How to Recruit Tutors: Job Ads, Screening and Scorecards With AI

A tutoring agency can use AI to write a clear job ad, summarise applications against stated criteria and prepare interview questions, then score every candidate on the same scorecard. People make every decision, run the reference and background checks, and review the process for bias. No applicant should be rejected by an automated tool alone.

The aim is a hiring process that is quick enough to keep good tutors interested and structured enough to be fair. Tutors often apply to several agencies at once, and the best often have a full diary, so a slow or muddled process loses them. At the same time, you are placing adults in front of children or young people, so careful checks matter more than speed. The way through is a short, fixed process that AI helps to prepare and people carry out.

Follow me on Instagram@sagnikteaches

Write down the role before you write the ad

An ad is only as clear as the role behind it. Before using AI, write a one-page role sheet. It becomes the input for the ad, the screening questions and the scorecard, so every later step measures the same thing. Include:

Connect on LinkedInSagnik Bhattacharya

• Subjects and levels. For example, lower-secondary maths and exam-year science, with the exam syllabuses you work with.
• Hours. The days and times pupils want, usually after school and at weekends, and the minimum hours a week a tutor must offer.
• Format. Online, in person at pupils' homes, at a centre, or a mix, and any travel.
• Pay or fee. A range or a fixed figure, how often paid, and whether the tutor is an employee or self-employed. The status depends on how you run the work, so take advice.
• Essential criteria. Subject knowledge, the right to work, a current background check for working with children, availability. Keep these few and real.
• Desirable criteria. Experience with a particular exam, a teaching qualification, experience of special educational needs.

Separating essential from desirable matters for fairness. If a degree is desirable but you reject candidates for not having one, you have turned a preference into a barrier. Test each essential criterion by asking whether a tutor could do the job well without it. If the answer is yes, move it to desirable.

Subscribe on YouTube@sagnikteaches

Draft the job ad with AI, then edit it with a sceptical eye

A good tutor ad is specific, honest about the hours and fair in its wording. Give the chatbot the role sheet, and ask for plain, inclusive language.

Write a job advert (about 250 words) for a tutoring agency recruiting tutors.
Use only these facts:
- Subjects: maths (ages 11 to 16), chemistry and biology (exam years)
- Hours: weekday evenings 4pm to 8pm and Saturday mornings; minimum 4 hours a week
- Format: online, with some in-person home visits within a 30-minute travel time
- Pay: $28 to $38 an hour depending on subject and experience, paid monthly; tutors are self-employed
- Essential: strong subject knowledge to at least the level taught, a current background check for working with children (we will arrange or verify it), two references, reliable internet
- Desirable: teaching or tutoring experience, a particular exam syllabus
- Apply: short form and a 10-minute call; we reply within five working days
Use plain, inclusive language. Avoid "rockstar", "native speaker", "young and energetic", "digital natives" and any phrase that suggests an age, gender or background. Do not promise a number of pupils.

A usable first draft is tidy but often slips in phrases that narrow the pool, such as "energetic" or "fresh graduates", or claims such as "a steady stream of pupils". Strike both. Check the pay and hours are exactly right, and add one sentence about what happens next. Then ask the chatbot to review the ad for words that could put off a qualified applicant, and to list any requirement not tied to the job. Treat that as a second opinion, not a verdict.

Before and after: an ad that excludes nobody

Before: "We are looking for an energetic young graduate to join our dynamic team and inspire pupils! Must be a native English speaker." After: "We are recruiting tutors for maths and science, with weekday evening and Saturday morning slots. You need strong subject knowledge, reliable internet and a current background check for working with children, which we can arrange. Pay is $28 to $38 an hour. Apply with the short form, and we will reply within five working days." The second version tells a qualified person whether to apply and does not exclude anyone for an irrelevant reason.

Screening questions that test the job, not the CV

A short application form beats a CV pile. Ask four to six questions that match the essential criteria, and use the same ones for everybody. Typical questions: Which subjects and levels can you teach, and to what standard have you studied them? Which days and times can you offer every week? How would you explain a hard concept to a pupil who has not understood it twice? How do you manage a pupil who has not done the homework? Do you have, or can you obtain, a current background check for working with children?

You can ask AI to propose the form and then reduce it. Keep it short, since long forms lose good candidates who are already busy.

Suggest six short screening questions for a tutoring agency application form.
Role sheet: [paste it]
Rules: each question must test an essential criterion, be the same for every applicant, be answerable in under 80 words, and avoid asking about age, health, family plans, religion, nationality, accent or anything unrelated to the role.
After the questions, list any question you think could disadvantage a group of applicants, and say why.

A good answer produces the six questions, and a second list that flags something like "Where did you go to school?" as irrelevant and potentially discriminatory. Remove anything that looks like that. Questions about the right to work and a background check are fine when asked consistently and in line with the rules that apply to you, so check the exact wording with an adviser.

Use AI to summarise, not to rank

When 40 applications arrive, a chatbot can speed up the read. Paste each application with the name, contact details, school and photograph removed, and ask for a plain summary against the essential criteria.

Summarise this application against our essential criteria only. Return a table with one row per criterion: subject knowledge, availability, background check, references, and a short quote from the application as evidence for each. Mark "no evidence given" where it is missing.
Do not score, rank or recommend. Do not comment on writing style, accent, age or anything not in the criteria.
Application (identifying details removed): [paste]

The output is a neat table that helps a person see at a glance which essential criteria are evidenced. A reviewer reads the table, compares it with the original and decides who goes to a call. If the model starts to make judgements ("a strong candidate"), tell it to stick to the evidence. Never let it send a rejection. The tutorial on bias checks every recruiter should run on AI screening goes into the testing in more depth.

An interview scorecard everyone fills in the same way

An unstructured chat rewards confidence and likeability, which are not the same as teaching well. A scorecard fixes what you are looking for before you meet anyone and records the evidence for each score. Keep it to five or six criteria on a 1 to 5 scale, with a line describing what a 1, 3 and 5 look like.

Ask the chatbot to draft two questions per criterion, with the follow-ups an interviewer should ask when the answer is thin. For "explaining to a pupil", a useful task is to ask the candidate to teach a short topic for five minutes, such as adding fractions with different denominators, to an interviewer playing a confused 12-year-old. That tests what the job is. The tutorial on building interview questions and scorecards with AI covers the general method, and this one stays on tutoring.

Two interviewers, one candidate, scored apart

Illustrative scores for one candidate, scored independently by two interviewers: subject knowledge 4 and 4; explaining 3 and 4; managing a lesson 3 and 3; parents 4 and 3; reliability 5 and 5; safeguarding 4 and 4. Total 23 and 23 of 30. The interviewers differed by one point on explaining and on working with parents, which is worth a short conversation before anyone decides. Write the evidence next to each score: "she tried a second example using pizza slices when the first did not work". Evidence stops scores from drifting towards a feeling.

Agree a pass mark in advance, such as 20 out of 30 with no score below 3 on subject knowledge or safeguarding, and keep it for every candidate. If an interviewer wants to override it, record why. Review the scores after a few hires: do high scorers do well with pupils? If not, change the criteria.

References, background checks and reminders

These steps protect pupils, and they are where good processes slip when the agency is busy. List every check you need before a tutor teaches a pupil alone, with the order and the owner. A typical list: identity and right to work, a current background check for working with children (the exact kind depends on your rules), two references, evidence of qualifications where relevant, and a short safeguarding briefing with a signed code of conduct.

AI cannot run a check, and it should not store the results. What it can do is keep the process on track. Ask the chatbot to turn your checklist into a tracker with columns for each step and a reminder date, and to draft the reference request.

Draft a short email (under 120 words) asking a referee for a reference for a tutor applicant.
Include: who we are, that the applicant named them, the five questions we would like answered (how long and in what capacity they have known the applicant, subject knowledge, reliability, conduct with children or young people, any reason they would not recommend them), and a reply-by date of [date].
Make clear that the reference is confidential, and thank them. Use [referee first name] and [your name]. Do not name the applicant in the subject line.

A usable draft is polite and clear. Adjust the questions to your rules and remove anything that invites opinions on matters unrelated to the role. When references arrive, a person reads them, and if something concerns you, the process stops until it is resolved. Record that the reference was received and by whom, but keep the contents in your secure system, not in a chatbot.

Set reminders for renewals too: background checks may need to be refreshed, qualifications can lapse, and tutors move between roles. A simple monthly list of "checks due in the next 60 days" prevents the awkward discovery that a tutor has been teaching on an out-of-date record. Pupils come first, and good records protect the agency as well.

The 10-minute screening call, scripted

A short call sits between the application and the interview and removes a lot of wasted time. Its job is to confirm the basics, check availability and hear how the person talks about pupils. Keep it to ten minutes and use the same outline for everyone: confirm the subjects and levels, confirm the weekly hours, ask one question about a time a pupil did not understand something, explain the process and the checks, and ask whether they still want to go ahead. Take two lines of notes against the essential criteria and nothing else.

An AI chatbot can prepare the call sheet, and a transcription tool can take notes, but check consent first. Tell the candidate if the call is recorded or transcribed, say why, and let them decline. If you use an AI note taker, the tutorial on candidate consent and tool choice for AI note takers explains what to ask and what to keep. If the candidate declines, take handwritten notes and carry on; nobody should be marked down for that.

Onboarding a new tutor in the first week

A tutor who has just been hired is the most open to guidance and the least familiar with your standards. A short onboarding plan makes the first lessons better and cuts the number of early complaints. It needs five parts: a welcome message with the schedule, a one-page guide to how you work (lesson length, reporting, cancellations, communication with parents), the safeguarding briefing and code of conduct, a first-pupil handover note, and a check-in after the first two lessons.

AI can help with the written parts. Give the chatbot your policies and ask for a plain one-page summary the tutor can keep, then read it against the originals, because a summary that drops a rule is worse than none. It can also draft the handover note for a new pupil from the information the family gave you, with the pupil's name and personal details left as placeholders that you fill in by hand. The tutor then sees what the pupil needs, what has been tried and what the parents asked for, before the first lesson.

Plan a feedback call after two lessons. Ask what went well, what surprised them and whether the pupil and parent seem a good match. Compare the answers with the scorecard: if a tutor scored 3 on managing a lesson and now reports trouble with an off-task pupil, you can offer support early. That loop also teaches you whether the scorecard predicts real performance, so adjust the criteria when it does not. Over a year, a simple record of scores, first-month feedback and parent comments will show which questions have been worth asking and which are only habit.

Pay, status and the paperwork that follows

Before the first offer goes out, settle whether tutors are employees or self-employed, because it changes the paperwork, the insurance and what you may require of them. The right answer depends on how the work is organised, so take advice from an adviser who knows the rules that apply to you, and do not rely on a chatbot for it. What AI can do is turn the decision into a contract summary in plain English for the tutor, once an adviser has approved the underlying terms: pay rate and payment dates, notice, cancellation, confidentiality, data handling and conduct. Give the tutor time to read it and ask questions before signing.

Illustrative sum for the pay range in the ad: a tutor charged out at $55 an hour and paid $34 leaves the agency $21 an hour to cover admin, insurance, checks, marketing and profit. At 20 paid hours a week per tutor, that is $420 a week of margin for the agency, or about $19,000 over a 46-week year. If recruiting and checking one tutor costs about $300 of time and fees, the $21 margin repays it within a week at 20 hours, or in about five weeks for a tutor who teaches three hours a week ($63 a week). Replace every figure with your own before you rely on it. The tutorial on how much to charge for tutoring shows the same sums from the tutor's side, which helps when you decide what to pay.

Run a bias check on your own process

Fairness is something you test, not something you declare. After each hiring round, look at who applied, who reached the call, who reached the interview and who was offered a place. If you collect equal-opportunities data, do it separately from the application, with consent and in line with the rules that apply to you. Look for steps where one group drops out more than others. If you do not collect that data, at least look for patterns in the reasons you gave for stopping applications.

A practical test for the AI step: take a sample of five applications, change only the name or a detail unrelated to the role, and run both versions through your summary prompt. The summaries should match. If they differ, stop using that prompt and review it. Also ask a colleague to read ten rejected applications and say whether they agree with each reason. The tutorial on using AI in hiring for a small shop shows the same ideas at a smaller scale.

Why no automated rejections

A tool that rejects people on its own is hard to defend. It can pick up patterns in past hiring that favour certain groups, it may misread a good answer, and the applicant has no one to ask. Depending on where you operate, data-protection law may limit decisions made solely by automated means, and employment and equality rules still apply to the outcome. Keep the rule simple and put it in writing: AI drafts and summarises, a person decides, and every rejection has a recorded reason tied to the criteria. If a candidate asks why, you can answer.

From application to first lesson in two weeks, with numbers

Speed keeps good candidates. An illustrative timeline for a tutoring agency that gets 30 applicants for six places: day 1 the ad goes live; days 2 to 7 a person reads the summaries and picks 14 for a ten-minute call; days 5 to 9 calls take place, and eight go to a 40-minute interview with a five-minute teaching task; days 8 to 12 two interviewers score separately, and six candidates pass the agreed threshold; days 10 to 14 references and background checks start for those six, with offers made once the checks are complete and a first lesson scheduled after the safeguarding briefing.

The AI work in that timeline is about three hours: an hour on the ad and form, an hour on application summaries, and an hour on scorecard questions and reference emails. The human work is much larger, and rightly so. If the agency hires six tutors a year, those drafts become templates, so the second round takes half the time.

Hiring slips that cost an agency good tutors

• A vague ad that draws 200 unsuitable applications. State hours, checks and pay, and watch the number fall and the quality rise.
• Skipping the teaching task. A brilliant mathematician may be unable to explain fractions to a ten-year-old. Five minutes of teaching shows the difference.
• Letting the first interviewer decide. Independent scoring protects against a charming first impression.
• Starting a tutor before checks are complete. It is tempting when a family is waiting. Do not do it; ask the family to wait or offer another tutor.
• Pasting candidate details into a public chatbot. Remove names and contact details, and use a business plan with data controls for anything more.

Once tutors are on board, the work moves to matching them with pupils, scheduling and reports, and if you are still building a client base, getting your first ten local clients comes first. If you would like to map your hiring steps with someone and decide which parts AI should draft, that is what my AI implementation consultation covers.

Questions agencies ask about hiring tutors with AI

Can AI reject tutor applicants for us?

It should not. Use AI to draft ads, summarise applications against your criteria and prepare interview questions, but let a named person decide who moves forward and record why. Automated rejections are hard to explain, may breach employment or data-protection rules, and can screen out good tutors for the wrong reasons.

What should a tutor job ad include?

The subjects and levels, hours and pay or fee range, whether work is online or in person, the checks you require, how to apply and when you will reply. A clear ad filters itself, because applicants who cannot meet the hours or the checks do not apply.

Do tutors need background checks?

If tutors work with children or vulnerable people, a background check for that kind of work is normal and often required. The exact check, who may request it and what you may do with the result depend on your rules, so ask your regulator or adviser and write the process down.

How many interviewers should score a candidate?

Two where you can, scoring separately before they compare notes. Independent scores reduce the pull of a first impression and make disagreements visible. Use the same questions and the same 1 to 5 scale for every candidate, and keep the scores with the record.

Further reads

• Best AI Tools for Tutors: What a Private Tutor Can Actually Use — Which AI tools your tutors may use, and which to avoid.
• Tutor Scheduling Software: Bookings, No-Shows and AI Reminders — Scheduling for a growing tutor team.
• AI Recruiting Tools for Small Businesses: What's Worth Paying For — Recruiting tools worth paying for at a small size.
• How to Test Job Candidates' AI Skills in an Interview — Test a candidate's AI skills in interview.
• How Tutoring Centres Use AI to Write Parent Progress Reports — What your tutors will write for parents each term.
• AI CV Screening for Recruitment Agencies: Setup and Safeguards — Screening setup and safeguards from a recruitment agency view.
• Tutoring Centre Admin With AI: Enquiries, Timetables, Invoices — Enquiries, timetables and invoices once tutors are placed.
• Tutor Safeguarding Checklist: Policy, Checks and Records With AI — What a tutoring agency's safeguarding policy, vetting checks and incident log must contain, with AI prompts for drafting and a legal-review step.
• How to Start a Tutoring Agency: Tutors, Pricing and AI Admin — The margin per lesson hour, tutor and parent terms, safeguarding basics and first-month admin for a new tutoring agency, with AI doing the drafting.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: the agency's own role requirements, employment and equality guidance for the reader's market (reader to check), safeguarding and background-check rules for working with children (reader to check), and the help pages of the AI tool used.