As employers race to automate hiring, a troubling pattern is starting to emerge: AI-written CVs may be performing better with AI screening tools than applications written without machine help. That raises an uncomfortable question for recruiters and job seekers alike — are artificial intelligence systems beginning to reward signs of AI use rather than genuine candidate quality?
A recent hiring experiment at international law firm Addleshaw Goddard has put this issue into sharper focus. Faced with a surge in applications for an IT role in HR, the firm tested an internal AI agent against human reviewers. What it found adds to growing concerns about bias in automated recruitment and the unintended consequences of AI-assisted job applications.
Why AI-written CVs are getting more attention
Recruiters have long used technology to sort résumés, but generative AI has changed the scale of the challenge. Tasks that once took candidates hours — tailoring a cover letter, aligning a résumé with a job description, highlighting key skills — can now be done in minutes. The result is a flood of polished applications.
In Addleshaw Goddard’s case, 178 applications arrived for a single role, far more than would have been typical a year earlier. To speed up screening, the HR team used an AI agent built through Microsoft Copilot, giving it the job requirements and desired technical experience.
Both human reviewers and the AI created shortlists. But the results did not align. The AI leaned toward especially slick applications featuring unusually impressive accomplishments, rapid career impact, and ambitious project claims despite limited experience. To human reviewers, some of these applications appeared too polished to be fully credible.
That disconnect highlights a key concern: AI-written CVs may not simply be easier to read — they may be receiving favorable treatment from AI systems.
How AI hiring bias may work
One explanation is relatively benign. If a role values AI literacy, an AI system might interpret a highly optimized application as a signal that the candidate knows how to use modern tools. But the more worrying possibility is what researchers call self-preferencing: AI models showing a preference for content created with AI, especially when the same or similar model is used on both sides of the hiring process.
A working paper cited in the source reporting examined simulated recruitment across 24 occupations, including consultants, engineers, chefs, and teachers. It found that candidates using the same large language model as the hiring tool were significantly more likely to be shortlisted than candidates relying on entirely human-produced applications.
That suggests the advantage may not come solely from better formatting or clearer writing. Instead, the model may be responding more positively to familiar language patterns, structure, phrasing, or optimisation choices commonly produced by AI.
The risks for employers
If AI hiring tools favor AI-written CVs, employers could face several problems:
- False positives: Highly polished applications may exaggerate competence.
- Missed talent: Strong candidates who write plainly or avoid AI may be overlooked.
- Bias amplification: A model may reward stylistic similarity over substance.
- Reduced trust: Candidates may feel forced into an AI arms race just to compete.
For organizations trying to improve efficiency, that creates a serious tension between speed and fairness.
What this means for job seekers using AI-written CVs
For candidates, the message is mixed. On one hand, AI-written CVs can help applicants communicate experience more clearly, match keywords in a job posting, and avoid weak formatting. On the other hand, overreliance on AI can create generic, inflated, or misleading applications that fall apart under human scrutiny.
Job seekers should think of AI as an editor, not a substitute for truth. A strong application still needs to be accurate, specific, and grounded in real achievements.
Best practices for candidates
- Use AI to improve clarity, not invent accomplishments.
- Customize your CV with concrete examples and measurable results.
- Keep language natural enough to reflect your real communication style.
- Review for exaggeration, vague buzzwords, and unsupported claims.
- Prepare to explain every line in an interview.
In short, AI-written CVs can be useful, but only when paired with honesty and careful editing.
How companies can reduce AI recruitment bias
The most practical takeaway for employers is that AI should support hiring decisions, not replace judgment. Researchers referenced in the source material suggest two promising ways to reduce this kind of bias:
- Instruct models to ignore origin signals and focus strictly on job-relevant content.
- Use multiple assessment models instead of relying on a single AI tool.
Companies can also strengthen their recruitment process by:
- Keeping humans involved in final shortlisting and hiring decisions
- Auditing AI outputs for patterns of favoritism or inconsistency
- Testing screening tools with varied résumé styles and formats
- Evaluating evidence of skills rather than presentation alone
The Addleshaw Goddard example offers a useful reminder. Despite the AI’s rankings, the successful hire was ultimately chosen by human reviewers — and it was not the candidate the AI preferred most.
The future of AI-written CVs in hiring
The rise of AI-written CVs is part of a broader transformation in recruitment, where both applicants and employers are using automation to gain an edge. But when AI starts favoring AI, the process risks becoming less about merit and more about machine compatibility.
The smart path forward is not to reject AI outright. It is to use it carefully, transparently, and with safeguards. AI-written CVs will likely remain a fixture of modern job hunting, but employers must ensure their screening tools reward genuine capability rather than algorithmic polish. In recruitment, human judgment still matters most.
