Our methodology

What the roast can see—and what it cannot

RoastMyResume is an automated second opinion on the words and structure in your resume. It is not a recruiter, a visual design audit, or a promise that an employer will call.

Maintained by RoastMyResume · Last reviewed September 26, 2026

What happens after you upload

  1. 1.The server accepts a PDF, DOCX, or plain-text file and extracts selectable text. It does not take a screenshot of the page.
  2. 2.The extracted text is sent to a general-purpose language model through Groq. The model receives instructions to critique only the information present in the resume and not invent missing facts.
  3. 3.The response is checked against a fixed output structure before it is shown. The uploaded file is then removed from the application server's temporary storage.

How Resume Boost rewrites work

Below each roast, you can ask for paste-ready fixes. The text already extracted from your resume is sent back to the model, with an optional job description you paste in. It returns a rewritten summary, before-and-after bullet rewrites, three structural fixes, and, when a job description is supplied, the posting's keywords your resume covers or misses.

  • The model is told to rewrite only facts that are already in the resume. A number or result it cannot know is left in [brackets] for you to fill in or delete.
  • A second instruction pass checks each rewrite against its original and removes invented claims. It is not perfect, so read every line before you paste it.
  • Keywords are filtered on our server so only terms that literally appear in the job description are shown.

How the four scores work

Overall

A combined judgment of clarity, evidence, organization, and readiness. It is a review score, not a hiring probability.

Content

Rewards clear ownership, relevant detail, concrete outcomes, and numbers that are already in the resume.

Format and structure

Looks at section order and text organization after extraction. It cannot judge margins, font choices, color, or visual balance.

General ATS readability

Checks conventional headings, chronology, readable text, and role-relevant language. It does not test your file inside every employer's software.

The important limitations

  • The roast does not see a job description, so its scores are not a job-match score. Resume Boost can list a posting's keywords, but it cannot know how a particular employer weighs them.
  • Text extraction can reveal broken reading order, but it cannot prove how every applicant tracking system will parse the original file.
  • The model does not know the context behind a layoff, career change, confidential result, or deliberately omitted detail.
  • Scores can vary slightly between runs. Use the specific criticism, not a one-point score change, to decide what to edit.

Editorial standard

Our guides are maintained by RoastMyResume. We do not invent a staff recruiter or pretend the model was trained on a private collection of successful resumes. Product claims are checked against the live application, and factual career guidance is sourced where a reliable public reference exists.

For practical resume guidance, we use sources such as the U.S. Department of Labor-sponsored CareerOneStop Resume Guide. For model-provider data handling, see Groq's current data documentation. Our own handling is described in the privacy policy.

Use it as a second opinion

Keep the useful criticism. Ignore any suggestion that would make your resume less accurate.

Roast my resume