Most "AI for job seekers" advice is generic. "Use ChatGPT to rewrite your resume." Cool. Then what?
I've been job-search adjacent for the last 3 months — building tools for it, talking to people in it, watching what works. The actual AI workflows that move the needle aren't the obvious ones. Here are 7 specific prompts I see real candidates using to land interviews in 2026.
Each is copy-paste ready. Replace the bracketed parts.
1. The keyword-aware resume rewrite
Most people ask AI to "make my resume better." That's vague and you get a generic result. This works:
Here's a job description: [paste full JD]
Here's my current resume bullet for a similar role: [paste 1 bullet]
Rewrite this bullet so it (1) uses the exact terminology from the job description where applicable, (2) adds a quantifiable result if missing, and (3) keeps the original meaning. Output 3 versions.
You then pick the version that sounds most like you and least like a robot. Repeat for each bullet.
2. The recruiter screen prep
The phone screen is where most candidates lose interviews they should have won. Use this 24 hours before:
You are a senior recruiter at [company]. Based on this job description [paste JD], generate the 8 most likely questions you'd ask in a 30-minute initial phone screen. For each, also list (a) what you'd be listening for in a strong answer, and (b) red flags that would end the call.
Then practice answers out loud. Not in your head. Out loud.
3. The "STAR" answer compressor
Your behavioral answers need to be tight — 90 seconds max. Use this to compress:
Here's my behavioral story: [paste 200-400 word story]
Compress this to a 90-second STAR answer. Specifically: 10 seconds situation, 15 seconds task, 50 seconds action with concrete details, 15 seconds quantified result. Strip any context that isn't load-bearing.
This is where most candidates ramble. The compression is brutal but necessary.
4. The application tracker
Most people lose track of where they applied. Have AI build a tracker:
Generate a JSON schema for tracking job applications with these fields: company, role, source, date_applied, contact_name, contact_email, status (5 stages: applied/screened/interviewing/offered/closed), follow_up_date, notes, salary_range, decision_reason. Output as a CSV header row and 3 example rows.
Save the CSV. Add rows as you apply. Sort by status weekly.
5. The follow-up that gets responses
Boilerplate follow-ups get ignored. This generates one that doesn't:
I applied for [role] at [company] [days] ago. I want to send a short follow-up to the hiring manager [name].
The job description mentioned [specific challenge from JD].
My background includes [1-2 specific relevant experiences].
Write a 5-sentence email that (1) references the specific challenge in their JD, (2) demonstrates I read the role description carefully, (3) ties one of my experiences to it, and (4) ends with a question that requires a yes/no/maybe response (not "let me know your thoughts").
The "yes/no/maybe" question is the trick. Open-ended questions get ignored. Specific ones get answered.
6. The salary research synthesizer
Don't go into negotiation blind:
For a [role] in [city or "remote"] with [years] years experience and skills in [stack], synthesize what you know about typical 2026 compensation ranges. Break it into base / bonus / equity. Cite the typical sources (Levels.fyi, Glassdoor, LinkedIn) and note the variance you'd expect. Then suggest a target range and a floor I should not accept below.
Use this to anchor your number before they ask.
7. The thank-you note that closes loops
After every interview, send a thank-you. AI helps make them not generic:
I had an interview today with [interviewer name] for [role] at [company]. They asked me about [specific topic].
I want to send a 4-sentence thank-you that (1) thanks them for their time specifically, (2) references something specific from our conversation, (3) adds one piece of relevant context I didn't get to share, and (4) leaves the door open without being pushy.
Make it sound like me — slightly informal, no jargon, no "thrilled" or "delighted."
The "no thrilled or delighted" instruction matters. AI defaults to those words and they read as fake.
What ties these together
Notice all 7 require specific input from you. The job description. Your stories. Your background. Your conversation.
Generic AI prompts give generic outputs. Specific prompts give specific ones. The prompt isn't the magic — your specific input is. AI just helps you process it faster.
A free toolkit for the specifics
If you want the underlying tools without crafting each prompt yourself:
- ATS Resume Checker — paste resume + JD, get the gap
- Keyword Extractor — extract terms from the JD before you write anything
- Cover Letter Generator — opens with a specific hook, not "I am writing to apply"
- Interview Prep — generates likely questions for your specific role
- Follow-Up Email Generator — implements prompt #5 above
All free, all browser-based, no signup. Full kit at charliemorrison.dev/tools.
What to do with this
Don't try all 7 at once. Pick prompt 1 and prompt 2. Use them on your next 3 applications. See if your interview rate changes.
If it does, add the rest. If it doesn't, the prompts aren't the problem — the inputs are. Improve the inputs first.
Which of these do you actually use? Or have you found something better? Comments are wide open.
Top comments (0)