25 Pre-Screening Interview Questions Your AI Agent Can Score (2026)

Good pre-screening interview questions do one job: they tell you who deserves a real interview before anyone books a call. Most applications arrive as a resume and a cover letter. So a recruiter reads each one, guesses at the gaps, and schedules a 20 minute phone screen to fill them in. Multiply that by 80 applicants and the first week of hiring is gone.

A scored pre-screen fixes that. You ask every applicant the same short set of questions, each answer carries points, and the total decides who moves forward. Because the rules are written down, an AI agent can apply them the moment an answer lands. Then your team only talks to people who already cleared the bar.

This guide gives you 25 pre-screening interview questions grouped by what they measure, with a scoring rule for each. After the list, you will see how to turn them into a form and how your agent builds and scores that form in one message.

Pre-screening interview questions: what to score and why

A pre-screen is not an interview. It is a filter. The goal is to confirm four things quickly. The person can take the job, they have the core skills, they want this role, and they can work the way the team works. Everything else belongs in the real interview.

That is why the questions below are short and mostly closed. Closed questions produce comparable answers, and comparable answers can be scored. The SHRM toolkit on screening and evaluating job candidates describes screening as the step that narrows the pool before deeper evaluation. However, narrowing only works when every applicant gets the same questions. A form guarantees that; a phone call does not.

The four signals

• Eligibility and logistics: can they start, commute, relocate, or work the hours? A "no" here ends the process, so ask it first.
• Skills and experience: do they have the two or three things the role cannot work without?
• Motivation and fit: did they apply on purpose, and do they understand the job?
• Work style and availability: will their way of working match the team's?

The score buckets

Give each answer option a number of points, add them up, and compare the total with the maximum possible score for that form. Then sort applicants into three buckets:

• Advance (70% or more): book the interview.
• Review (40% to 69%): a recruiter reads the open answers and decides.
• Decline (under 40%): send a polite no and keep the profile on file.

Use percentages, not raw points, because you will not use all 25 questions on one form. A percentage keeps the buckets stable when you swap questions between roles.

25 pre-screening interview questions with scoring rules

Pick 8 to 12 questions per role. Replace the example answer options with your own requirements, and keep one open text question in each block so the agent has something to read beyond checkboxes. For open answers, give the agent a one line rubric, for example: "10 if they name a specific tool and a result, 4 if generic, 0 if blank."

Eligibility and logistics (questions 1 to 7)

Question 7 is the one most teams skip because it feels awkward. Ask it anyway. A mismatch of 30% discovered in the final round wastes everyone's month.

Skills and experience (questions 8 to 14)

Question 11 is where the agent earns its keep. A human recruiter skims 80 paragraphs in an hour. An agent reads all of them against the same rubric and never gets tired by applicant number 60.

Motivation and fit (questions 15 to 20)

Motivation questions rarely disqualify anyone. Instead, they separate the top of the Advance bucket from the middle. Two candidates with the same skills score will differ most here.

Work style and availability (questions 21 to 25)

Question 25 is optional for scoring but useful for the recruiter. In practice, it surfaces accommodations and expectations early, when they are easy to handle.

How to turn the questions into a pre-screening form

The list above only pays off when every applicant answers the same version of it. Here is how to build the form so that it is short enough to finish and structured enough to score.

Keep it under 12 questions

Candidates abandon long forms, and the ones who finish a 40 question screen are not always the ones you want. So pick two or three questions per signal and cut the rest. The screening questions to ask candidates for a warehouse role are not the ones for a backend engineer, which is why the template is a menu, not a script.

Put disqualifiers first

Work authorization, schedule and location go at the top. If a candidate cannot pass them, the form should end politely instead of asking 10 more questions. Conditional logic does this: show the rest of the form only when question 1 and question 6 are a "yes".

Write the scoring into the form, not into a spreadsheet

The mistake most teams make is to collect answers in one tool and score them in another. Points live in a spreadsheet. A recruiter copies answers across, and the rules drift within a month. Instead, attach the points to the answer options themselves. Then the score exists the second the applicant hits submit, and the routing can run from it.

This is the same model we use for lead qualification forms: every answer option carries points, the total lands in a bucket, and the bucket decides the next step. Hiring and sales look different, but the scoring logic is identical.

How your AI agent builds and scores the pre-screening form with dForms

Here is the part that removes the manual work. dForms is the form builder for AI agents, so your agent can build this form, publish it and read the scored answers without you opening a builder. You connect it once, with an API key from the Developers page or through the dForms MCP server at mcp.dforms.ai inside Claude. Then you ask for the form in plain language.

Paste a prompt like this one into your agent:

Build a pre-screening form in dForms for a Customer Support Specialist role.
Use these questions: work authorization (yes/no, "no" disqualifies), earliest start date
(within 2 weeks = 10 pts, 2 to 4 weeks = 8, 1 to 2 months = 5, later = 2),
schedule including one weekend day (yes = 10, partly = 4, no disqualifies),
expected salary range (inside 45k to 55k USD = 10, up to 10% above = 6, more = 2),
years in a support role (under 1 = 2, 1 to 3 = 6, 3 to 6 = 10, 6+ = 9),
tools used (Zendesk, Intercom, HubSpot: 3 pts each),
one open question about a difficult ticket they solved (rubric: specific issue and outcome = 10,
generic = 4, blank = 0),
and why they applied (concrete = 10, generic = 3).
Show the schedule and salary questions only if work authorization is "yes".
Score every answer, label totals of 70% or more as Advance, 40% to 69% as Review, under 40% as Decline.
Publish the form and give me the link.

The agent creates the form with the options and points you described, adds the conditional logic, publishes it and returns the public link. When applications come in, ask it to list the Advance bucket or summarize the open answers for the Review bucket. It can also send the Advance candidates to your calendar for an interview slot.

A few facts worth knowing before you hand this to an agent. The key acts with the same role as the person who created it, inside that workspace. An agent with a key can publish. So if you want a human to approve the form first, ask the agent to leave it in draft and publish it yourself. Edits live in a draft until published, and a published form can be reverted. Responses are free and unlimited, and the project is open source if you prefer to self-host.

If your agent builds other forms too, the same connection covers them. For example, the AI form generator guide shows everything an agent can do through the API. The AI SDR post shows the same scoring model applied to inbound leads.

Phone screening questions versus a scored form

Teams often ask whether a form can replace the phone screen. For most roles, the answer is that it replaces the first half of it. Phone screening questions about availability, salary and basic experience are exactly the ones a form handles better. Every applicant answers them, the answers are comparable, and nobody has to schedule anything.

What the form cannot do is hear how someone communicates. So keep a short call for the Advance bucket. Spend it on the questions that need a conversation: how they think about a problem, how they would approach the first 30 days, what they want to learn. Because the logistics are already settled, that call gets shorter and more useful.

Common mistakes with pre-screening interview questions

• Asking for a full resume twice. If the applicant already uploaded one, do not make them retype it. Ask only what the resume does not answer.
• Scoring open answers by gut. Write the rubric down, even if it is one line. Otherwise, two recruiters will score the same answer differently, and the agent will have nothing to follow.
• Hiding the disqualifiers at the end. A candidate who spends 10 minutes on a form and then learns the role is on-site will not apply again.
• Never revisiting the points. After a hiring round, compare the pre-screen scores with who actually got hired. Then adjust the points. The rules should improve each cycle.
• Mixing legal questions into the screen. Keep the questions about the job and its requirements. If you are unsure whether a question is allowed where you hire, check with your legal team before adding it.

The short version: the best pre-screening interview questions are the ones you ask every applicant, score the same way and review each quarter. A form makes the first two automatic, and an agent makes the third one cheap.

Frequently asked questions

How many pre-screening interview questions should a form have?

Eight to twelve is the practical range. Fewer than eight leaves gaps in the four signals; more than twelve lowers completion. Start with two or three questions per signal and remove any question whose answers never change your decision.

Can an AI agent score open text answers fairly?

Yes, when you give it a written rubric. The agent applies the same rule to every answer, which is more consistent than a tired human at the end of a stack. Keep a human in the Review bucket for borderline cases, and check a sample of the agent's scores each round.

Does my agent need special permissions to publish the form?

No. An API key in dForms acts as the member who created it, with that member's role in the workspace, and it can create, edit and publish forms. There are no per-key scopes, so if you want to approve the form before it goes live, ask the agent to leave it in draft.

Can I use these questions for phone screening instead of a form?

You can, but you lose the comparability. On a call, questions drift, answers go unrecorded, and scoring happens from memory. Use the form for the questions with fixed answers, then keep the call for the candidates who advanced.

Does the form work with Claude, ChatGPT and other agents?

Any agent that can call an HTTP API with a key can use dForms, and Claude can also connect through the dForms MCP server at mcp.dforms.ai. The questions and the scoring rules live in the form, so the agent you use does not change the results.

Is dForms free for candidate screening forms?

Yes. Responses are not counted or billed, the project is MIT licensed, and you can self-host it. Build the pre-screen above, publish it, and the applications start arriving scored.