I spent the last month running an experiment: use AI for every stage of job searching, from finding openings to negotiating offers. Not just ChatGPT-write-my-resume stuff. Custom tools, scripts, actual systems.
Some of it worked. Some was a complete waste of time. Here's the breakdown.
What Actually Worked
1. Keyword Extraction From Job Descriptions
The single most useful thing: pulling keywords from job postings and matching them against your resume before applying.
I built a browser-based ATS checker that does this. Paste your resume, paste the job description, get a match score. No server, no data collection. It runs entirely in your browser.
The insight: most rejection happens before a human sees your resume. ATS systems filter by keyword density. If the posting says "Kubernetes" five times and your resume says "container orchestration," you're probably filtered out. Dumb? Yes. Fixable? Also yes.
Time saved per application: ~8 minutes. Over 30 days and ~45 applications, that's 6 hours I didn't spend guessing whether my resume matched.
2. Cover Letter Templates With Variable Slots
Writing cover letters from scratch for each application is miserable. Writing none at all tanks your response rate. I tested this: 12% with cover letter vs 3% without, from 45 apps.
The fix: build templates with slots for company name, role, specific requirements, and one personal hook. I made a cover letter generator that handles this. Pick a tone, fill in the details, get a draft.
Is it perfect? No. You still need to tweak it. But it takes 3 minutes instead of 20.
3. LinkedIn Headline Testing
Your LinkedIn headline is the first thing recruiters see in search results. I tested 4 different headline formats over 4 weeks:
- Week 1: Generic title ("Senior Developer")
- Week 2: Title + specialty ("Senior Developer | React & Node.js")
- Week 3: Result-oriented ("I Build Web Apps That Handle 10M+ Requests")
- Week 4: Problem-oriented ("Helping Teams Ship Faster With Better CI/CD")
Week 3 got the most profile views. Week 4 got the most recruiter messages. Your field may differ, but the pattern held: specific beats generic, outcomes beat titles.
I built a headline generator if you want to test variations quickly.
4. AI Interview Prep
This one surprised me. Feeding a job description to Claude and asking "What are the 10 hardest technical questions for this role?" produced genuinely useful prep material. Better than generic "top interview questions" listicles because it pulls from the actual requirements.
Follow-up prompt: "For each question, give me a STAR-format answer outline based on this experience: [paste resume]." The outlines weren't usable verbatim, but they eliminated blank-page syndrome before every interview.
What Didn't Work
AI-Written Resumes
Letting AI write your resume from scratch produces something that reads like a corporate mission statement. Every bullet point starts with "Spearheaded" or "Leveraged." Recruiters spot this immediately.
Better approach: write your own bullets, then use AI to tighten the language and add quantified results. "I worked on the payment system" becomes "Reduced payment processing errors by 34% across 2M monthly transactions." You need the real numbers though. Making them up is worse than being vague.
Mass Application Bots
Tempting, but counterproductive. Tools that auto-apply to 100 jobs per day sound efficient until you realize your response rate drops to near zero. Companies track application quality, and spray-and-pray signals desperation.
Focused applications, 2-3 per day with tailored materials, consistently outperformed bulk approaches in my experiment.
AI Networking Messages
"Hi [Name], I noticed your work at [Company] and would love to connect." Everyone knows this is AI-generated now. Every LinkedIn user gets 15 of these per week. They go straight to the trash.
What worked instead: actually reading something the person wrote or built, and referencing it specifically. Takes 5 minutes per message. Send 3 good ones instead of 30 generic ones.
The Numbers
After 30 days:
- Applications sent: 45 (targeted, not bulk)
- Response rate: ~16% (7 responses)
- Interviews: 4
- Offers: 1
- Time spent on applications: ~2 hours/day (down from ~4 hours/day before AI tools)
The AI tools didn't make the job search painless. Nothing does. But they cut the grunt work in half, which meant I could spend more time on the parts that matter: networking, interview prep, and targeting the right companies.
Tools I Used
All free, all browser-based:
- Resume ATS Checker -- keyword matching against job descriptions
- Cover Letter Generator -- template-based drafts with tone selection
- LinkedIn Headline Generator -- test different headline formats
- Salary Negotiation Script Generator -- scripts for offers, raises, counters, and promotions
For deeper prep, I put together a Job Search AI Toolkit with 100+ prompts covering salary negotiation scripts, behavioral interview prep, and networking templates. The free tools above cover the basics though.
If you're currently job searching: the ATS checker alone will probably change how you think about resume optimization. Try it before your next application.
Top comments (0)