29 SEPTEMBER 2026 · 8 MIN READ

We read 70 job application forms. The same question is asked 25 different ways.

Search for how to write a CV that gets past the applicant tracking system and you will find the same advice repeated for a decade: use a single column, avoid tables, mirror the keywords, and above all remember that some very large fraction of applications are rejected by a machine before a human sees them. Almost none of it is measured. So we measured the part that is actually measurable — not what the software supposedly does to your CV, but what the form in front of you actually asks for.

What we did

Otto fills in job applications on people's behalf, which means it has to read the form first. Every one of the ten hiring systems we support publishes the structure of its application form — every question, whether it is required, and the full list of options — over an ordinary web request. No login, no browser, no scraping heuristics. The form is public because it has to be: the page you fill in is built from it.

On 26 September 2026 we read 7 live forms from each of the 10 systems — 70 forms in total — and catalogued every question on them. That produced 1,059 question-appearances, which collapse into 484 distinct ways of asking something. Of those 484, 156 are a box the hiring system puts on every form it generates and 328 belong to one of 51 individual employers who wrote their own.

The catalogue records how employers ask. It contains no applicant's answer to anything — that is a deliberate design constraint of the system it comes from, not a redaction we applied afterwards.

An application is fifteen questions, not one

The mental model most advice assumes is a CV going into a slot. The reality across 70 forms averages 15.1 questions per application. Here is what gets asked most, counted by how many of the 70 forms carried it:

The counts exceed 70 in places because a single form can carry the same question twice — 75 CV-upload boxes across 70 forms means some employers ask for a second document in a second slot.

The cover letter is not universal, and it is not usually a file

The advice industry treats the cover letter as a given. Across our 70 forms, 33 asked for one at all — 47%. Rather more than half of the applications we read had nowhere to put a cover letter even if you had written one.

Where it is asked for, the shape depends entirely on the system rather than the employer. Greenhouse, Recruitee, Pinpoint, Personio, Ashby and Paylocity want a file — 21 of the 33. Workable and Breezy want text typed into a box — the other 12. A beautifully typeset PDF is worth nothing in a textarea, and a paragraph of plain text is not what a file upload wants.

The same question, twenty-five different ways

This is the finding that surprised us. Two questions in our catalogue were never once asked the same way twice.

Right to work: 25 appearances, 25 distinct wordings. Every single employer who asked it phrased it themselves. A sample, verbatim:

• “are you legally authorized to work in the united states”
• “what is your right to work in the uk”
• “are you legally authorized to work in canada for any employer”
• “are you authorized to work in the country you currently reside in”
• “are you legally eligible to work in canada without sponsorship”
• “do you have a valid working permit to work in the netherlands”
• “what right to work documents do you have”

Those are not stylistic variations. “Authorized to work in Canada” and “authorized to work in Canada for any employer” are different questions with different correct answers, and “eligible to work without sponsorship” folds two questions into one. One employer asked in French and English in the same box.

Visa sponsorship: 15 appearances, 15 distinct wordings. Same pattern — and here the tense is what moves:

• “do you require immigration sponsorship to work for <company> in the united states”
• “will you now or in the future require sponsorship for a visa to remain in your current location”
• “do you currently require visa sponsorship, or might you in the future”
• “would you require future sponsorship in order to work for our company in canada”

“Do you require sponsorship” and “will you now or in the future require sponsorship” have opposite answers for a large number of people — anyone on a visa that currently permits work but expires. Both were asked, on different forms, of candidates for comparable roles.

Salary is the most fragmented question of all

Salary expectation appeared on 47 forms in 33 different wordings — the highest ratio of phrasings to appearances of any question that gets asked often. Employers are asking for an annual figure, an hourly rate, a range, a minimum, a currency, “compensation expectations”, and a free-text box, and they are doing it with no shared vocabulary whatsoever.

This matters more than the phrasing suggests, because salary is the one answer on the form that is a negotiating position rather than a fact. There is no equivalent of getting your email address slightly wrong.

Roughly a third of what gets asked is sensitive

Of the 484 distinct questions, 148 — 30.6% fall into categories that need careful handling rather than a lookup: right to work, sponsorship, salary, age, gender, race, veteran status, sexual orientation, disability, and background-check consent.

The demographic block alone — gender, race, veteran status, disability, pronouns — accounted for 92 appearances across 58 distinct wordings. In most jurisdictions those questions are voluntary and legally walled off from the hiring decision. They are still two-thirds of the way down a form that also wants your phone number.

So what about the ATS rejecting your CV?

Here is what we can say from data rather than folklore. We know precisely which systems are doing the receiving, because we index the postings: 661,133 live openings across ten systems, with SmartRecruiters at 31.2% and Greenhouse at 24.8%. The full distribution is on our ATS coverage page, and we wrote up what the market shares imply in which hiring systems companies actually use.

What we did not find, anywhere in 70 forms, is a keyword filter. What we found is an average of fifteen questions, several of which are gates with a hard right answer — right to work, sponsorship, location, notice period, onsite days. Those are structured fields with structured options, and they are evaluated exactly as written.

That reframes the advice. Formatting your CV for a parser is guarding a door that may not be the one you are being stopped at. A wrong answer in a required dropdown is unambiguous and unappealable in a way that a two-column layout is not.

What we are not claiming

Seventy forms is a survey, not a census. It is seven per system, chosen to span the ten systems rather than to be representative of employers within them, and it skews toward companies whose roles our users actually apply to. Percentages here carry real slack and should be read as the shape of the thing rather than the precise size of it.

We also did not measure outcomes. Nothing here tells you how any employer weighs an answer, whether a human reads the cover letter that 47% of forms asked for, or what proportion of applications are filtered automatically. We counted what the form asks. Anyone telling you what happens to your CV after you press submit — including us — is inferring.

And the catalogue grows from live applications, so it is biased toward employers who are hiring into roles people want. It says nothing about the long tail of the companies in our index that were not in the sample.

The part that matters if you are job hunting

Three things follow from the numbers, and none of them are about fonts.

• Budget for fifteen questions, not one. The CV is the part you have already done. The other fourteen answers are what the next twenty minutes go on, per application.
• Have your right-to-work and sponsorship answers settled before you start. They will be asked in wording you have not seen before, and the difference between “currently” and “now or in the future” is doing real work.
• Write the cover letter as text, not as a layout. More than a third of the forms that want one want it pasted into a box.

This is also the reason Otto exists in the shape it does. The reading of the form, the fifteen answers and the twenty-five phrasings of one question are the actual work of applying, and they are the part that does not get easier with practice. How Otto handles it is the same catalogue described here, pointed at your answers instead of at a survey.

Method: 70 live application forms — 7 each from Greenhouse, Lever, Ashby, SmartRecruiters, Workable, Recruitee, Breezy, Pinpoint, Personio and Paylocity — read on 26 September 2026 from each system's own published form definition. Every question was classified by what it asks; counts are question-appearances across forms, so a question on all 7 forms of a system counts 7. 1,059 appearances, 484 distinct wordings, 51 named employers. The catalogue records employers' questions only and holds no applicant answers. Market-share figures are from the Otto Roles index on 29 September 2026.