Candidates Are Using AI to Beat Your Screening. Here's How to See Through It

The CV in front of you was probably touched by AI. Not faked, in most cases, but tightened, reworded, and quietly optimised to hit the exact language of your job advert. Candidates now have free tools that rewrite a CV to mirror a job spec in seconds, and a lot of them use them. That is not cheating. It is what anyone would do. But it does break something in the way most recruiters screen.

When everyone can match the keywords, keywords stop telling you who is actually good. The candidate who spent three years doing the work can look weaker on paper than the one who spent three minutes tailoring their wording. If your screening rewards the second person, you have a problem, and it is getting worse every month.

Why CVs stopped being reliable signals

For years, screening leaned on a simple assumption: the words on a CV roughly reflect the person behind it. AI writing tools have loosened that link. A candidate can now feed your advert into a tool and get back a CV that echoes your required skills, seniority language, and even your phrasing, whether or not the underlying experience is a strong match.

The honest ones use it to present real experience more clearly, which is fine. The weaker matches use it to paper over gaps. From the outside, on a keyword scan, the two can look identical. That is the crux of the issue: the tools that made CVs easy to optimise also made keyword screening easy to fool.

What keyword screening gets wrong

Traditional filters, including most ATS boolean searches, reward the presence of words. If the phrase is there, the candidate passes. If it is missing, they are cut, even if they have done exactly that work under a different label. This creates two failure modes at once:

• False positives. A keyword-tuned CV with thin real experience sails through because it says all the right things.
• False negatives. A genuinely strong candidate gets filtered out because they described their Oracle migration as "led the finance systems move" instead of using your exact keyword.

We walked through a real version of that second case in our sample shortlist, where a strong candidate was buried simply because of how they had worded their experience. Keyword screening would have missed them. That is the quiet cost: not just letting weak CVs through, but throwing away the people you actually wanted.

Read context, not keywords

The fix is not to detect AI or ban polished CVs. It is to screen for evidence instead of vocabulary. Good AI screening reads a CV the way an experienced recruiter does: it asks what the person actually did, how recently, for how long, and at what level, rather than whether a magic word appears.

That approach is far harder to game, because you cannot fake depth and specificity as easily as you can insert a keyword. A tool that weighs real experience will not be impressed by a CV that simply repeats your advert back to you, and it will still surface the quietly strong candidate who did the work but described it plainly.

How to see through an AI-polished CV

Whether you screen manually or with software, the same instincts apply:

• Chase specifics, not adjectives. "Transformational leader" tells you nothing. "Cut time-to-hire from 45 to 28 days across three offices" tells you plenty.
• Be suspicious of perfect alignment. A CV that mirrors your advert word for word deserves a second look, not an automatic pass.
• Look for outcomes and timelines. Real experience leaves a trail of dates, scope, and results. Padding rarely does.
• Screen against real criteria. Use a tool that scores candidates on defined role requirements and explains its reasoning, so a keyword-stuffed CV cannot quietly outrank a stronger one.

Where Lucuma fits

Lucuma reads context, not keywords. It scores each candidate against the criteria you set and gives a plain-English reason for every score, so a polished-but-thin CV does not slip through and a plainly-written strong one does not get lost. When a CV lacks the detail to judge fairly, it flags it for review rather than guessing. Because every score is explained, you can also stand behind your decisions, which matters under UK rules on automated hiring, covered in our guide to ICO-compliant AI screening, and it is the difference we dig into in black box vs glass box screening.

You can see a real anonymised shortlist to watch it separate substance from wording, or read what to look for in AI CV screening software before you choose a tool.