How to Make an ATS Resume with ChatGPT (One Prompt, Done Right)
You don't need a course to make an ATS resume with ChatGPT. You need one good system prompt, one user prompt, and an honest read on where the output stops being enough. This guide hands you both prompts and the line where the machine quits.
Most "ChatGPT resume prompt" advice gives you a one-liner that produces a generic rewrite that reads like every other ChatGPT resume. The prompt below is built differently: a structured role, a numbered procedure, and hard rules that keep it from faking your experience or breaking the parser.
Key Takeaways
- ChatGPT is strong at the rewrite layer, language, structure, accomplishment framing. Use it for that.
- It can't see the live listing's ATS-weighted keywords or score your draft against the parser. That gap is yours to close.
- Paste the full job listing as text. ChatGPT can't open a URL in most flows.
- Use one system prompt (the rules) plus one user prompt (your resume and the listing). Save the system prompt in Custom Instructions or a Project so you don't rebuild it every time.
- ChatGPT loves tables, bold, and fancy characters. Those break ATS parsers. Flatten the output to plain text before you send.
Why ChatGPT is good at the resume layer, and the one thing it can't see
ChatGPT is a strong language model. Feed it your resume and a job listing as text and it will tighten your verbs, convert duties into accomplishments, and align your summary to the role in about a minute. That's real work, and it's the part most people get wrong by hand. Don't dismiss it.
Here's the wall. ChatGPT doesn't know which keywords the employer's Applicant Tracking System weights heaviest. The parser runs on the company's side. It scores placement, frequency, and proximity in ways no chat model can see, because the model never touches it. ChatGPT also can't open a job listing's URL in most consumer flows, so it's working blind unless you paste the text in. A Harvard Business School study on hiring found that more than 75% of qualified applicants are filtered out before a human ever reads the resume — which means clearing the parser matters at least as much as the rewrite itself.
That's the 70/30 split. ChatGPT gets you about 70% of a strong, tailored resume, the language and the structure. The last 30%, matching the listing's actual scored terms and confirming the format survives the parser, is on you. The model writes the draft. It does not score it. Harvard Business Review's research on hiring argues that ATS-driven hiring processes filter out qualified candidates before a human ever evaluates them — which means surviving the parser matters at least as much as having the right experience. ChatGPT can help you write it. It can't confirm the filter ever lets it through.
The ChatGPT ATS resume system prompt
This is the rulebook. It sets ChatGPT up as a resume writer and ATS analyst, defines the procedure, and bolts on the guardrails that stop it from inventing experience or formatting your resume into a shape the parser chokes on. Paste it once into Custom Instructions or a Project (see below), and it applies to every resume you run through that chat.
You are an expert resume writer and ATS (Applicant Tracking System) analyst. Your job is to rewrite a candidate's existing resume so it mirrors the language and keywords of a specific job description, while staying completely truthful to the candidate's real experience.
Follow this procedure every time, in order:
1. Read the candidate's resume and the job description in full before writing anything.
2. Identify the 8 to 12 most emphasized terms in the job description: skills, tools, methodologies, certifications, and repeated phrases.
3. For each emphasized term, check whether the candidate's resume shows real evidence of it. Only use terms the candidate actually supports.
4. Rewrite the resume so supported terms appear naturally, prioritizing the summary and the most recent roles.
5. Rewrite each work-history bullet as: strong action verb + specific action + outcome. Quantify an outcome only when the original resume already supports a number. Do not invent metrics.
6. Output the finished resume, then a short list of job-description terms you found NO evidence for, so the candidate knows their real gaps.
Hard rules:
- Do not invent experience, employers, titles, dates, degrees, or numbers. Never fabricate.
- Do not add a skill, tool, or achievement the candidate has not demonstrated.
- Use ATS-safe formatting only: plain text, no tables, no columns, no text boxes, no graphics, no icons, no information in headers or footers.
- Use standard, parser-recognized section headings: Summary, Experience, Skills, Education, Certifications.
- Output the resume as plain text. No markdown tables, no bold, no special characters, no emojis. Use simple dashes for bullets.
- Keep every claim factual and verifiable against the source resume.
If the resume or job description is missing, ask for it before writing.
The user prompt
This is what you send in the chat after the system prompt is in place. Paste your current resume where it says so, then paste the full text of the listing you're targeting. Don't paste a link, ChatGPT can't open it.
Here is my current resume and the job description I'm targeting. Rewrite my resume to match this listing using the rules above. Do not invent anything I haven't actually done.
MY CURRENT RESUME:
[PASTE YOUR RESUME]
THE JOB DESCRIPTION:
[PASTE THE JOB DESCRIPTION HERE]
How to run it
You have two ways to load the system prompt so you're not pasting it every time.
Custom Instructions. Open ChatGPT settings, find Custom Instructions, and paste the system prompt into the field that asks how you'd like ChatGPT to respond. Now every new chat starts with the rules already loaded. Good if you're tailoring to a handful of jobs this week.
A Project. If you're on a plan with Projects, make one called "Resume tailoring" and drop the system prompt into the project instructions. Every chat inside that project inherits the rules, and you can keep your master resume in the project files so you only paste the listing each time. Better if you're running a long search.
Either way: system prompt sets the rules once, user prompt carries the resume and the listing for each job. Paste the listing as text. The model is blind to a URL.
The honest friction
Here's where doing it yourself stops being free, because it costs you time and judgment instead of money.
You re-prompt for every single job. The system prompt is reusable. The work isn't. Every new listing means a fresh paste, a fresh run, and a fresh read-through. Twenty applications is twenty rounds.
You're the only judge. ChatGPT doesn't score your draft against the parser, because it can't reach the parser. There's no number telling you the resume will clear the filter. You eyeball it against the listing and hope you caught what matters. That's a guess, not a score.
The format will betray you. ChatGPT defaults to markdown, bold text, tables, and tidy little symbols. Those are exactly what break ATS parsers, which read plain text and choke on the rest. You have to flatten every output by hand: strip the bold, kill the tables, replace fancy bullets with plain dashes, paste it into a clean document. Miss one table and the parser can scramble the whole section.
No interview script. A tailored resume gets you in the room. It does nothing for the conversation once you're there. ChatGPT can mock-interview you in a separate session, but it won't build prep from the same listing-and-resume context automatically. That's a second job.
It can't open the listing. Worth repeating because people keep pasting URLs. ChatGPT in most consumer flows can't browse to the link. Copy the listing's responsibilities, requirements, and "about the role" sections as text, or the model is guessing at a job it never saw.
If you want the other model versions of this exact method, the Claude ATS resume prompt leans on XML-tagged inputs and a thinking step, and the Gemini ATS resume prompt adds extra guardrails against over-formatting. Same backbone, tuned per model.
FAQ
Can ChatGPT make an ATS-friendly resume? Yes, for the language and structure layer. ChatGPT rewrites bullets into accomplishments, aligns your summary to the role, and outputs a clean draft in about a minute. It can't see the live listing's ATS-weighted keywords or score your draft against the parser, so you still do a manual keyword pass after.
Is the free version of ChatGPT enough for this? Yes. The system prompt and user prompt above work on the free tier. The only limitation is that free ChatGPT can't browse the web, so you paste the job listing text directly — which is the right move regardless of your plan, since a URL in the prompt is useless without live browsing.
Why does the formatting break in the ATS?
ATS parsers read plain text. When ChatGPT formats output with bold headers, markdown tables, or special bullet characters, many parsers interpret those as garbage or drop the lines entirely. The system prompt above forces plain text, but you should still paste the output into a plain .txt file and strip any surviving markup before submitting.
Skip the friction
The prompt above is real and it works. The tedium is also real: re-prompting every job, flattening the formatting by hand, self-judging output you can't score, and then building interview prep from scratch.
That's the exact stack Gate Crashers handles in one session. You paste your resume and the listing. You get three differentiated resume versions tailored to what that specific listing weights, already in clean ATS-safe formatting, plus a 12-question interview script built from the same context. No re-prompting, no flattening, no guessing. Pay once, no subscription. $4.99. Files are yours.
ChatGPT writes the rewrite. You score the last 30% by hand, every time, for free, or you pay $4.99 once and skip it. Both paths are honest. Pick the one that fits the week you're having.