Your ATS Already Has AI Screening. Here's Why It Still Misses Your Best Candidates.
· 6 min read
Bullhorn, Vincere and Greenhouse all ship AI matching now. This is the honest difference between keyword matching in your ATS and a screener built on your own criteria.
Bullhorn, Vincere, JobAdder and Greenhouse all ship AI matching now. So a fair question before you pay anyone to build a screener is: why not just use what our ATS already gives us? Here is the honest answer, including when the ATS is genuinely enough.
What your ATS AI actually does
The AI built into most applicant tracking systems is a matching engine. It compares the words in a CV to the words in a job description and returns a similarity score. It is fast, it is included, and for some roles it is fine.
The catch is in how it decides. Keyword matching rewards CVs that sound like the job description. It cannot tell the difference between someone who did the work and someone who listed the right words. And it does not know anything about your desk.
The three things it cannot learn
Your ATS scores against the job description. It does not score against you. So it cannot learn:
- Your placement history. That your best hires for this client all came from two competitors, and CVs from those firms deserve a second look — the ATS has no idea.
- Your unwritten rules. That your team quietly passes on anyone who job-hopped three times in two years for this kind of role. The ATS will happily rank them first.
- Your reasoning. A keyword score is a number with nothing behind it. When a recruiter asks "why is this person ranked third?", there is no answer to give.
A custom screener is built on your actual criteria and your history, and every score comes with a note explaining the placement. That note is the difference between a tool your recruiters use and a tool they quietly ignore.
Keyword match vs. criteria-based screening
| ATS AI matching | Criteria-based screener | |
|---|---|---|
| Scores against | The job description text | Your rubric + placement history |
| Handles unwritten rules | No | Yes, once captured |
| Explains each score | Rarely | Every score, in writing |
| Setup | Already on | Built once, tuned to you |
| Cost | Included | A build + optional tuning |
When the ATS is enough
We would rather tell you this than sell around it. If you run a low volume of straightforward roles, your criteria really are just "these keywords," and no one on your team can name an unwritten rule they screen on — the ATS matching is probably fine. Adding a custom layer would be effort for little gain.
The case flips when your roles are nuanced, your volume is high, or your best recruiters screen on judgment they have never written down. That judgment is your edge. A keyword engine cannot see it; a screener built on your rubric can. If you are not sure which side of the line you are on, the free audit exists to answer exactly that.
The bottom line
Your ATS AI is not bad — it is generic. It scores every firm the same way, because it does not know your firm. The value of a custom screener is not that it is smarter in the abstract; it is that it is yours: your criteria, your history, your reasons, on every candidate.
New to the topic? Start with how AI CV screening cuts shortlisting time, then come back here. Or see the packages on the services page.