What 647,795 application forms actually ask
Search for why an applicant-tracking system rejected your résumé and you will find the same article about thirty times. Most of it rests on one statistic — that around 70% of résumés are filtered out before a human sees them — which is repeated everywhere and sourced nowhere. We are not going to replace it with a number of our own, because we do not have one either. Nobody outside the ATS vendors does.
What we do have is the forms.
freehire captures the application form behind every posting it crawls, because our auto-apply feature has to fill them in. As of 2026-09-09 that is 647,795 forms across five platforms, stored the way the platform sent them — the employers’ own wording, not our summary of it.
| platform | forms |
|---|---|
| Greenhouse | 283,363 |
| Lever | 117,424 |
| Ashby | 87,988 |
| Recruitee | 79,584 |
| Workable | 79,436 |
That is not the job market. It is the postings from five platforms that we crawl and could capture a form for, which skews toward tech, and toward companies large enough to buy an ATS at all. Everything below is about those.
How much a form asks
The simplest question first: how many fields does an application form have?
| platform | questions, average | median | longest form |
|---|---|---|---|
| Greenhouse | 19.9 | 19 | 113 |
| Lever | 14.6 | 15 | 154 |
| Workable | 14.1 | 13 | 78 |
| Ashby | 10.5 | 10 | 50 |
| Recruitee | 8.5 | 7 | 60 |
The averages are safe to read here, and that had to be checked rather than assumed: on every platform the median sits within 1.5 of the mean, so no handful of enormous forms is dragging the figure. If the mean had been 19.9 against a median of 6 it would have been describing something almost nobody meets.
So the same application costs you about 2.3 times more work on Greenhouse than on Recruitee — and you did not pick either one. The employer did, probably years ago, for reasons that had nothing to do with you.
The longest form in the corpus has 154 fields. It is on Lever, whose typical form is one of the shorter ones.
How much of it you cannot skip
Fields are one thing; mandatory fields are another. Here the measurement gets narrower, and the reason why is worth more than the number.
Greenhouse, Ashby, Workable and Recruitee each publish an explicit “this field is
required” flag through their APIs. Lever does not. Its form has to be read out of the
page markup, and requiredness inferred from a ✱ glyph next to the label. So we
checked what that inference actually yields: 48.8% of Lever forms come back with no
required field at all, against 0.0% for each of the other four.
A form with no mandatory field is not a thing that exists — every one of them needs a name or an address to reply to. So that 48.8% is not Lever being relaxed. It is the failure rate of reading a form out of markup instead of an API, and it means any required-field number we published for Lever would be measuring us rather than them. Lever is therefore left out of the table below and kept in the one above, which counts fields and never reads the flag.
| platform | required, average | median | most required |
|---|---|---|---|
| Greenhouse | 13.2 | 13 | 76 |
| Workable | 10.3 | 9 | 75 |
| Ashby | 8.3 | 8 | 40 |
| Recruitee | 6.8 | 5 | 48 |
Two thirds to four fifths of a form is mandatory, consistently: 66% on Greenhouse, 73% on Workable, 79% on Ashby, 80% on Recruitee. The optional questions are the minority everywhere, which is the opposite of how application forms are usually described.
And somewhere out there is a Greenhouse form with 76 required answers on it. For one job. Against a median of 13, that is not typical — but somebody built it, and somebody is expected to fill it in.
What they ask, once the basics are done
Name, email, phone, CV and cover letter are on nearly every form. They are the floor, not the finding. Underneath them the forms diverge, and the wording is the employers’ own.
Counting this honestly needs a caveat about method. Labels are stored verbatim, so one
question fragments across spellings: resume/cv, resume, cv, lebenslauf, lettre de motivation and currículo/cv are one field; linkedin profile, linkedin url and linkedin are one question; how did you hear about this job?, how did you hear about us? and how did you hear about this opportunity? are one
question asked three ways. No mechanical rule merges those — they differ in wording,
not in formatting — so the families below were grouped by hand, and every label folded
into each one is listed in our working notes rather than hidden behind a total.
Your other profiles: 308,948 forms ask for LinkedIn. Then website (84,692), portfolio url (59,543), github url (58,455), twitter url (44,907). A photo is
asked for on 46,323.
Where you are: 243,452 forms ask your current location, plus separate city, state, country and postcode fields on top.
Two of these deserve care rather than a headline. latitude and longitude appear
118,203 times each — but on exactly one of the five platforms. So do disabilitystatus and veteranstatus, at 104,970 each. Those are one vendor’s form
template, not something the industry decided. We nearly wrote “application forms
collect your exact coordinates”, which would have been true of the count and false
about the world.
The contrast is what makes it worth saying: gender appears 113,269 times across all
five platforms, and race 106,774 times across four. Those are practices. Coordinates
are a product decision at one company.
Work authorization: 70,635 forms, mostly variations on whether you can legally work in the United States and whether you will need sponsorship.
“How did you hear about us”: 61,762. An age gate: 30,337. Salary expectations: 24,762. Notice period: 7,841.
What this does not tell you
It does not tell you why Greenhouse forms are longer. A platform’s default template, the kind of employer that buys it, and the questions those employers add are three different explanations, and counting fields separates none of them.
It does not tell you how long a form takes. Thirteen required checkboxes and thirteen required essay boxes count the same here and are not the same afternoon.
It does not tell you how many employers ask something. A company with 4,000 open postings puts its custom question on 4,000 forms, and these are per-posting counts.
And it tells you nothing at all about whether any of this filtering works, or whether that 70% figure is close. We measured what the forms ask. What happens to your answers afterwards is behind a door none of us can see through.
What you can do about it
Not much about the form. It arrives the length it arrives.
What you can change is the document you feed into it — and that part is not mysterious. The first thing every one of these systems does is try to turn your CV into text. If it cannot, nothing else you did matters.
We put that check on a page you can use without an account: drop a CV into freehire.me/roast and it will tell you what a parser reads, what it misses, and which skills the open postings in your field are asking for that your CV does not mention. It stores nothing and it does not send your CV to a model.
The forms are not going to get shorter. The CV can get more readable.