Auto apply vs autofill: what actually gets sent

Search for a way to automate job applications and you get three different products under one name. One pastes your details into the boxes and hands the page back to you. One fires the same résumé at hundreds of listings and shows you the count. One reads the posting, decides whether it is worth your name, fills the employer's own form and presses submit. The words on the pricing page are nearly identical. What arrives at an employer is not. Here is how to tell them apart from the outside, in about five minutes, including on us.

Three things people mean by “auto apply”

• Autofill. A browser extension recognises the fields and pastes your name, your email and your links. You still read the posting, still write the extra answers, still press submit yourself. It removes typing, which was never the slow part.
• Bulk sending. Volume sold as the feature: hundreds of listings a day, nothing read, one résumé everywhere. The number on the dashboard goes up every day. The replies do not.
• A real submit. The software reads the posting, checks it against rules you wrote, opens the employer's own form in a real browser, fills every field with your answers, attaches the résumé, clicks submit — and keeps proof of what the employer's server received.

Most of what is on sale is the first, described in the language of the third. That is not an insult to extensions: an extension that fills five boxes does exactly what it says. The problem starts when a page counts those five boxes as an application and reports a number back to you.

What autofill leaves on your plate

Autofill is usually measured by the fields it recognises. Measure it instead by the minutes it removes. We took a real form apart field by field — the one 147 employers were using last month — and the five fields an extension is good at are the five that were never slow. First name, last name, email, phone, country: half a minute by hand.

Everything under them is where the hour goes. School is a search against a remote catalogue, not a text box: typing “Centro Universitário FIEO” returned nothing after 6.5 seconds when we timed it. Degree is a short fixed list of levels that refuses “B.S. Computer Engineering”. Then the employer's own questions begin, and they are unbounded — one form asked for consent to have interviews recorded and transcribed, plus thirty required checkboxes of countries. “How did you hear about this job?” only registers when you click the option in the list; pressing Enter leaves it empty while it looks filled. The certification at the end is a text field where you type your own name.

The last row is the one worth arguing about. Nothing in an autofill workflow tells you an application arrived, because nothing in an autofill workflow is watching after the click. You assume it went, the way you assume a letter you posted arrived.

What a real submit looks like on the wire

When you press submit, the browser sends a request to the employer's server and the server answers with a number. Two hundred means it took the application. Four hundred and something means it refused it. That answer is the only place in the whole process where a machine can tell the truth about what happened, because it comes from the other side.

Almost every tool skips it and reads the page instead: the button went away, a green box appeared, therefore submitted. That is a guess about a picture. A validation error also makes the page change. So does a session that quietly expired. And so does the trap we measured on one very common form: after the click, a block titled “Security code” appears with eight one-character boxes, and an email arrives with the code. Over ten submissions on 21 August we watched it get sent the first time an address submits and then reused, so one code arrived four minutes before the submit that needed it. Until those eight boxes are filled, nothing has been sent — and the page looks exactly like a page that has finished.

So nothing here is called sent on the strength of the page. After the click, the software looks at what the browser really sent: a submit request to the form's own site, and the number the employer's server answered with. When it finds one, that goes into the application's timeline in those words — the method, the status, the address. When the network cannot witness it, a confirmation written on the page itself is required instead. With neither, the software is refused permission to call the application sent, and the row stays open. The screenshot of the moment it went, the log of what went into each field and the times are filed beside it. When an attempt dies against a wall instead, more is kept than a picture: the page's own HTML, the numbered outline of controls the software was reading off, and the network log — so that “the submit never left” and “the submit left and came back 422” stop being the same observation. The whole mechanism is written out on the auto-apply page.

Why 10 a day beats 1,000

The first answer is supply. In four days of a real search we read 360 postings across eleven job sites. Forty-seven of them matched the target — 13% — and nine became applications. On the accounts we run today the number of roles a day genuinely worth someone's name lands between one and eight. A tool offering a thousand a day is not offering you a thousand good roles. It is offering to stop reading.

The second answer is what happens on the other side. A hundred applications from one person in one morning is visible to a recruiter, and hiring software has started answering that pattern with fraud checks. The tools that promise volume invert the ratio that makes a candidate credible, and the cost of that lands on the candidate, not on the tool.

So the ceiling here is ten a day and three hundred a month on the paid plan, one a day and twenty a month on the free one — forever, with no card. There is no setting that turns ten into a thousand. And ten is a ceiling, never a target: on a day when nothing clears your bar, nothing is sent, and that is the product working, not failing.

Five minutes to test any tool, including this one

• Ask for one submit on the wire. Not a screenshot, not a status word in a dashboard: the request that left and the answer the employer's server gave. A tool that cannot show it does not know either.
• Send one application and watch your mailbox. Most employers send an automatic confirmation. If a week of “submitted” produced no confirmations at all, the count was never applications.
• Ask what it does with a question your data cannot answer. The only acceptable answer is a truthful conservative one, flagged for you to correct. Anything that quietly invents a work permit is spending your reputation.
• Read the price twice. A subscription that then sells applications in packs is a different number at the end of the month from the one on the pricing page. Ask what the total is for a month of real use.
• Ask which account it signs into. Anything that wants your LinkedIn login is betting your professional profile on their automation not being noticed. That account is not worth the convenience.
• Ask how often it refuses. A tool that never says no to a posting is not reading them. Ask for the share it dropped last week, and why.

Run that list on us. One: yes — the method, the status and the address of the request the browser sent, quoted in the application's own timeline, or a confirmation lifted from the page when the network cannot see it. Two: employer confirmations land in a mailbox opened for your search, filed against the right application. Three: a form question with no truthful answer in your data gets the conservative one, flagged for you to correct. Four: US$29 a month covers the applications, with no credits on top. Five: we never sign into your LinkedIn, for reading or for applying. Six: 87% of what we read in that four-day run never became an application. The rest is on the FAQ.

Read next

• What the Greenhouse application form asks you — field by field, including the eight-character code that arrives after the button.
• The robot that applies to jobs: how it really works — what the press left out, and what the proof of a submit is made of.

Try the test on your own search: create an account, paste one job link, and you get the filled application back with the request, the status and a photo of the submit.