Every CV looks polished now. How to find the real candidates in an AI application flood

I keep hearing the same sentence from recruiters: “We have more applicants than ever, and fewer people we actually want to talk to.” It sounds like a contradiction. It isn’t. It is what happens when applying for a job stops costing the applicant anything.

Why the inbox is overflowing

In a July 2026 interview with Fortune, the CEO of the hiring platform Greenhouse said that around 254 people apply for every job ad on its platform, and that applications per recruiter have risen by 412%. He described tools costing about $20 that use AI to apply to every job a candidate can find.

On the other side of the table, Gartner predicts that by 2028, one in four candidate profiles worldwide could be fake. In the same research, reported by HR Dive, 6% of the candidates Gartner surveyed admitted to interview fraud, such as posing as someone else.

So the problem is not a shortage of applications. It is a shortage of signal. Each CV has been polished by the same tools, against the same job description, using the same vocabulary. When everyone is a “results-driven team player with strong stakeholder management”, nobody is.

Why keyword screening makes it worse

Most applicant tracking systems still filter by keywords. That made sense when a CV was written by a person, for a person. Today a keyword filter mostly measures how well a candidate’s AI tool copied your job ad. The candidate who pasted your requirements into a chatbot ranks first. The experienced claims handler who described her work in her own words ranks somewhere on page nine.

I wrote more about this in why keyword filters miss your best candidates. The short version: matching words is not the same as matching experience.

What to screen for instead

You cannot stop candidates using AI, and you shouldn’t want to. Many use it for the same reason you do: to save time on writing. What you can do is change what you look for. These checks work whatever tools you use:

• Specifics over adjectives. Products handled, team sizes, systems used, regulators dealt with, numbers achieved. Generated text is fluent but vague. Real experience is specific.
• Consistency across the record. Do dates, titles and responsibilities line up? Does the experience described match the seniority claimed?
• Evidence for the requirements that matter. Decide your three or four must-haves before you open the first CV, and look for evidence of each one, not for the words.
• A follow-up question. The fastest way to tell substance from polish is to ask one question the CV doesn’t answer: “Tell me about a difficult case you handled last month.” A tool can write a CV. It is much harder to fake a conversation about your actual work.

How we approach it at Talenda

We built Talenda around those checks, because they are what experienced recruiters already do when they have time. My co-founder Jenny Chow spent more than ten years screening candidates for frontline banking and insurance roles. Her rule is simple: a CV is only the starting point.

Here is what that looks like in the product:

• Every CV becomes a structured profile. Talenda extracts experience, skills, education and other details into the same fields for every candidate. Comparing two people stops being a matter of reading two different writing styles.
• Matches come with reasons. When Talenda suggests candidates for a role, it shows why: the relevant experience it found against your requirements. Recruiters review those reasons rather than trusting a score.
• The follow-up question happens early. Candidates can complete an AI voice screening interview in English, Cantonese or Mandarin through an invitation link. Your team reviews the transcript and summary before deciding who to speak to. (More on this in the first-round phone screen.)
• People make the decision. Talenda does not reject or advance candidates on its own. AI does the groundwork; your recruiters choose.

A simple process for your next high-volume role

1. Write down the four requirements a successful hire must meet, and the evidence that would prove each one.
2. Structure every application the same way before anyone reads it.
3. Shortlist on evidence, and record the reason for each shortlist decision.
4. Ask every shortlisted candidate the same two or three follow-up questions, in the language they work in.
5. Spend your interview time on the people whose answers were specific.

The flood of applications is not going away. But the answer to AI-written applications is not more keyword filters. It is getting back to the questions that only a real candidate can answer, and giving your team the time to ask them.