In recruitment, an AI-on-AI war is rewriting the hiring playbook

Roei Samuel, founder of networking platform Connectd, has been hiring at speed — 14 roles in six months. But he’s begun to wonder if candidates’ answers are genuine, even on video calls. “I can see their eyes shifting across the screen,” he says. “Then they come back with the perfect answer to a question.” The trust gap between employer and jobseeker is widening, and it’s fast becoming one of the trickiest knots in modern hiring.

From ChatGPT-polished CVs to full-blown applications submitted by bots, GenAI has hit the job market hard and gone fully mainstream. For a sizable generation of jobseekers — 68% of European tech workers were actively looking for a new role at the end of 2024 — it’s commonplace to use AI to tweak a CV or even complete an entire application.

Tools like Sonara, LazyApply, and JobCopilot have made it easy to shoot off dozens of applications in a day. In June, data from TestGorilla found that just over a third (37%) of UK jobseekers are using AI to complete applications. Among early-career candidates, it jumps to 60%, up from 38% the year before, according to Bright Network, which connects graduates and young professionals with recruiters.

Startups are at the forefront of this AI arms race. With smaller teams, shorter runways, and a culture of speed, they’re particularly exposed to this strange new world of suspiciously shiny applicants and AI-assisted code challenges. Most aren’t fighting it: 85% of employers now actively accept AI-assisted applications. But their acceptance doesn’t equal apathy. Amid a deluge of blatant AI use, how are Europe’s most agile companies working out who’s being real — and if they’re worth bringing onboard?

A new normal

Using AI to troubleshoot and tailor a CV has become par for the course. For most jobseekers, GenAI acts like a digital sidekick — smoothing grammar, sharpening phrasing, and cranking out tailored applications faster than ever. According to Canva’s January survey of 5,000 employees across countries including the UK, France, Spain, and Germany, 45% had used GenAI to build or improve a resume — and have yielded positive results. But hiring managers aren’t entirely sold. In the UK, 63% believe candidates should disclose if AI played a role in their application materials, signifying that trust is on shaky ground. 

Other research suggests that attitudes depend on the context. A global survey by Experis (part of workforce giant ManpowerGroup) found that 28% of tech leaders are fine with AI if used to personalise a resume or cover letter, 26% with help on problem-solving tests, and 24% even with answering interview questions. Just 15% said AI use is unacceptable across the entire job application process. 

For Duco van Lanschot, co-founder of fintech startup Duna, it’s all about the role. “If an engineer used ChatGPT to polish a written application, fine. That’s not the job. But for a growth or sales hire to use it very obviously is a big red flag,” he says. “The job itself involves public-facing comms and emailing stakeholders — and in a sea of generic, AI-generated copy, I want us to be as human as possible.”

Tech employers and startups are adapting — albeit in different ways and at different speeds. Some are setting ground rules for usage, some are bulking out human-only processes, and some are cutting away entire parts of the “traditional” hiring approach. “AI hasn’t broken hiring,” says Marija Marcenko, Head of Global Talent Acquisition at SaaS platform Semrush. “But it’s changed how we engage with candidates.”

Goodbye to CVs 

In the words of Khyati Sundaram, ethical AI hiring expert and CEO of Applied, we’re in the middle of “an AI-on-AI war.” And in the fallout, traditional application materials are losing their sway. In the tech sector, cover letters fell into obsolescence long ago, and CVs are next on the chopping block. “A huge upside is that it’s exposing résumés for what they are — a broken artefact,” says Sundaram, whose team works with the likes of Unicef UK, BLab, and the Equality and Human Rights Commission. “Putting résumés into keyword scanners or GenAI tools isn’t solving the problem for those hiring, because when it comes to the interview, the candidate falls apart,” she explains. 

Instead of cover letters and cut-and-paste CVs, employers are turning to structured questionnaires and skill-based tasks — tools that measure how someone thinks, not just how well they can write a prompt. “Skill-based hiring is no longer just a tech hiring thing,” Sundaram adds. “We’re seeing that crop up in more white-collar roles across the board.” According to TestGorilla, 77% of UK employers now use skills tests to evaluate candidates, with the same proportion saying these tests outperform CVs in predicting job success. This should have a positive effect in the long term: LinkedIn’s Economic Graph Institute found that a skills-based approach globally could expand talent pools by 6.1x, and help broaden gender and minority representation. 

At Semrush, the shakeup is already in full swing. Hiring managers are trained to sniff out fluency without depth, spotting signs of AI in real-time coding challenges or task-based interviews. “We’ve replaced the usual ‘Tell me about yourself’ prompts with in-depth interviews that explore experience, soft skills, and thinking patterns,” says Marcenko. “It’s hard to fake those, with or without AI.”

Applied’s own system uses a mix of automation and human insight. “We don’t believe in AI detectors — they’re rarely accurate, so we train reviewers to pattern match like an AI, comparing submissions to known GPT outputs,” explains Sundaram. “If five responses sound suspiciously identical, humans can flag them.”

Elsewhere, startups are getting more creative and more human. Alessandro Bonati, Chief People Officer at travel scaleup WeRoad, has ditched cover letters in favour of more creative, human-centric formats like curated portfolios or briefs of the “show, don’t tell” type. The company, which has over 210 staff in offices across Italy, Spain, Germany, and France, actively encourages candidates to use AI. “But that’s also accompanied by traditional in-person interviews to assess candidates’ thinking, communication, and cultural fit in real time,” says Bonati. His team also leans on real-time scenario-based exercises that reflect how candidates would collaborate, not just how well they can prepare.

The return of the reference and in-person interviews

Another ripple effect: references are back on the menu. Santiago Nestares, co-founder of the accounting startup DualEntry, is spending more time face-to-face on Zoom with candidates. “Experience is hard to fake,” says Nestares. “You can usually tell when someone’s just read about something versus having lived it.” He’s also going deeper on references; not just the usual ones, as those are always glowing, but backchannel conversations with people who’ve worked with them directly. “It’s so we can find out how someone handles pressure, works with a team, and shows up day to day,” says Nestares.

Through building the team for Connectd, a platform which enables angel investors and founders to effectively manage their startups, Samuel has noticed that candidates are cutting through the lack of trust by building more social proof around themselves. “For hiring managers, instead of taking a candidate’s word for it, we’re diving into references more than ever,” he says.   

The dreaded take-home task is on the way out now too. Unpaid and time-consuming, candidates have long despised them, and now that there’s the option of using GenAI to fake it (until they make it), employers are souring on them too. Live interviews, technical walk-throughs, scenario-based challenges, and even roleplay simulations are becoming the new standard, particularly in product, design, and marketing roles. “AI detectors are being used,” says Andreas Bundi, founder of Berlin-based HR consultancy Bundls. “But most companies are asking — why bother with take-homes when you can just do a live assessment?”

Bundi, who works with clients like Pitch, Cradle and Telli, says hybrid companies with mandatory office days are ditching take-homes to keep the interview process aligned with on-site work. With more candidates on the market, jobseekers are going further — relocating, or even flying in for interviews.

In the same vein, well-funded companies are increasingly comfortable bringing people on-site for tasks. “When travel isn’t possible, I’ve arranged in-person meetings with interviewers who happen to be nearby,” says Bundi. “I recently scheduled interviews around a conference both the candidate and interviewer were attending.” This “networking meets hiring approach” works surprisingly well.

Bundi says he sees that it’s the AI-first companies that are more relaxed about candidates using tools like ChatGPT — but it’s rarely made explicit. Recently, one of his Lead Data Scientists blew an interview by manually wrangling messy data instead of automating it. “They thought they needed to demonstrate their raw coding skills,” says Bundi. “But the company wanted strategy, not a human data janitor. That’s what ChatGPT is for.” As AI becomes standard, candidates and companies will need to get clearer on where it fits in the process. Until then, these frictions will continue.

Hire for skills and values, not a job description

Despite the ubiquity of AI-powered applications, most companies still haven’t formalised their approach. “It’s confusing for candidates, as some companies don’t want AI used in applications, even though those same roles involve AI tools every day,” says Sundaram. Indeed, 40% of employers using BrightNetwork services said they still haven’t set guidelines for AI usage in their processes, although 28% plan to for the next recruitment season. Of those who have set guidelines, 44% don’t allow candidates to use AI.

“The vanguard employers want everyone to use it and demonstrate their AI literacy,” says Sundaram. Some of Applied’s clients have even added the question: ‘How will you use AI in this job?’ She warns, however, that many of the quick fixes employers reach for — AI detectors, video screening with facial tracking, voice sentiment tools — raise huge ethical concerns. “If companies are tracking facial expressions for emotional nuance, it gets creepy,” she says. “Where do we draw the line?”

Instead, she argues, the fix will lie in redefining what candidates are being tested for. Applied has shifted from traditional job architecture to task architecture, evaluating not just skills, but values like resilience, adaptability, and mission alignment. “These are the human traits that will matter even more as jobs evolve,” she says. “Especially in startups, where everyone’s a generalist.”

Without doubt, generative AI is fundamentally reshaping the hiring process. The most forward-looking startups aren’t resisting the change — they’re building better processes around it. CVs may be broken, cover letters outdated, and applications increasingly synthetic — but the real differentiator is still very human. “We need people who can adapt, not just apply,” says Sundaram. “Because the job they’re hired for today may not exist in six months.” Startups that understand this and structure their hiring accordingly aren’t just future-proofing their teams, they’re rewriting the rules of work for the AI era.