How long should your résumé be?

As long as your evidence, set well. For most careers that is one full page; for a genuinely long senior career it is two, and it is almost never three. I know that reads like a dodge on a question people want a number for, so the rest of this post is an attempt to earn it. The page-count question turns out to be an honesty question wearing a formatting costume.

Some background on why I have data about this at all. jobhunt.coach writes tailored résumés for its users. In its LaTeX path, a model writes the content and a compiler typesets it into a PDF. We evaluate candidates for that job, and among the things the eval measures is layout: how many pages came out, and how full the last one is.

Earlier this year one candidate model nearly won the job. It compiled cleanly every time, invented nothing, and took a perfect grounding score. It also produced, in ten out of ten runs, a two-page document whose second page was between 10% and 24% full. A few stranded lines under most of a page of white space. Typesetters call a single abandoned line an orphan; we borrowed the word for the page. That one defect disqualified an otherwise excellent model. Curious whether we had caught a quirk, we ran a page-fill probe across its whole model family. Underfill was systemic: roughly 72–85% fill even on the single-page documents. The note in the eval log says it plainly: no model self-manages page density.

I suspect the same is true of most humans, though I can't put a percentage on that. Deciding what to say and deciding how it sits on the page are different acts, and the second one usually gets skipped, so documents end wherever the material happens to run out. A senior résumé dribbling 15% of its content onto page two. A junior résumé inflated until it touches the bottom margin. Those look like opposite problems, but they misstate the same thing: how much substance is actually behind the document. The spill claims your evidence needed two pages when it didn't. The inflation claims a page of evidence you don't have.

The junior case is where we had to be most careful, so we wrote the trap down before running the eval. The test profile held roughly 250 words of career material, about two years of work, for a fictional candidate. The pre-registered rule: if the model fills 90% or more of a page from that little source, that is a red flag for padding, and a human reads the PDF hunting for filler, repetition, or bloated rephrasing. The honest outcome, we wrote, is a well-set, possibly underfull single page. The model then in production passed: ten out of ten runs came out as one page filled 76–83%, and the manual read found normal typesetting and no stretching. A page with some air at the bottom was the passing grade.

I want to underline that, because most résumé advice pushes juniors the other way: fill the page, look substantial. Filling a page from 250 words of material means writing things that are not material.

The rules that ended up in our full-document LaTeX prompt work fine for people. When a page runs short, first add supported substance that got trimmed in editing. If there genuinely isn't any, scale the typography instead: a slightly larger font, wider margins, more generous spacing between sections. The prompt's own phrasing is "Adjusting whitespace is honest; inventing content is not." It also carries a working number: a filled page at 10–11pt with normal margins holds roughly 430–520 words of body content, double that for a true two-pager. And when in doubt between running slightly under budget and spilling over, it says: "a 90%-filled final page beats a nearly-empty extra page every time." The default job-specific LaTeX path now uses a server-owned layout with fixed typography; there, the model adjusts content selection instead.

The LaTeX path also has a one-shot repair for a nearly-empty extra page. The server asks for compression: drop or tighten existing bullets and spacing, without adding factual content. This is a repair attempt; it does not guarantee a shorter document.

The word band is approximate, though. A denser template or a different font moves it, and we have had to tune it per style. Treat 430–520 as a starting estimate, not a law.

So, practically. Count the body words of your résumé, everything except the header. Divide by about 470. If the answer rounds to one, make it one page. If it rounds to two and you have a long record of substantial, relevant roles, make it two full pages. Then look at the rendered last page. If the content stops less than halfway down, cut until the document is a page shorter. And if a thin history leaves your single page a little underfull, adjust the spacing, and leave it alone.