How do you create an AI screening interview from a job description? Give the AI a trimmed version of the job description and let it draft interview topics. Cut those topics down to the three to five skills that decide success in the role, write instructions that tell the AI how to probe and score, and take the interview yourself before you invite candidates.
Most job descriptions are written to attract applicants, not to evaluate them. That is why the draft an AI builds from one still needs a recruiter's edit. This guide covers the four steps that turn a JD into an interview worth sending: preparing the JD, editing the draft, writing instructions, and testing the result. If you are new to the format, start with how AI interviews work.
1. Prepare the job description
A job description is the closest thing most teams have to a job analysis, which is the foundation of a structured interview. The catch is that the AI treats everything you paste as a requirement. Ten listed skills become ten topics. Perks and culture copy turn into questions nobody needs answered. A short edit up front saves a long one later.
- Cut what isn't about the work: company history, benefits, the equal opportunity statement, and application instructions.
- Mark three to five must-haves. Only skills a hire cannot succeed without belong in a first-round screen.
- Rewrite vague requirements as observable work. "Excellent communication skills" gives the AI nothing to test. "Explains product issues to non-technical customers on calls and in email" does.
- Add what the JD leaves out. Which skill matters most? What usually makes someone fail in this role? Most tools let you add this as extra context alongside the JD.
Then pick a length. A 15-minute screen with three to five topics suits a first round. A 30-minute interview with more topics suits a shortlist or a senior role. Not sure the role needs an adaptive conversation at all? This comparison of one-way vs conversational AI interviews helps you decide.
On Equip, you paste the JD, set the experience range, add context, and the AI drafts topics in seconds. The help center has the step-by-step setup.
2. Edit the AI's draft
Read each drafted topic and ask one question: would a strong first-round recruiter spend time on this? Then make four kinds of edits.
- Cut what another round already tests. If a skills assessment covers a tool or technical skill, spend interview time on reasoning and experience instead.
- Match the format to the item. Eligibility checks like notice period or location need fixed wording and no follow-ups. Skills and past experience need an open topic the AI can probe. Case studies and code reviews work best as a scenario with reference material on screen. For ideas on fixed questions, see these preliminary screening questions.
- Spend follow-ups where depth matters. Follow-ups are what separate a conversational interview from a questionnaire, but each one uses time. Give them to the skills you most need to test.
- Weight the make-or-break skill highest. Otherwise a candidate who does well on easy topics but misses the core skill can still top your list.
Write each topic as a brief, not a label. "Account prioritization" is a label. "How the candidate decides which accounts get attention when several renewals are at risk in the same month" is a brief. Check the help center for details on questions, topics, and scenarios on Equip.
Example: a Customer Success Manager screen
Here is how four lines from a Customer Success Manager JD become a 15-minute first round.
| JD says | Interview item | Weight and follow-ups |
|---|---|---|
| Handle escalations from customer leadership | Topic: handling an escalation from a senior stakeholder | Highest weight, 2 follow-ups |
| Own renewals for mid-market accounts | Topic: prioritizing when several renewals are at risk | High weight, 2 follow-ups |
| Run onboarding and quarterly business reviews | Topic: structuring a business review for a new account | Standard weight, 1 follow-up |
| Hybrid, three days a week in Bengaluru | Question: notice period and office attendance | Standard weight, no follow-ups |
Two JD lines were cut. CRM proficiency is better checked in a skills assessment. "Strong communication skills" is not a topic at all: the AI hears the candidate communicate in every answer, so it belongs in the scoring instructions instead.
3. Write instructions for the AI interviewer
Good instructions cover six things:
- Context: the role, the seniority, and the kind of company.
- Tone: how formal the conversation should feel.
- Pacing: whether to start easy and raise difficulty as answers get stronger.
- Vague answers: what to do when a candidate talks in generalities.
- Boundaries: what the AI should never do or ask.
- Scoring cues: what earns a higher score and what earns a lower one.
Here is a sample for the Customer Success Manager screen above.
Sample instructions
You are interviewing candidates for a Customer Success Manager role at a B2B software company. Candidates have three to five years of experience.
Keep a friendly, professional tone and use plain language.
Start each topic with an open question. Raise the difficulty only if the first answer is strong.
If an answer is general, ask for one specific example from the candidate's own work: what they did and what happened as a result.
Do not give hints, suggest answers, or say whether an answer was correct. Do not ask about age, marital status, family plans, religion, caste, health or disability, or salary.
Score higher when the candidate describes their own actions rather than the team's, names a measurable outcome, and explains things clearly enough for a non-technical customer. Score lower for answers that only describe a general process.
Write each line as a rule you could check against the transcript later. "Ask for one specific example when an answer is general" is testable. "Be thorough" is not. Rules like these are also what let a recruiter question a score, which is the difference between a real AI interviewer and agent-washing.
4. Configure the round and test it yourself
A few decisions shape the candidate experience as much as the questions do:
- When candidates can start. An open window suits most roles. A fixed window helps for campus drives and hiring events.
- Whether a dropped attempt can resume. Allow it if candidates may be on unreliable connections.
- What candidates are told. Say the round uses AI, roughly how long it takes, and that a person reviews every result.
- Video or audio only. Audio-only lowers the barrier for candidates who are camera-shy or on weak bandwidth.
- A space for code or written answers. Useful for technical roles, where candidates need to show work as well as talk it through.
- Proctoring. Tab-switch detection and a second camera catch most AI-assisted cheating. Here is why traditional proctoring fails against ChatGPT.
FYI, Equip lets you configure all of the above.
Before you invite anyone, take the interview as a candidate. Check that every question makes sense spoken aloud, that the AI follows up where you expected, and that the report weights the skills you care about. If a question confuses you, it will confuse candidates.
See it from the candidate's side
Take a demo AI interview, then read the report
The demo takes a few minutes. The sample report shows how each topic is scored.
Weighing this against your current first round? Read conversational AI interviews vs traditional phone screens. Screening hundreds of applicants a month? This volume hiring strategy shows where the interview fits.
FAQ
Can AI generate interview questions from a job description?
Yes. AI interview tools read the JD and draft topics from it, then generate questions around each topic during the live conversation. Edit the draft first, because a JD usually lists more skills than a first-round screen can test.
How many topics should an AI screening interview have?
Three to five for a 15-minute first round, which leaves room for follow-ups. Use more for a 30-minute interview with a shortlist.
What should instructions for an AI interviewer include?
The role context, tone, pacing, how to handle vague answers, what never to ask, and what earns a higher or lower score.
Should I paste the full job description?
No. Paste a trimmed version with responsibilities and must-have skills, and add which skill matters most as extra context.