How Job Application Autofill Actually Works (And Where It Breaks)

Every autofill tool does roughly the same three things: read the form on the page, match each field to something it already knows about you, and type the answer in. The differences between a good one and a frustrating one show up entirely in what happens when a field doesn't have an obvious match.

What autofill can reliably handle

Standard fields — name, email, phone, LinkedIn URL, current company, years of experience — are easy. They show up on almost every application in a predictable shape, so a tool built for Greenhouse, Lever, or Ashby forms (the platforms most mid-size and large companies actually use) can fill these correctly close to 100% of the time.

Resume upload is also reliable, since it's just attaching a file. Where things get genuinely harder is anything the tool has to generate an opinion about.

Where it breaks: custom questions

Almost every application has at least one free-text question specific to that company — "Why do you want to work here?", "Are you authorized to work in this country?", "What's your expected salary?". These don't have a standard answer sitting in your resume, and a tool that guesses badly here does more harm than good: a generic, obviously-copied answer to "why us" is worse than leaving it blank.

The honest approach — the one worth actually using — is to ask you the first time a genuinely new question shows up, save that answer, and reuse it automatically on every future application. Some tools now also try an AI-drafted first pass from your resume and preferences before asking, which works well for straightforward factual questions (work authorization, notice period) and poorly for anything requiring real judgment about the specific company.

Why the review step matters

The single biggest mistake an autofill tool can make isn't filling a field wrong — it's submitting without you seeing the result first. A dropdown that defaulted to the wrong option, a salary field that got the currency wrong, a question it answered with a low-confidence guess: these are all recoverable if you catch them before Submit, and genuinely damaging if you don't.

This is the actual design tradeoff every tool in this space makes, whether it says so or not: faster and less careful, or slightly slower with a real review step. If you're evaluating one, that's the question worth asking before you ask about speed.