Sophisticated ATSs use CV parsers such as Text Kernel, Rchili, and Dextra.
They don't just parse; they also return structured data from the CV, such as personal information, skills, work history, and dates.
Even for LLMs, I wrote a CV parser that uses Mistral OCR to extract the text and an LLM to structure the data, with great success, even for multilingual CVs.