Improve Your Resume with AI: Prompts That Work

To improve your resume with AI you don't need a special tool: ChatGPT, Gemini or Copilot do the writing well if you give them good instructions. A good prompt is worth what it forbids the model from doing: inventing numbers, filling your resume with tools you never used, and turning your experience into generic filler. We tested every prompt on this page with a real resume against a real job posting, and we only published what worked. What AI doesn't do on its own is see your PDF the way an ATS reads it or check it against the job posting; that's covered below too.

Already have a resume and want the software to read the whole thing before you start? Convert it to ATS format for free: upload the PDF, see which parts the machine can't read, and get the same content in a clean single column. You can also choose the Executive template. No job posting or credit card needed.

How do you make a resume with AI?

With an AI chat and your current resume, or with a resume builder. The two routes serve different purposes. ChatGPT, Gemini, Copilot or Claude can take your resume plus a job posting and do the hard editing work in seconds: detect what the posting values, find the gaps, rewrite weak bullets in the employer's language and draft the cover letter. Used well, an AI-written resume is a real improvement over a blank page, and this guide shows you exactly how to do it.

The honest fine print isn't capability. It's that a chat gives you text, and a job search needs more than text: a document that machines read well, a version for each job and a way to know which version went where. We get to that after the prompts, because the prompts are what you came for.

One rule before you paste anything: the model only knows what you give it. Each prompt below expects two inputs: your current resume as plain text and the full text of a specific job posting. The better the inputs, the better the output.

Which AI is best for making a resume?

It depends on the step, not the brand. There are three types of tool, and each one solves a different part:

If you already have a resume, the short path is a chat to write and a reviewer to check. If you're starting from zero, a builder saves you the design work, but check that the template is one column, like the Harvard format resume. For writing, the large chats are capable and the prompts in this guide work in them. The real difference isn't between models; it's between writing the text and checking the file.

Prompts to improve your resume with AI (copy them)

They're written and tested in ChatGPT, and they work in any AI chat. We ran the nine prompts with a real software engineering resume against a real full-stack developer job posting. Two failures kept showing up in testing: the model invents metrics that aren't on the resume, and it adds the posting's tools to your skills even when you've never used them. The prompts below include the clauses that stopped both. Don't delete them.

1. Adapt your resume to a specific job posting

Use it as the main prompt: one posting, one rewrite.

Here are my resume and a job posting. Rewrite my resume to target this posting. Rules: use only experience, tools and numbers that are already in my resume. Don't invent metrics. Don't add skills from the posting that I don't have; instead, list them separately at the end under the heading
"Missing vs. the job posting" so I can decide what to do about them.
Reorder and rephrase so my most relevant experience is at the top, using the posting's vocabulary where it truly applies to my case.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

A good output looks like your own resume, reordered and refined, with an honest list of gaps at the end. In our test, without the "don't invent" rules, the model added an invented percentage and slipped a database the candidate had never used into the skills. With the rules, it flagged them as gaps.

2. Gap analysis before rewriting anything

Use it first, when you don't know whether it's worth tailoring the resume to that posting.

Compare my resume against this job posting. Give me three lists:
1) requirements I clearly meet, 2) requirements I meet partly, and
3) requirements I don't meet. For each item in list 2, quote the exact line of my resume you're basing it on, and suggest an honest way to word it that makes the match visible without exaggerating.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

The rule to quote the line is what makes it work: in testing, it forced the model to justify every "partly" instead of rounding in your favor.

3. Keyword check against the job posting

Use it to see the match the way a filter sees it, before and after you edit.

List the technical skills, tools and phrases this posting repeats or emphasizes. Then check my resume against that list and mark each one: present (quote my line), present under a different name (show both words) or missing. Don't rewrite anything yet.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

Expect output in three groups. The valuable one is "different name": in our test, the resume said "RESTful APIs" where the posting asked for "server-side API routes," a match a literal filter can miss.

4. Rewrite one bullet, result first

Use it bullet by bullet in your experience section.

Rewrite this bullet from my resume so it starts with the result and has no filler words. Keep all the facts; don't add any number or tool that I haven't given you. If a number would make it stronger and I haven't given it to you, ask me for it instead of inventing it.

BULLET: [paste a bullet]

CONTEXT (the job I'm targeting): [paste the job posting]

A good output is one punchier line, or a question back. In testing, the clause that asks instead of inventing is the only thing that separates you from a made-up "improved by 35%."

5. Find the numbers you didn't know you had

Use it when your resume has responsibilities but no numbers, which was exactly the state of our test resume.

Read my resume and ask me at most five questions that let us add real numbers: team sizes, frequencies, volumes, deadlines, money, users. Only ask about things my resume already mentions. When I answer, rewrite the corresponding bullets using my answers and nothing else.

MY RESUME: [paste your resume]

In our test, it produced exactly the right questions ("how many people are on the team you lead?", "how often were the releases shipped before and after the pipeline you built?"), which is how quantifying should work: the numbers come from you.

6. Rewrite your professional profile

Use it when the body of the resume is already tailored.

Rewrite my professional profile for this posting in three sentences at most: who I am professionally, my strongest points most relevant to this posting using its own vocabulary, and where I'm headed. Use only what's in my resume. No empty words, no "passionate," no claims you can't point to in my experience.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

7. Ask for the recruiter's objections

Use it when the resume seems finished and you want the cold read.

Act as the person hiring for this posting. You have thirty seconds with my resume. Tell me plainly: what would make you hesitate, what looks confusing or unexplained, and what you'd want to ask me before shortlisting me. Don't rewrite anything; just give me the objections in order of importance.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

Our test surfaced objections a proofreader would never catch: no database in a full-stack resume, and a jump from junior to lead at the same company that needs a line of context.

8. Write the cover letter

Use it after tailoring the resume, so the letter inherits the same vocabulary.

Write a cover letter of 200 words at most that connects three concrete things from my resume with three concrete requirements in this posting.
Direct, neutral tone; no "I'm passionate about." Don't invent experience.
If the posting asks for something I don't have, leave it out or address it honestly in a single line.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

The honest-line option earned its place in testing: for a requirement we didn't meet, the model wrote a believable sentence about nearby experience instead of pretending.

9. Prepare for the interview from your gaps

Use it once the application has been sent.

Based on my resume and this posting, list the ten questions I'm most likely to be asked in the interview. Mark the ones that point to the gaps between my resume and the posting's requirements. For the three hardest ones, give me an outline of a strong, honest answer using only my real experience.

MY RESUME: [paste your resume]

JOB POSTING: [paste the job posting]

How to improve your resume with AI, step by step

The prompts work better in a fixed order, and the order isn't "rewrite first." This is the sequence that held up in testing, whichever chat you use:

1. Gap analysis first (prompt 2). If the posting asks for five things you don't have, no rewrite fixes that. Better to know in two minutes.
2. Keyword check (prompt 3), to see which words the posting repeats and which of yours need renaming.
3. The tailored rewrite (prompt 1), now that you and the model know where the real match is.
4. Bullet passes (prompts 4 and 5) on the three or four lines that carry the application.
5. The profile last (prompt 6), because it should summarize the tailored resume, not the old one.
6. The cold read (prompt 7) before you send anything.

The method itself isn't tied to any AI: what to underline in a posting, how to reflect its language honestly, and what a recruiter really looks for are explained without relying on any tool in how to tailor your resume to a job description. An AI chat is a fast way to run that method.

Hoja de vida with AI

In Colombia, Ecuador or Peru, the method is the same: the prompts work just as well if you say hoja de vida instead of "CV." What changes are the conventions of the final document. A hoja de vida for a portal reads best in one column, with sections in Spanish and no photo, and you don't need the Minerva or SIGEP format for a private company. The full rules, with an example, are in ATS-format hoja de vida. Write the text with AI and paste it into a structure like that, not into a design template.

Is it worth improving your resume with AI? Yes, with two limits

For the writing itself, it's well worth it. Rewrites are fast, the vocabulary adjustment is strong, and with the clauses above, the honesty problem is under control. If the question is whether improving your resume with ChatGPT or another AI beats doing it alone in a text editor, the answer from our tests is a clear yes.

The first limit: the chat gives you text, and the employer's filter reads your PDF. No AI chat has any idea what your final document looks like or whether a parser can read it in order. That part is entirely on you, and it fails more often than people expect: we ran four polished Canva templates through a real parser, and two were read badly out of order, one losing the entire work history. Whatever ChatGPT writes, the layout where that text ends up decides whether a machine can read it. The two-minute test: select all the text in your finished PDF, paste it into a plain text editor and see whether it comes out in order.

The second limit: volume. Getting interviews today takes many more applications than most people expect, and no one should count on results from ten. Tailoring your resume isn't a trick you do once for your dream job; it's something you repeat with every serious opening, week after week. Do it in a chat and the mess arrives fast: thirty openings means thirty conversations, with versions of your resume buried in the history, no view of the document, no saved versions and no record of which text went to which company. Writing was never the problem. The workflow is.

How to review your resume with AI before applying

Reviewing isn't asking a chat to "improve" the resume again. It's checking two things about the file you're about to upload: that it reads in full, and that it contains the job posting's words. A chat can help with the second if you paste the posting (prompt 3 above), but it can't check the first, because it never sees your PDF.

That's what resume reviewers are for. Dopplio is one, and it uses AI for what a chat doesn't do: it reads your PDF the way an ATS extracts it, compares it with the specific job posting you paste, and tells you which words are missing and where to put them, without inventing experience. Here's how to use it:

1. Just want to know whether your resume passes ATS filters? The ATS checker gives you a score against a job posting, free and without an account.
2. Want to tailor it to the opening? Compare your resume with the job posting in Dopplio: you see the missing requirements, the rewritten bullets with the posting's vocabulary, and the version ready for that application. It requires a free account.
3. Did you write your resume in a design template? Convert it to ATS format first, so the text you wrote with AI reads in order.

A workspace for each version (what a chat can't do)

This is the part we built, so here's the description without beating around the bush. Dopplio is a workspace for your resume: it tailors it to a job posting the way the prompts above do, and then does everything a chat can't. You edit the result in a real editor, you see the actual document while you work, every version is saved, each resume's history stays attached to it, and you can track which version went to which application. It also runs an ATS-style review on the result, and the templates are machine-readable by design: one column, real text, standard headers, so the layout problem from the section above can't happen.

With the free account you get three applications per month; the Pro plan, at $15 a month, makes them unlimited. The honest way to decide: if you apply now and then, the prompts above plus discipline are enough. The moment you apply in volume, convert your resume to ATS format, bring it into the workspace and let it carry the versions, because the mess isn't a flaw on your part. It's what a chat produces at scale.

Frequently asked questions