You asked a chatbot for help, it produced something that reads like a brochure, and you can tell it is not right without being able to say exactly why. That instinct is correct. An AI draft is a reasonable starting point and a poor finished article, and humanising an AI generated CV is a specific editing job with a clear method.
Here is a five-pass framework. Work through them in order; each pass is quick and does one thing.
Why AI Drafts Read the Way They Do
A language model producing a CV has no access to your actual work. It has the shape of a strong CV and the statistically likely vocabulary of professional achievement. So it generates confident, fluent, general text, grand verbs, round numbers, no constraints, no setbacks, uniform tone throughout.
The result has three characteristic faults: it is abstract where it should be specific, it is grandiose where it should be plain, and it is American where it should be British. All three are fixable in about forty minutes.
Pass 1: Delete the Clichés
Certain phrases now function as a signature of generated text. Search for each and remove it.
Verbs to replace:
- spearheaded to led
- orchestrated to ran, coordinated
- used to used
- utilised to used
- championed to pushed for, argued for
- pioneered to introduced, was first to
- facilitated to ran, chaired, helped
- drove to led, delivered
The plain verb is always stronger. "Led the migration" is more credible than "spearheaded the migration", because nobody says spearheaded out loud.
Phrases to delete outright: evidence of my dedication; proven track record; results-driven professional; passionate about delivering excellence; dynamic and highly motivated; wealth of experience; hit the ground running; think outside the box; go-getter; team player with excellent communication skills.
None of these survive contact with a reader. Every one is a self-assessment that cannot be checked, and they occupy the space where evidence should be.
Pass 2: Anglicise It
Models default to American English and American workplace conventions. On a UK application this reads as a CV written for somewhere else.
- Spelling: organised not organized, specialise not specialize, programme (initiative) not program, analyse not analyze, centre not center, licence (noun) not license, fulfil not fulfill, catalogue not catalog.
- Words: CV not resume, sick leave or annual leave not PTO, secondary school not high school, degree classification not GPA, mobile not cell, autumn not fall.
- Register: British professional writing is more understated. Trim superlatives. "Reduced processing time by 30%" lands harder than "achieved an outstanding 30% reduction".
- Dates: use the UK convention consistently, Mar 2022, Sep 2024.
- Currency and units: pounds, metric, UK date order.
Also strip anything American models add by habit: objective statements, references with contact details, photographs, marital status, and "References available upon request".
Pass 3: Replace Abstraction with Specifics
The most important pass. This is what converts a generated draft into an actual CV.
Go through each bullet and ask: could this sentence have been written by someone who had never done my job? If yes, it needs your detail.
Add, wherever honest:
- The real number. Not "significantly improved" but "cut it from 9 days to 4". Not "a large team" but "a team of 14".
- The specific system, place or process. The actual tool, the actual department, the actual client type.
- The incidental detail nobody invents. The version that caused the problem, the legacy system you had to work around, the seasonal peak that broke the old approach.
Then check the numbers the model produced. Generated figures cluster on tidy values, 30%, 50%, doubled, £200,000, because those look like strong results. Any figure you did not supply yourself must be deleted or replaced with the real one. Carrying an invented number into an interview is the single worst outcome of using AI on a CV.
Pass 4: Put the Difficulty Back
Models remove friction. Every project in a generated CV proceeds smoothly to a successful conclusion, which is not how work happens and not how experienced readers expect it to be described.
For each of your two or three most important examples, restore one honest complication: the constraint you worked within, the approach you abandoned, the thing that broke first, the resistance you had to work through.
This feels like admitting weakness. It is the opposite. Constraints and setbacks are close to impossible to fabricate convincingly, so they read as true, and interviewers reliably pick them as the thing they want to discuss. It is the highest-value sentence in your CV.
Pass 5: Fix Attribution and Read It Aloud
Search for "we". Generated text drifts into collective language, which erases your personal contribution, and evidence-based screening is specifically built to distinguish what you did from what your team did. Name your part honestly: "I owned the data migration within a five-person replatforming team" is both accurate and far stronger than "we replatformed the CRM".
Then read the whole thing aloud. You will hear the remaining artificial sentences immediately. They are the ones you stumble over or feel faintly embarrassed saying. Those are the sentences an interviewer will also notice.
The test to apply: could I talk for two minutes about every claim on this page? If not, that claim is either not yours or not yet specific enough to be useful.
Frequently Asked Questions
How Can I Tell if My CV Sounds AI-generated?
Look for grand verbs (spearheaded, used, orchestrated), abstract achievement language with no numbers, uniform tone across your whole history, tidy round percentages, and no mention of anything that was difficult or went wrong. Reading it aloud surfaces all of it quickly.
Will a Recruiter Reject Me for Using AI?
Not for using it. AI detection tools are unreliable and unfairly flag non-native English writers, so responsible employers do not screen on them. What costs you is the result, fluent text with no evidence behind it, which reads as telling them nothing.
Should I Rewrite an AI Draft from Scratch?
No need. The structure is usually serviceable. Run the five passes: clichés out, anglicise, specifics in, difficulty restored, attribution fixed. Forty minutes of editing beats starting again.
Is It Fine to Keep Some AI Phrasing?
Yes, where it is genuinely clear and plain. The goal is not to remove every trace of assistance. It is to ensure every claim is accurate, specific and yours. Clear writing is clear writing whoever tightened it.
What Is the Single Most Important Pass?
Pass three. Replacing abstraction with real specifics is what converts a well-written document into a CV that scores well and earns an interview. Everything else is tidying by comparison.
The Principle
Use the model for expression and supply the substance yourself. The passes above are really one instruction repeated: put back everything specific, difficult and personal that a system with no knowledge of your life could never have known.
CVSense scores the context, scale and outcome behind each claim rather than its polish, which is exactly why an edited draft full of real detail outperforms a pristine one full of confident generalities.
Sources
National Careers Service. CV Sections and Advice.
https://nationalcareers.service.gov.uk/careers-advice/cv-sections
Chartered Institute of Personnel and Development.
https://www.cipd.org/uk/
InsightCircle
