How many screening questions should a role have?

Four to six. Each answer runs one to three minutes, which keeps the whole screening under fifteen minutes and still gives the summary enough material to work with.

What makes a good AI candidate screening question?

One that asks for a concrete project and the candidate's own part in it. Questions that ask for opinions or self-assessment produce answers an AI summary cannot ground in evidence.

Should questions include sub-points?

Yes. A short lead sentence plus two or three bullet sub-points guides the candidate through everything you want covered, and the screening flags anything a candidate did not address.

Can candidates type instead of recording?

Yes. Every question can be answered by voice or in writing, and candidates can edit or re-record any answer before submitting.

Recruiter guide

AI candidate screening: questions and templates that produce a usable summary

An AI screening summary is only as good as the questions behind it. These are the question patterns we see work best in VoltFlows — copy them into a role as they are, or adapt the wording to your client.

What separates a good screening question from a weak one

Do

Anchor every question in a real project: "the project you spent the most time on".
Ask for the candidate's own contribution separately from the team's.
Ask them to name tools and technologies and say where each was used.
Use two or three bullet sub-points per question so nothing is missed.

Avoid

Do not ask "how strong are you in X?" — self-ratings cannot be evidenced.
Do not paste the role requirements into the question text; that context can be mistaken for the candidate's own experience.
Do not stack five topics into one question — split them.
Do not ask yes/no questions; they produce nothing to summarise.

Copy-ready question templates by role

Each set is written to be pasted straight into a role's question list. Four to six questions is the sweet spot.

Software engineer

Aim for projects and personal contribution, not opinions about technology.

Walk us through the project you spent the most time on in the last two years. What was the system for, who used it, and what part of it was yours? - the problem the system solved - your specific tasks in it - the languages, frameworks and tools you used yourself
Describe a piece of code or a service you built end to end. How did it get from an idea to production, and who else was involved?
Tell us about a bug or an outage you personally investigated. What did you look at, what did you change, and how did you confirm it was resolved?
Which parts of your stack do you use daily versus occasionally? Please name them and say where you used each one.
What kind of work would you like to do more of in your next role, and what would you rather step away from?

DevOps / platform engineer

Ask for the environment and the ownership boundary — that is what a hiring client wants to read.

Describe the infrastructure you worked with most recently. - cloud or on-prem, and roughly how many services - what you owned versus what another team owned - the tooling you used for CI/CD, monitoring and infrastructure as code
Tell us about a deployment or migration you ran. What was moved, what was the plan, and what did you do when something did not go as expected?
How are alerts handled where you work now, and what was your part in on-call?
Describe a change you made that reduced manual work. What was manual before, and what did you build?

QA / test engineer

Keep the focus on the process the candidate actually ran.

Describe the testing setup on your last project — what was automated, what stayed manual, and which of it did you build?
Walk us through how a feature moved from a developer to release on your team, and where you stepped in.
Tell us about a defect you found that mattered. How did you find it, and what happened afterwards?
Which test tools and frameworks have you used hands-on, and on which project did you use each one?

Project / delivery manager

Ask for scope, people and the decisions they made — as they reported them.

Describe the last delivery you led. - what was being delivered and for whom - the size and make-up of the team - your day-to-day responsibilities on it
Tell us about a delivery that slipped. What changed, what did you decide, and how did you communicate it?
How do you track progress and report it to a client or a steering group? Please describe the tools and the rhythm you used.
Which parts of delivery did you handle yourself and which were handled by other roles around you?

The closing questions every role should include

Availability — when could you start, and what notice period are you on?
Expectations — monthly gross salary for permanent roles, or an hourly rate for freelance work, with the currency you prefer.
Anything we did not ask about that you would want a hiring team to know.

Frequently asked questions

How many screening questions should a role have?: Four to six. Each answer runs one to three minutes, which keeps the whole screening under fifteen minutes and still gives the summary enough material to work with.
What makes a good AI candidate screening question?: One that asks for a concrete project and the candidate's own part in it. Questions that ask for opinions or self-assessment produce answers an AI summary cannot ground in evidence.
Should questions include sub-points?: Yes. A short lead sentence plus two or three bullet sub-points guides the candidate through everything you want covered, and the screening flags anything a candidate did not address.
Can candidates type instead of recording?: Yes. Every question can be answered by voice or in writing, and candidates can edit or re-record any answer before submitting.

Try these questions on a live role

Paste a set into VoltFlows, send one link, and read the structured summary that comes back. See how VoltFlows screening works.