A viral thread claims a man "fired yesterday" landed a remote US job by uploading his resume to an AI tool and blasting 500+ applications. Let's look inside the funnel these tools actually create — and compare the marketing math to evidence-based reality. Drag the funnel to explore.
An Applicant Tracking System parses your resume into structured fields (titles, skills, dates, education), then scores it against the job description's exact keywords, titles, and required skills. Say "ML" when the posting says "machine learning" and you can score zero on that requirement. A mass-blasted identical resume can't match 500 different keyword sets, so most copies score below the recruiter's cutoff and are never seen by a human.
Newer tools map your resume and each job posting into high-dimensional vectors (embeddings) and rank jobs by cosine similarity — how close the vectors point. This catches synonyms that keyword matching misses, but it's still lossy: it can't verify seniority, location eligibility, visa status, or whether you can actually do the work. "500 matched jobs" usually means "500 postings above a similarity score," not 500 jobs you're competitive for.
Slide personalization to the right and watch the funnel: conversion at every stage rises, and even though tailoring costs more minutes per application, the effort-per-interview number usually improves. The evidence-backed playbook is boring: pick a small set of roles you genuinely match, mirror the job description's real vocabulary in your resume, write two specific sentences instead of a template letter, and pursue referrals — referred candidates interview at several times the rate of cold applicants. Automation is fine for finding jobs; it is a poor substitute for being a fit for them.