Pretty resumes are failing where plain ones succeed, and the fix turns out to be a 1970s typesetting language.
Job seekers are discovering that using ChatGPT to generate resume code in LaTeX - a document markup language originally built for academic papers - then compiling it through free tools like Overleaf, produces clean PDFs that applicant tracking systems (ATS) can actually read. The approach is gaining traction among applicants who designed polished resumes in tools like Canva only to get ghosted by automated screening.
The core problem is straightforward. Most ATS platforms are text parsers at heart. They pull your resume apart looking for job titles, dates, skills, and education. Multi-column layouts, text boxes, custom fonts, and layered design elements from visual design tools produce PDFs where that parsing breaks down. The ATS sees garbled text, misattributes your job title to the wrong company, or skips sections entirely. Your resume looks great to a human but reads like nonsense to the robot that screens it first.
LaTeX output avoids this because it produces structurally simple PDFs with a clear text hierarchy. There are no hidden layers, no floating text boxes, no embedded design elements that confuse parsers. The text flows in a single, predictable stream.
The ChatGPT workflow is simple: paste your resume content into a prompt asking for LaTeX code formatted as a professional resume, copy the generated code into Overleaf (a free online LaTeX editor), compile it, and download the PDF. No need to learn LaTeX syntax yourself.
This is not a magic bullet. A well-parsed bad resume is still a bad resume. And some ATS platforms handle designed PDFs just fine. But if you have been applying consistently with zero callbacks, the formatting layer is worth testing before you rewrite your entire work history. The switch from a visual template to a LaTeX-compiled PDF takes about ten minutes with ChatGPT handling the code.