The easiest thing to read on any CV is the skills list. It's also the least reliable.

Java, Spring Boot, Kafka, Docker, Kubernetes, AWS. Six words, ten seconds, nothing to verify. A candidate only has to remember the name of a technology to put it there.

The harder part sits right underneath: the job descriptions. That's where the dates live, the scale, the constraints, and the things people reveal without meaning to.

The problem is you've got 80 CVs this week and no time to read any of them properly.

So how do you tell, in five minutes, whether a skill is real experience or just a word on a page?

Why every CV reads like a good one

A skills list is a claim with no receipt

Everything else on a CV is attached to a date. The skills list isn't.

It doesn't tell you how long, on which project, or in what role.

It tells you one thing only: the candidate knows the word.

The skills list is the claim. The job history is the receipt.

Every CV now wears the same suit

Same templates. Same polished phrasing. Strong verbs on every single line.

• Optimized the system.
• Designed the architecture.
• Owned the pipeline.
• Improved performance.

Read ten CVs and all ten sound like a tech lead.

You're matching keywords, not reading context

Fast. Also wrong in both directions.

The word Kafka is there, but the actual work was editing one config file. The words Spring Boot are missing, but the candidate shipped three services to production.

Keyword screening is inspecting the shipment by reading the label.

The familiar one: 18 months, 12 technologies

A Data Engineer CV: Kafka, Spark, Airflow, dbt, Snowflake, Kubernetes, Terraform, Redis.

Experience: 18 months, one company, one project.

This person isn't lying. But twelve technologies in eighteen months means each one got a few weeks — and you have no idea which few.

80 CVs a week = normal.

3 minutes per CV = also normal.

You = the one who explains a bad shortlist.

The Named–Used–Owned ladder for screening CVs

You don't need to read more carefully. You need to read in a different order.

Every time a skill shows up, ask one question: which rung is it standing on?

Rung 1 — Named

The skill appears. No action attached to it.

The tell: "Project stack included Kafka, Redis, Elasticsearch."

That sentence describes the project. Not the person.

Rung 1 isn't bad. It just hasn't told you anything yet.

Rung 2 — Used

There's a verb, and there's a scope.

The tell: "Built the consumer that pushed order events into the warehouse."

Now you can believe the basics. What's still missing is scale and constraint.

Rung 3 — Owned

There's a decision, a trade-off, and a consequence for getting it wrong.

The tells:

• A real number
• Why A instead of B
• An incident they handled
• Something they maintained after shipping

Example: "Split the consumer group to cut lag from 40s to 3s during the sale."

That's experience. Everything below it is information.

Running the ladder in five minutes

Four moves, no second pass:

• Minute 1: read the timeline
• Minute 2: pick 3 must-have skills
• Minutes 3–4: assign a rung to each
• Minute 5: write questions for rungs 1 and 2

Minute 1 matters more than it looks. If no role lasted long enough to learn the skill, every line above it needs a question.

The output isn't a verdict. It's three questions for the phone screen.

And here's the part that matters most: a gap between rung 1 and rung 3 usually isn't dishonesty. Plenty of strong engineers write terrible CVs.

You're placing rungs to find out what to ask, not who to cut.

The Knoot Angle

Reading fast isn't reading less. It's knowing where to stop.

But when the pile is 300 CVs deep, even a five-minute routine stops fitting into the day. That's where Knoot's AI Screening comes in.

It doesn't read CVs in the abstract. It compares each profile against criteria pulled from your own JD, then marks the spots worth a second look: a skill claimed but never mentioned inside any role, timelines that overlap, a seniority claim the dates can't support. Every shortlisted candidate arrives with the reasoning attached.

Nobody gets rejected automatically. A flag is only a flag. It tells you what to ask in the first fifteen-minute call.

AI finds the gap. You're still the one who asks about it.