AI-generated cover letters are now the norm, not the exception: a StepStone study of 704 recruiters and 3,495 applicants found 61% of candidates already use AI to help write them. The recruiter's job has shifted from spotting AI to judging substance. This guide gives you a fair, DACH-compliant framework to evaluate them — plus what the EU AI Act and your works council mean for your process.

How common are AI-assisted cover letters right now?

Common enough that treating an AI-assisted application as a red flag by default would mean rejecting most of your pipeline. In the same StepStone study, 61% of applicants reported using AI to edit their cover letters, and recruiters had mixed feelings about the result. 74% appreciated the more professional presentation, but 73% said AI-assisted applications feel less authentic, 69% found them less individually tailored, and 75% were bothered by exaggerated qualifications.

The honest takeaway: AI has raised the floor on polish and lowered the ceiling on individuality. Roughly 80% of recruiters in that study rated the overall quality of incoming applications as only medium or poor. So the real problem is not "a candidate used AI." It is "AI-generated text makes it harder to see the actual person behind the application." Your evaluation has to adjust to that — not fight the tool.

What AI-generated cover letters typically look like

You do not need a detector to recognise the pattern. AI-written cover letters tend to share a recognisable shape, especially when the candidate pasted a job ad into a generic tool and copied the output with little editing. Watch for these tells:

Two caveats before you act on any of this. First, none of these tells prove AI use — a nervous first-time applicant can produce the same generic letter unaided. Second, and more important, a polished AI-assisted letter can still sit on top of a genuinely strong candidate. Tells tell you where to look harder, not whom to reject.

Can — and should — you reject a candidate for using AI?

Some recruiters do. In a TopResume survey of 600 US hiring managers, 19.6% said they would reject a candidate outright for an AI-generated résumé or cover letter, 20% treated heavy reliance on AI as a red flag, and 14.5% believed candidates should not use AI at any stage. The same survey found 33.5% were confident they could spot an AI-generated document in under 20 seconds.

That confidence is the trap. Current AI-text detectors are not reliable enough to serve as evidence of authorship, especially on hybrid text that a human has edited. They also produce systematically higher false-positive rates for non-native English speakers and for neurodivergent writing styles — a real fairness and discrimination risk if a rejection hinges on a detector score. Treat any detector output as a weak hint at most, never as proof, and never document it as a rejection reason.

The defensible position: do not reject for the tool. Reject — or advance — based on substance. A candidate who used AI to structure a clear, specific, honest letter has demonstrated exactly the kind of judgement most roles reward. A candidate whose AI-polished letter collapses under one specific follow-up question has told you something real. Focus your evaluation on the skills and evidence behind the words, not the mechanism that produced them.

A 3-question evaluation framework for AI-written applications

Instead of asking "did AI write this?", ask three questions that separate substance from surface. This framework works whether or not AI was involved, which is exactly the point — it is AI-proof because it evaluates the candidate, not the writing tool.

Question three is where AI-inflated applications fail, because a candidate can generate an impressive letter in seconds but cannot generate lived experience on demand in an interview. Build your process so that written claims always get tested in conversation. That single habit neutralises most of the risk AI introduces.

The DACH angle: AI-assisted screening and your works council

Here is what the generic candidate-advice articles miss entirely, and where DACH recruiters need to be careful. The moment you use AI to help evaluate or rank applications — not the candidate writing them — German co-determination law can apply.

Under § 95 of the German Works Constitution Act (BetrVG), guidelines for personnel selection — the criteria you use for hiring, transfers, regrading and dismissals — require the works council's consent where one exists. Crucially, § 95 Abs. 2a, added by the 2021 Works Council Modernisation Act, explicitly extends this co-determination right to cases where AI is used in establishing those selection guidelines. In plain terms: if you introduce an AI tool that shapes how candidates are screened or ranked, that is not purely an HR decision. Your works council has a formal say.

The practical implications for DACH teams:

If you want a deeper checklist covering GDPR and works-council requirements for HR software specifically, our DACH talent-management software guide lays that out alongside vendor criteria.

The EU AI Act and recruitment AI: what actually triggers it

The second DACH/EU-specific point competitors ignore: recruitment AI is regulated. Under Annex III of the EU AI Act, AI systems used for recruitment and selection — including targeted job advertising, filtering applications, and evaluating or ranking candidates — are classified as high-risk. Systems that decide on promotions, terminations or task allocation based on behaviour fall under the same heading.

High-risk classification does not ban these tools. It attaches obligations — risk management, data governance, transparency, human oversight and record-keeping — with requirements phasing in for high-risk systems through 2026 and 2027. Penalties for the most serious breaches can reach €15,000,000 or 3% of global annual turnover, so this is not a footnote for enterprise HR.

What this means in practice: a candidate using ChatGPT to write a cover letter is not your compliance problem. An AI system you deploy to screen or rank those letters is. Before adopting any AI screening capability, confirm the vendor can support the human-oversight, transparency and documentation obligations the high-risk regime requires — and that your works-council process (above) is aligned with it.

A structured interview workflow to verify what's real

Because you cannot and should not rely on detection, verification moves into the interview. The goal is simple: test whether the confident written narrative holds up in a live, unscripted conversation. Use a consistent, structured sequence so every candidate is measured on the same axes.

Notice that this workflow makes the cover letter's authorship irrelevant. It does not matter whether AI helped write the application if the interview verifies the substance behind it. That is the reframe experienced recruiters land on: stop policing the tool, start structuring the verification.

Where a structured AI co-worker fits

There is a fair, useful role for AI on the recruiter's side of the table — not to detect candidate AI, but to make evaluation more consistent, auditable and bias-resistant. That is the thinking behind Atlas, sprad's AI co-worker for hiring: it supports structured, criteria-based screening where every candidate is assessed on the same defined axes and the reasoning is recorded.

Used well, that directly serves the two DACH concerns above. Consistent, documented, criteria-based evaluation is exactly what makes a selection process defensible to a works council under § 95 BetrVG, and it maps onto the human-oversight and record-keeping expectations of the EU AI Act's high-risk regime. The point is not to remove human judgement — it is to give every applicant the same fair, transparent read, and to leave an audit trail when someone asks how a decision was made. AI on the candidate side raised the noise; structure on the recruiter side is how you cut back through it.

Frequently asked questions

Can recruiters tell if a cover letter was written by AI?

Not reliably. Experienced recruiters recognise typical AI patterns — generic openings, buzzwords, no company specifics — but these are hints, not proof. AI-text detectors are not accurate enough to establish authorship on edited text and carry a real bias risk against non-native speakers and neurodivergent writers, so they should never be the basis of a rejection.

Should I reject a candidate for using AI to write their cover letter?

Not for the tool alone. Some hiring managers do — around 20% in a TopResume survey — but with 61% of applicants now using AI, blanket rejection means discarding much of your pipeline. Evaluate substance instead: specific evidence, fit to the exact role, and consistency between the letter, CV and interview.

Do works councils have a say when AI is used in hiring?

In Germany, yes. Under § 95 BetrVG, personnel-selection guidelines require works-council consent, and § 95 Abs. 2a explicitly extends that co-determination right to cases where AI is used in setting those selection criteria. Involve the works council before deploying any AI-assisted screening tool.

Is recruiting AI high-risk under the EU AI Act?

Yes. Annex III of the EU AI Act classifies AI used for recruitment, application filtering and candidate evaluation or ranking as high-risk, which attaches obligations around transparency, human oversight and documentation. It does not ban such tools, but it does regulate how you deploy them.

Next step

Stop asking whether a cover letter was written by AI — that question is already unwinnable and, in DACH, legally beside the point. Build your process around the three evaluation questions and the structured interview workflow above, involve your works council early, and make sure any AI you use on the recruiter side is transparent and documented. That is how you stay fair to candidates and compliant at the same time.