How to read an AI match score

If you treat 84% like a hire, you’ll get burned

Models flatten people. They also miss closed-source work, weird titles, and “I led the migration” that isn’t in a skills list. Fine. That’s why you still have a job.

Use the score to sort. Use the write-up to think. If those two disagree with what you see on the page, trust the page-and maybe the JD was vague.

What Insayab is trying to say

A match is against your job description, not “are they a good engineer.” Different req, different number. That’s the feature.

You get a score, strengths, and gaps. The gaps are the useful part. That’s where you either pass, or you go into the call with one sharp question instead of a tour of their life story.

A simple way to use it

Don’t build a religion around bands. Roughly:

• High + gaps you can live with: reach out
• Middle + gaps that are the job: skip or save for a different req
• Low + a LinkedIn you still like: read why. Sometimes the JD is buzzword soup

Two platforms, two numbers

LinkedIn 80 and GitHub 40 isn’t “60.” It’s a story. Maybe they don’t publish code. Maybe the GitHub is a student account they forgot. Look, then decide.

If you only have one platform, one number is enough to start. Don’t wait for a perfect file.

Common questions

Should I only message 80+?

No. That’s how you miss people with ugly titles and real work. Set a floor if you must, then read the gaps on anything near it.

Can the score be wrong?

Yes. Garbage JD in, garbage match out. If the description is a keyword pile, fix that before you blame the model.

Is this replacing me?

It’s replacing the ten minutes you spend squinting and still guessing. You’re still the one who sends the message.