a wrong answer is not a typo — it is a misrepresentation on a legal document, and in several cases the honest answer depends on details no stored fact captures. They are configured to be left empty for a human to fill, not filled with a best guess. A blank field that a person completes beats a plausible-looking machine answer, every time.

The 17 categories — the ordered match rules, in full

Each cleaned label (widget prefix and option list stripped, whitespace collapsed, lowercased) is tested against these rule sets in this order; the first set with any matching regex claims the label. A label no set claims lands in the residual bucket, which is published above, not hidden. Patterns name concepts, never companies.

1. Age & protected identifiers — \b18 years\b · years of age · age or older · date of birth · \bbirth ?date\b · what is your (current )?age\b · your current age\b
2. How did you hear about us — how did you (first )?hear · where did you (first )?hear · hear about (us|this|the) · how did you learn · where (did|have) you learn · where you learned about · referral source · source of (referral|application) · \breferr
3. Security clearance & export control — clearance · export control · export-controlled · \bitar\b · polygraph · top secret · public trust · \bcuba\b · sanction · embargo
4. Demographics & EEO self-identification — \beeo\b · equal employment · self[- ]identif · \bgender\b · pronouns · transgender · non-?binary · \brace\b · ethnic · hispanic · latin[ox] · veteran · military (spouse|service|status) · armed forces · disabilit · \bdisabled\b · sexual orientation · lgbt · demographic
5. Prior or current relationship with the employer — former(ly)? (an? )?(employee|employed|intern|contractor) · currently or formerly · current or former · currently (an? )?(employee|employed|contractor) · previously (worked|employed|interviewed) · (ever|previously) been employed · have you (ever )?worked (at|for) · immediate family · family member · relative · close personal relationship · (personal|family) relationship · your relationship to · conflict of interest · nepotism · fiduciary · economic interest · competing business · prior[- ]employer · rehire · re-hire · employee of (the )?u\.?s\.? government · post-government · government employee · do you know (anyone|someone|any employee) · (work|worked|working) (at|for) (this|our|the) company · current(ly)? work(ing)? on .{0,25}project · secondary non- · outside (employment|business|work|activit)
6. Work authorization, sponsorship & citizenship — sponsor · work authori[sz] · employment authori[sz] · authori[sz]ed to work · legally authori[sz]ed · right to work · \bvisa\b · citizen · nationality · permanent resident · green card · work permit · eligib\w* to work · work eligibility · immigration · h-?1b · lawfully
7. Language proficiency — language · \benglish\b · \bspanish\b · \bfrench\b · \bgerman\b · fluen · bilingual · native speaker · ^(reading|writing|speaking|listening|comprehension|overall)$
8. Employment-history dates — ^(start|end) date · ^(from|to)$ · ^(month|year)$ · dates? of employment · employment dates
9. Compensation expectations — salary · compensation · \bpay\b · pay (range|rate|expectation) · hourly rate · \bwage\b · remuneration · \bctc\b · \bote\b · \bbase pay\b · \brate\b.{0,20}(hour|annual|year)
10. Travel & on-site expectations — travel · on-?site\b · \bin[- ]office · in the office · hybrid · \bremote\b · work from home · return to office · commute · days (a|per) week
11. Location & relocation — reloc · locat · based in · where are you based · what (country|state|cit(y|ies)|province) · which (country|state|cit(y|ies)|province) · \bcit(y|ies)\b · zip ?code · postal code · time ?zone · geograph · reside\b · residence · city and state
12. Start date & notice period — notice period · start date · when (can|could|are) you · earliest.{0,20}start · available to (start|begin) · availab · date available · join(ing)? (us|date)
13. Contact details & profile links — website · linkedin · github · portfolio · \burl\b · profile · phone · mobile · telephone · e-?mail · twitter · contact you · how should we (communicate|contact) · \bsms\b · text message · address · (preferred|first|last|legal) name
14. Education & credentials — school · university · college · education · degree · \bgpa\b · graduat · \bmajor\b · transcript · academic · certif · licen[sc]e · discipline · field of study · diploma · bachelor · master's · \bphd\b · doctorate
15. Years of experience / skills matrix — years? of .{0,30}experience · how many years · years of · experience (with|in|using|working|as) · hands-?on experience · do you have experience · \bskills?\b · skill level · rate your · self-?rate · proficien · familiar(ity)? with · how long have you · expertise · percentage of your (work|time)
16. AI-tool usage questions — \bai\b · artificial intelligence · generative · chatgpt · copilot · large language model · \bllm\b
17. Legal agreements & consents — ^i (have read|agree|understand|acknowledge|certify|consent|confirm|accept|attest|authorize|hereby|declare) · \bi hereby\b · \bi attest\b · \bi acknowledge\b · \bi certify\b · privacy (notice|policy|statement) · data privacy · gdpr · arbitration · terms (and|&) conditions · terms of (use|service) · consent · acknowledg · by (typing|signing|submitting|checking) · signature · e-?sign · affirmation · code of conduct · background (check|screening) · drug (test|screen) · non-?compete
18. Other — matched none of the above (the residual row).

Reproducing this

The numbers above came from these queries, run on 2026-11-07 against the live database and log that the automation writes to (the same strings, machine-readable, are at /stats.json):

SELECT count(*), sum(n_seen), min(first_seen)::date, max(last_seen)::date FROM answer_gap_queue; → 6,521 · 61,717 · 2026-09-03 · 2026-11-06

SELECT state, count(*) FROM answer_gap_queue GROUP BY 1 ORDER BY 2 DESC, 1; → open 6,256 · intentional_blank 213 · per_instance 38 · answered 14

The per-platform split was computed from logs/answer_gaps.log, one line per encounter, in the form <timestamp> | source=<ats> | label="<the form's own words>" | … — 61,132 lines, split by grep -oP 'source=\K\w+' logs/answer_gaps.log | sort | uniq -c, with any line the parser cannot attribute counted in the unattributed row rather than silently dropped. The category table is answer_gap_buckets.py in the same tree — it exits nonzero if its rows do not sum to the denominator.