How it works
A rules engine first. A writer second. Nothing invented.
Referred is two things bolted together with a strict seam between them: a deterministic check that knows the federal rules, and a rewrite that is only allowed to use what your resume and your answers give it.
1. The check
Twelve requirements, each a rule with a source: OPM's Agency Guidance on the Two-Page Limit on Resume Length (September 2025), the USAJOBS Help Center, and the announcement you paste. The check is ordinary code, not a model. It reads your file in your browser, counts the pages the way USAJOBS does, finds each role and its dates, and looks for what the rules require. The same resume gives the same report every time, which is why every row can carry its citation.
2. Measured against real announcements
We don't train a model on your data; we test our rules against the real thing. The announcement reader was measured on 6,000 USAJOBS announcements pulled in September 2026: it reads the "This job is open to" section and the closing date back correctly on every one, finds the stated requirements on 77% of them, and a filing clerk's resume "addresses" 1 of 4,790 requirement clauses, so a match means the words are really there. The Title 38 page-limit exemption was checked the same way, because applying the two-page rule to a VA nurse would be wrong.
3. The rewrite
A top-of-the-line language model rewrites your resume under rules it cannot break: two pages, every role keeps its title, employer, months and hours, the announcement's own wording wherever your experience supports it, every cut listed. It is given the requirement clauses the check extracted and the facts you typed, verbatim. Where a required fact is missing, it writes a marker instead of a guess, and the review shows you exactly where.
4. The second read
A separate top-of-the-line model reads the announcement and the rewrite blind, never told it is a rewrite, and scores fit. Then the twelve rules run again on the result, in your browser. The count you see on the review page is that second check, not the writer's opinion of itself.
Compared with asking a general chatbot
A general model is a fine writer. It is not a checker, and on a certified federal application the difference matters.
| Question | Referred | A general chatbot |
|---|---|---|
| The two-page limit | Counted from the file, the way USAJOBS enforces it, and the Title 38 exemptions read from the announcement. | Doesn't know the rule exists unless you tell it; can't see your page count. |
| Required content | Hours per week, month and year dates, employer, contact details, education, GPA where claimed: each checked, each with its source. | Sometimes mentions them; never checks them line by line. |
| The announcement | Who may apply, the closing date, the documents required and the stated requirements are read from the page you paste and matched against your resume. | Reads the announcement as prose; can't tell you what it requires or whether you answered it. |
| Facts it doesn't have | Asks you. A blank stays marked as needed. It never invents hours, months, employers, degrees or numbers. | Fills gaps with plausible text, which on a certified federal application is a false statement. |
| The same answer twice | The check is deterministic: the same resume gives the same report today and next month. | Different answers each time you ask. |
| A second opinion | A separate model scores the rewrite blind, then the twelve rules run again on the result, on your device. | Grades its own work, if at all. |
| Your resume | The check runs in your browser. Only a rewrite you ask for sends text to our server, and it isn't stored. | Kept and sometimes used for training, depending on the plan and the settings. |
What a human writer charges
A federal resume specialist charges $210 an hour (The Resume Place); a written federal resume typically runs $300 to $600, with a range from $149 to over $2,500 (2026 pricing guide). A writer sees one announcement and rarely tells you which rule the result meets. Referred is $49 for up to five rewrites of one announcement, and cites every rule, every time.
What we don't do
- We don't promise a referral, an interview or a job. Agencies decide.
- We don't write the Merit Hiring Plan essays. Those must be your own words.
- We don't store your resume. The check never sends it; the rewrite doesn't keep it.
- We aren't affiliated with OPM or USAJOBS. We read their published rules like anyone can; we just check them line by line.