Your AI Resume Writer Wants to Make You a Fraud (Here's How to Stop It)

The Problem with AI Resume Builders

I have spent way too much time on AI resume builders.

Most of them follow the same pattern: paste your old resume, tell it the job description, and watch it spit out a keyword-optimized version that looks nothing like your actual experience. It adds action verbs you would never use. It inflates responsibilities you never had. And worst of all, it makes claims you cannot back up in an interview.

The whole point of using AI should be to help you articulate what you actually did, not to fabricate a fantasy version of your career. But most tools and prompts out there treat resume building like SEO spam. Just jam in the keywords and hope the ATS lets you through.

That approach has problems. Big ones.

Why "Just Optimize for ATS" Is Bad Advice

Yes, ATS systems scan for keyword matches. Yes, you need to frame your experience in terms the system understands. But here is what the "optimize for ATS at all costs" crowd does not tell you:

Fabricating resume details can get you fired. Not just rejected. Fired after you get hired, when the background check comes back or when your manager realizes you cannot do what you said you could do.
Employers verify claims more often than you think. Especially for roles with security clearances or sensitive access. They check. And when your "led a team of 10" claim turns out to be "occasionally helped two interns," you are done.
Interviews expose lies fast. You might get past the ATS with keyword spam, but the moment someone asks you to walk through a project you did not actually work on, the conversation falls apart.

The better approach: use AI to help you present your real experience clearly and confidently. Not to make things up.

The Prompt That Asks Instead of Invents

I have tested a bunch of resume prompts. Most of them fail because they assume too much or fabricate details to fill gaps. This one is different.

Instead of taking your old resume and rewriting it in one shot, this prompt asks you questions. A few at a time. It collects the information it needs, section by section, and builds your resume from your answers.

When it does not know something, it asks. When you say "I do not have that," it moves on. No fabrication. No guessing.

It uses Claude's canvas feature to build the resume iteratively, so you can see each section as it develops and make changes before moving to the next part.

The Full Prompt (Copy This)

Here is the exact prompt I have been using. Copy it, paste it into Claude, and follow the conversation.

You are a resume writing assistant. Your job is to help me build an accurate, well-formatted resume by asking me questions about my experience.

RULES:
1. NEVER fabricate or invent details. If you don't know something, ask me.
2. Ask 2-3 questions at a time, not a wall of questions.
3. Collect ALL necessary information for a section before drafting (including dates, numbers, etc.)
4. When you have enough information for a section, present it as a draft.
5. After each draft section, ask if it's accurate before moving on.
6. Follow this order: Contact info → Professional Summary → Experience → Skills → Education
7. Use action verbs and quantify achievements when numbers are provided.
8. If I say "I don't have that," accept it and move on. Don't push.

For Experience entries, always collect: title, company, dates (month/year), and 3-5 specific accomplishments or responsibilities.

START by asking: (1) the role they're targeting, (2) their most recent position with company and dates, and (3) a brief description of what they do/did.

That is it. No complex setup. No examples to fill in. Just paste it and answer the questions.

How It Works

The prompt is built around a simple idea: conversational prompting improves output quality. Instead of trying to guess what you did or how to frame it, the AI asks you directly and builds from there.

It follows a specific order (contact info, summary, experience, skills, education) so you do not jump around. Each section gets drafted and confirmed before moving to the next one.

The key rules:

No fabrication. If the AI does not know something, it asks. Period.
Small batches of questions. You are not filling out a 50-question form. You answer 2-3 questions, then get a draft to review.
Explicit confirmation. After each section, it asks if the draft is accurate. You can correct it or move on.
Real numbers matter. If you managed a budget or led a team or shipped a product, the AI will ask for specifics. If you do not have numbers, it will not invent them.

What to Expect in the Conversation

When you paste the prompt, Claude will start by asking three things:

What role are you targeting?
What is your most recent position (title, company, dates)?
What did you do in that role?

Answer those, and it will draft your contact info and ask follow-up questions about your experience. From there, it builds section by section.

You will get questions like:

"What were your main responsibilities in that role?"
"Did you manage a team? If so, how many people?"
"Can you share a specific project you worked on and the outcome?"
"What tools or technologies did you use regularly?"

If you do not have an answer, say so. The AI will move on. It will not fill in blanks with guesses.

Running It Step by Step

Here is how a typical session goes.

Start the Session

Open Claude (you need a Claude account for canvas support). If you want to run your own local AI setup instead, I have written about how I use self-hosted tools for this kind of work.
Paste the full prompt.
Hit enter.

The AI will start by asking for your target role, most recent position, and a brief description of what you did. Answer in a few sentences. Do not overthink it.

Answer the Questions Honestly

The AI will ask 2-3 questions at a time. Answer them based on what you actually did, not what you wish you had done.

Examples of good answers:

"I was a junior developer on a team of 5. I worked on bug fixes and small features for the first 6 months, then took ownership of the user settings module."
"I managed a budget of around $50k for vendor contracts. No direct reports."
"I do not have exact numbers for how many users were affected, but the feature launched to all customers in Q3."

Examples of answers that will lead to problems:

"I basically ran the whole engineering team." (Were you the manager? Did you have direct reports? Be specific.)
"I increased revenue by 40%." (Did you personally drive that, or was it a team effort? What did you actually do?)

If you do not know something, say so. The prompt is designed to handle gaps.

Review the Canvas Draft

After you answer a few questions, Claude will draft a section in the canvas view (a live document that updates as you go). Read the draft carefully. Check for:

Accuracy. Does this reflect what you actually did?
Clarity. Is it easy to understand, or is it full of jargon?
Specificity. Are the accomplishments concrete, or are they vague and generic?

If something is wrong, tell the AI. It will revise the section.

Iterate Until Done

The AI will move through each section in order. After each one, it will ask if you are ready to move on.

If you want to make changes, speak up. If a section looks good, confirm it and keep going.

By the end of the session, you will have a full resume draft built from your actual experience, formatted cleanly, and ready to tailor for specific applications.

What This Prompt Won't Do

Let me be clear about what this prompt does not do:

It will not invent experience you do not have. If you have a 6-month gap in employment, it will not fill it with a fake consulting gig.
It will not guarantee you pass the ATS. It helps you articulate your experience clearly, but keyword matching depends on the role and your actual background.
It will not write your resume in one shot. This is a conversation. It takes 20-30 minutes. If you want instant results, this is not the right tool.
It will not make you sound like someone you are not. The resume it builds will reflect your real career, not a fantasy version.

This is a tool for articulation, not fabrication.

It helps with formatting, phrasing, and structure. It does not create experience out of thin air.

Sample Session (Redacted)

Here is a quick example of what a session looks like. I have redacted personal details, but the flow is real.

User: (pastes the prompt)

Claude: Got it. Let's build your resume step by step.

What role are you targeting?
What is your most recent position (title, company, and dates)?
Can you give me a brief description of what you do or did in that role?

User: I am targeting junior developer roles. My most recent position was Software Developer Intern at [Company], June 2024 to December 2024. I worked on bug fixes, wrote unit tests, and built a small internal tool for the support team.

Claude: Thanks. A few follow-up questions:

What technologies or languages did you primarily use?
Can you tell me more about the internal tool you built? What problem did it solve, and what was the outcome?
How big was the team you worked with?

User: Mostly JavaScript and Python. The tool automated a manual CSV export process the support team was doing every week. It saved them about 2 hours per week. I worked with a team of 8 developers, but I was the only intern.

Claude: Great. I will draft your contact info and professional summary. What is your email, phone number, LinkedIn, and GitHub (if you have them)?

(The session continues from there. By the end, the user has a full resume built from real answers.)

Frequently Asked Questions

Will this prompt work with ChatGPT or just Claude?

The prompt itself will work with ChatGPT, but you will lose the canvas feature. Claude's canvas lets you see the resume build in real time in a side panel, which makes it easier to review and edit as you go. ChatGPT will still ask the questions and build the content, but you will need to copy/paste sections into your own document. If you have access to Claude, use it. If not, ChatGPT will still work.

How long does a session usually take?

Plan for 20-30 minutes if you have your details ready. It can go faster if you have a clear sense of your accomplishments and dates, or longer if you are working through how to frame your experience. This is not a "paste and go" tool. It is a conversation. Budget the time.

What if I don't have much experience to draw from?

Then be honest about that. The prompt will not invent experience for you, and that is a good thing. If you are early in your career, your resume will reflect that. Focus on what you did do: coursework, projects, internships, volunteer work, freelance gigs. The AI will help you articulate those clearly. A short, honest resume is better than a long, fabricated one.

Can I use this for cover letters too?

Yes, with modifications. The structure of this prompt is resume-specific (contact info, summary, experience, etc.), but the core idea—ask questions, collect real answers, build iteratively—works for cover letters. You would need to adapt the prompt to focus on the role, the company, and why your experience is relevant. I might write a separate prompt for that if there is interest.

What about LinkedIn optimization?

This prompt focuses on resumes, not LinkedIn profiles. LinkedIn has different conventions (first-person voice, longer descriptions, multimedia). You could use a similar question-based approach for LinkedIn, but you would need to adjust the format and tone. If you want help with LinkedIn specifically, check out the tools I recommend or reach out.

Your Turn

Copy the prompt. Run it. See what comes out.

If it asks a question you cannot answer honestly, that is data about your resume, not a bug in the prompt. It means you need to think about how to frame that experience, or you need to acknowledge that you do not have that particular skill or achievement yet.

This approach is not about gaming the system. It is about presenting your real experience as clearly and confidently as possible.

And in a job market full of inflated resumes and keyword spam, that might actually be your competitive advantage.

Want to dive deeper into how I think about career tools and tactics? Learn more about my approach here. If you are curious about running your own AI setup for this kind of work, I wrote about my editor's desk setup.