AI is becoming a bigger part of recruitment.
Candidates are using tools like ChatGPT to write and polish their CVs. Employers are using AI tools to screen applications, rank candidates, and speed up hiring decisions.
On the surface, this sounds efficient.
But new research suggests there may be a hidden problem.
AI screening tools may not simply be identifying the strongest candidates. In some cases, they may be favouring CVs that sound like they were written by AI.
What the research found
A recent paper, AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights, studied how large language models evaluate CVs and resumes.
The researchers used 2,245 human-written resumes and created AI-generated versions using several major language models, including GPT-4o, GPT-4-turbo, LLaMA, Mistral, Qwen and DeepSeek.
They then asked AI models to compare different versions of the same candidate’s resume.
The key point is important:
The candidate’s experience, skills and background stayed the same.
Only the wording changed.
The research found that AI models often preferred resumes written by themselves over human-written versions of the same candidate profile. In some cases, larger models showed self-preference bias of around 67% to 82%, even after controlling for content quality. The researchers also simulated hiring pipelines and found that candidates using the same AI model as the evaluator could be 23% to 60% more likely to be shortlisted than equally qualified candidates using human-written resumes.
That should get the attention of any hiring team using AI in recruitment.
Why this matters for employers
The risk is not just that AI makes mistakes.
The bigger issue is that AI may reward the wrong signals.
A candidate with a polished, AI-written CV may appear stronger than another candidate with the same or better experience, simply because their CV is written in a style the screening tool prefers.
That creates a serious problem.
Companies may think they are improving hiring efficiency, but they may actually be filtering for:
✅ Better CV formatting
✅ Better AI-style writing
✅ Better keyword alignment
✅ Better use of generative AI tools
Not necessarily better capability.
For specialist roles, especially in software engineering, data, cloud, cybersecurity and product, that distinction matters.
The best person for the job is not always the person with the most polished CV.
The Thailand hiring angle
This is especially relevant in Thailand and Southeast Asia.
Many strong technical candidates are highly capable but may not write perfect English CVs. Some may be excellent engineers, developers, infrastructure specialists or security professionals, but their CVs can be too brief, too modest, badly structured, or not written in the style an AI screening tool expects.
That does not mean they are weak candidates.
It means their CV may not fully reflect their ability.
In Thailand, this matters because many regional and international companies already require candidates who can work in English. If AI screening tools place too much weight on polished English presentation, employers may accidentally overlook strong local talent.
The result?
Good candidates get missed.
Hiring managers see fewer relevant profiles.
Companies assume there is a talent shortage.
But sometimes the issue is not the talent market.
It is the screening process.
What this means for candidates
For candidates, the takeaway is not that you should fake your experience or over-optimise your CV for AI.
That is a bad idea.
But it does mean your CV needs to be clear, structured and easy to understand.
A strong CV should show:
✅ What you have worked on
✅ What technologies you have used
✅ What problems you helped solve
✅ What impact you had
✅ What level of responsibility you carried
AI tools can help improve clarity, but they should not replace authenticity.
The best CVs are not generic AI-written documents. They are accurate, specific and easy for both humans and systems to understand.
What this means for companies hiring tech talent
For employers, the lesson is simple:
AI can support recruitment, but it should not replace judgement.
Used properly, AI can help organise information, summarise CVs, identify keywords and reduce admin. That can be useful.
But relying too heavily on automated screening creates risks.
You may filter out candidates who are technically strong but weaker at CV writing. You may overvalue candidates who are better at presenting themselves than actually doing the job. You may also create a process where every CV starts to look and sound the same.
That is not better hiring.
It is just faster filtering.
Why recruiters still matter
This is where a good recruiter still adds value.
A recruiter should not just forward CVs.
They should understand the role, the market, the candidate’s real experience, and the context behind the profile.
That includes knowing when a CV undersells someone.
It also means knowing when a polished CV is not backed up by real depth.
In specialist tech recruitment, that judgement matters.
A good recruiter can help companies look beyond the document and understand the person behind it.
That is especially important in competitive markets like Thailand, where strong candidates are often passive, cautious, and not always easy to assess through a CV alone.
Practical recommendations for hiring teams
Companies using AI in recruitment should be careful.
AI screening tools can be useful, but they need guardrails.
A few sensible steps:
✅ Use AI to support screening, not make final decisions
✅ Review rejected candidates periodically to check for false negatives
✅ Avoid over-relying on CV summaries or keyword matches
✅ Compare candidates against role requirements, not writing style
✅ Make sure human reviewers are involved before ruling out potentially relevant candidates
✅ Work with recruiters who understand the local talent market
The goal should not be to remove humans from hiring.
The goal should be to help humans make better decisions.
Final thought
AI is changing recruitment.
Some of that change is positive. It can reduce admin, speed up processes and help hiring teams manage large volumes of applications.
But faster does not always mean better.
If AI screening tools favour AI-written CVs, companies may end up rewarding presentation over substance.
For employers hiring tech talent, the key question is not just:
“Can AI help us screen faster?”
It should be:
“Are we still identifying the right people?”
Because the best candidate is not always the best CV writer.
And increasingly, they may not be the best AI prompt user either.