How to use AI to rewrite your CV without crossing the line into fabrication

The ATS Arms Race

Job hunting has become a software war. On one side, companies use automated Applicant Tracking Systems (ATS) to filter out hundreds of resumes before a human eye ever sees them. On the other side, candidates are using LLMs to tailor their CVs for every single application.

It makes perfect sense to level the playing field. Using models like Gemini or Claude to rephrase your bullet points to match a job description is a sensible strategy. It saves hours of manual editing.

But there is a tipping point. There is a fine line between translating your real-world experience into the keywords an ATS wants, and fabricating a professional persona that you cannot actually back up when you sit down in the interview room.

Your CV might tick all the algorithmic boxes, but if the human on the other side of the desk meets a completely different person, you have failed. Here is how to use AI to optimise your CV ethically, without losing your professional identity.

1. Keyword Alignment vs. Experience Inflation

The most common use case for AI in CV writing is keyword matching. You paste the job description, paste your CV, and ask the AI to align the two.

This is ethically fine when it is a matter of terminology. For example, if you wrote "led a team of three developers" and the job description asks for "experience with agile team coordination," using the AI to rephrase your bullet point to reflect "agile team coordination" is simply translating your work into the employer’s vocabulary.

However, it becomes unethical when the AI introduces skills you do not possess. If the job description requires "production experience with Kubernetes," and the AI rewrites your minor Docker hobby project to sound like you have spent three years orchestrating enterprise Kubernetes clusters, you have crossed into fabrication.

To prevent this, use a strict prompting constraint. When tailoring your CV, try this prompt layout:

> "Review this job description and my raw CV. Identify keywords in the job description that match tasks I have actually performed in my history. Rephrase my existing bullet points to use their terminology. Do not add new technologies, frameworks, or responsibilities that are not explicitly present in my raw CV."

2. The "Action-Impact" Hallucination Hazard

A great CV bullet point follows the classic formula: Accomplished [X], as measured by [Y], by doing [Z].

If your raw notes only have the [X] and the [Z]—for example, "I rebuilt the checkout page using React"—the LLM will often try to invent the [Y] to make the bullet point stronger. It might output: "Rebuilt the checkout page using React, resulting in a 25% reduction in cart abandonment and a £50k increase in monthly revenue."

Unless you have the analytics dashboard to prove those exact numbers, you cannot include them. If you get asked during an interview, "How did you attribute that 25% drop specifically to the React rebuild rather than the marketing campaign?" you will freeze.

If you lack concrete quantitative data, ask the AI to focus on qualitative impact or technical outcomes instead. For example:

• Instead of: "Boosted team efficiency by 40%."
• Use: "Reduced deployment bottlenecks by automating the CI/CD pipeline, cutting average release times from three days to under an hour."

This is still highly impressive, entirely verifiable, and completely honest. If you are struggling to get your prompts to output these balanced, non-inflated bullet points, check out our prompt engineering hub for structured system prompts designed for professional writing.

3. Preserving Your Genuine Voice

If you let an LLM rewrite your entire CV from scratch, it will almost certainly sound like a corporate robot wrote it. It will be stuffed with words like leveraged, spearheaded, dynamic, results-driven, and synergy.

When a hiring manager reads fifty CVs in a row that all use the exact same polished, GPT-style syntax, they tune out. Worse, if your CV is written in a hyper-formal, flowery tone, but your introductory email or initial phone screen is casual and conversational, the jarring disconnect raises immediate red flags. They will suspect you didn’t write your own CV.

To keep your CV authentic:

• Keep it punchy: Tell the AI to use active verbs and a direct, unpretentious tone. Avoid passive voice.
• Review it aloud: Read your tailored bullet points out loud. If you would never say those words in a professional conversation with a peer, rewrite them.
• Don't outsource the summary: Write the "About Me" or professional summary section yourself. This is the only place on your CV where your personal voice can truly shine. Let the AI edit it for grammar, but do not let it write your professional philosophy for you.

4. The Interview Test

The ultimate ethical (and practical) litmus test for your CV is simple: Can you comfortably discuss every single word on this page for ten minutes without preparation?

If your AI assistant added a bullet point about "championing cross-functional alignment across stakeholder groups," and you can't immediately recall a specific, real-world meeting where you did exactly that, delete it.

AI is a brilliant tool for clarity, formatting, and structural polishing. But your career is yours alone. Keep your accomplishments grounded in reality, and you will walk into your next interview with the quiet confidence of someone who actually did the work.

(If you are using Google’s suite of models to edit your documents and encounter formatting or integration issues, you can find troubleshooting steps on the [Gemini Support Page](https://googlegemini-support.com).)