Caregiver Recruiting Strategies: AI for Ads, Screening and Follow-Up

A home care agency can use AI to draft caregiver job ads, write the same screening questions for every applicant, and message lapsed candidates, while people read every application, run the interviews and check references. Keep AI away from ranking or rejecting applicants; use it to speed up the writing, not the judgement.

Generic recruiting advice for care agencies says "post on more boards" and "offer referral bonuses". Those help, but they ignore where office time actually goes: writing ads that do not sound like everyone else's, answering the same questions twenty times, and losing applicants in the silence between first contact and interview. Those are the jobs to hand to AI, in four parts: the ad, the screening, the follow-up and the checks that keep it fair. The judgement about people stays with you at every step.

Follow me on Instagram@sagnikteaches

The job ad is where most agencies lose people first

A caregiver reading ads on a phone makes a decision in a few seconds. The ads that lose them lead with the agency's history, bury pay at the bottom and use words like "passionate" and "rewarding" that every other ad uses. Applicants want to know four things at once: what it pays, when they would work, where, and whether they will be left alone on day one.

Connect on LinkedInSagnik Bhattacharya

AI is good at turning your messy facts into clean copy, provided you give it the facts and forbid it from inventing any. A reliable prompt:

Subscribe on YouTube@sagnikteaches

Write a job ad for a home care agency hiring care assistants for home visits.
Use ONLY these facts, and write "[to confirm]" for anything missing:
- Pay: $[rate] an hour, plus mileage between visits
- Shifts: mornings 7-11 and teatime 4-8, 20 to 30 hours a week, every other weekend
- Area: [towns or postcodes covered]
- Needs: driving licence and own car (mileage paid), willingness to be trained
- Support: paid induction, three shadow shifts, a named buddy, on-call number
- Checks before start: [list as your regulator requires]
Format: a headline with the pay in it, two short paragraphs, then a bulleted
"what you'll need" list. 180 words maximum. Plain words, no "passionate", no
"rewarding career", no "fast-paced".

A sample output (illustrative): "Care assistant, $[rate] an hour plus mileage. Mornings 7-11 and teatime 4-8, 20 to 30 hours a week, every other weekend. We visit people at home in [towns]. You'll need a driving licence and your own car; we pay mileage between visits. You do not need to have worked in care before. Your first week is paid induction with three shadow shifts alongside an experienced carer, and a named buddy for your first month. Checks are required before you start; we explain them when you apply." What you fix: confirm the pay line and mileage rate; check that "you do not need to have worked in care" is true for the roles you are hiring; add how to apply.

Three ads for three people

One ad cannot serve everyone. Ask the AI for variants from the same facts, and test them over a few weeks.

• The career-changer: leads on paid training and a buddy. A former hospitality worker wants reassurance more than a care-sector vocabulary.
• The experienced carer: leads on pay, rota stability and visit-length rules, such as no 15-minute calls. This is the person other agencies are trying to poach.
• The school-hours parent: leads on hours that fit, such as 9.30 to 2.30, and on the guarantee that the rota is published a week ahead.

Keep the facts identical across all three. Only the order and emphasis change, so nobody sees an ad that promises something the others do not.

Checking the ad for bias before it goes live

AI drafts can repeat the habits of the text they learned from, including phrases that put off older applicants, men, or people with disabilities who could do the job perfectly well. Add a second pass in a fresh chat, so the AI is reviewing and not defending its own draft:

Review this job ad for wording that could discourage a qualified applicant
because of age, gender, disability, nationality, accent, religion or family
status. List each phrase, say who it might put off, and suggest a neutral
alternative. Then list every requirement and say whether it is essential to
the work or merely a preference. Do not rewrite the ad. Ad: [paste].

Typical catches (illustrative): "energetic young team" (age); "must be physically fit" (blunt; better to describe the real task, such as helping someone to stand, and say reasonable adjustments are considered); "native English speaker" (say "confident speaking and writing clear English for care notes" if that is the real need); "female carers only" (in many places a sex-specific requirement is lawful only in narrow cases, such as intimate personal care a client has asked for, so take employment-law advice before using it). The AI flags; you decide. If you want a broader treatment, writing job adverts with ChatGPT without biased language covers the wording side in detail.

Screening questions: the same ones for everyone

Fair screening starts with consistency. If one applicant is asked about their experience and another is asked about their childcare, you have a problem however smart your tools are. A short written question set, asked of everyone in the same order, reduces that risk and gives you comparable answers.

Ask the AI to draft questions that test the things the job needs, and to keep each one open and about behaviour. The prompt:

Write 7 screening questions for applicants to a home care assistant role,
to be answered in writing or by phone in about 10 minutes. Each must:
- relate to a duty in this list: [lone working, dignity, reliability,
  recording, reporting concerns, driving between visits, working with
  people living with memory loss];
- ask for a real example or a "what would you do";
- avoid anything about age, family, health, nationality or religion.
For each question add what a good answer includes (2-3 bullet points) and
one sign that needs a follow-up question.

A sample question and guide (illustrative): "A client you have visited twice says they do not want to take their tablets today. What do you do?" A good answer: stays calm, does not force or trick, offers a choice, records it and tells the office or supervisor, follows the care plan. A follow-up sign: "I'd just leave them to it" with no mention of telling anyone. The guide is for the interviewer; it is not an automatic score.

Using AI to summarise answers, not to score them

When forty applications arrive in a week, summaries save real time. Paste each answer set (with the name removed and a number substituted) and ask the AI to list, for each question, what the applicant said in one line, and to flag anything the guide says needs a follow-up. Do not ask it to say who is "best". Then read the original answers yourself for anyone who is borderline, because summaries can flatten a thoughtful but short answer into nothing.

Two guardrails help. First, remove names, photos, addresses and dates of birth before pasting anything. Second, run a quick check once a month: take ten applications, read them without the summaries, and see whether you would have picked the same people to interview. If you disagree often, the summary prompt needs work.

Why not let the tool rank people?

Some recruitment tools offer automatic match scores. For a small agency the risks outweigh the time saved. A score cannot see context, such as a career gap caused by caring for a relative, a foreign qualification that does not map neatly, or a short answer from someone who is brilliant in person. It also hides its reasoning, so you cannot explain a rejection to an applicant who asks. Rules on automated decisions about people differ by place, and some give applicants the right to a human review, so take advice before using any scoring feature. The checklist in how to evaluate AI features in recruitment software gives questions to put to the vendor before you switch one on.

A simple policy that is easy to defend: AI may draft, summarise and schedule; a named person decides who is invited to interview, who is offered a job and who is declined. Write that sentence into your recruitment procedure.

Re-engaging lapsed applicants

This is the highest-return use of AI for most agencies, and the one the generic advice ignores. Over a year, an agency collects dozens of applicants who went quiet: they did not answer a call, found another job and then left it, or could not start until the school term ended. They already know your name. A short, human message costs almost nothing.

Start by sorting your applicant list into three groups with a spreadsheet or your recruitment tool: applied but never interviewed; interviewed but did not take the job; offered and withdrew. Each gets a different message. Use only people who gave consent to be contacted, and who were not rejected for a reason that still applies.

Write three short text messages (under 50 words each) from a small home care
agency to someone who applied [months] months ago and never interviewed.
Use the first name [first name]. Say we have new shifts in [area], pay is
$[rate], and ask if they would like a 15-minute call. Include "Reply STOP
and we won't message again". Tone: friendly and low-pressure, no pushing.
Make the three different in opening and length.

A sample (illustrative): "Hi [first name], it's [your name] at [agency]. You applied with us a while back. We now have morning and teatime shifts in [area] at $[rate] an hour plus mileage. Would a 15-minute call this week suit? Reply STOP and we won't message again." Send it from a proper business number, at a sensible hour, and send replies to a person. If someone says no, record it and do not message again.

Seventy lapsed applicants, one afternoon: the numbers

Say an agency with 18 caregivers needs four more. Over the past year 120 people applied; 60 were never interviewed, 35 interviewed but did not start, 25 withdrew after an offer. Suppose consent covers 70 of those. The office sends 70 messages, which AI drafted in three variants and the coordinator checked in 20 minutes. If 12 reply, that is a response rate of about 17 per cent; 6 agree to a call; 3 reach interview; 1 or 2 start. These figures are illustrative, but the shape is realistic: you will not rebuild a team from old lists, yet a handful of warm starters from a free 20-minute exercise is often worth more than another week of paid advertising. Track your own numbers so you know what the list is really worth.

Keeping applicants warm between steps

Applicants drift away during silence. The most common gaps in a small agency are between application and first call, between first call and interview, and between offer and start date, which also overlaps with pre-start checks. For each gap, write a standard message and let AI vary the wording lightly so it does not read like a template.

A realistic sequence: an acknowledgement within a day with what happens next; a first-call invitation with two time options; an interview confirmation with directions and who they will meet; a message after the interview, whether yes or no, within three working days; and a "here is what happens before you start" note once an offer is made. Applicants who are declined deserve a courteous message, written by a person or reviewed by one. The point is to keep people informed, not to maintain an automatic stream of messages.

The week after the offer is the riskiest. A starter who has heard nothing for nine days may accept another job. Pair each offer with a dated plan of the checks and the induction, which a caregiver onboarding checklist with an AI-drafted induction plan sets out step by step.

Where the same ad should go, and how to adapt it

One master ad becomes several short pieces for different places. Ask the AI to cut the master into a 40-word social post, a 90-word job-board listing, a 25-word message for a local community group, and a text that existing caregivers can forward to a friend. Give each the same facts and ask it to keep the pay and the shift pattern in the first line.

Your own caregivers are the best channel, and an AI-drafted forwardable message makes it easy for them to share: "We're hiring morning and teatime carers in [area], $[rate] an hour plus mileage, paid induction and a buddy. Message [first name] at the office if you'd like to hear more." If you pay a referral bonus, state the terms plainly and have a person check them, since AI will happily make up an amount.

A phone-screen script for the first ten minutes

The first call filters out mismatches kindly and early. A short script keeps it fair. Ask the AI to turn your screening questions into a call sheet with an opening line, two practical checks and a closing line. The practical checks are facts, not judgements: can they work the shifts on offer, can they get to the area, do they have the right to work and the documents you will need, and are they happy with the checks required before starting?

A sample opening (illustrative): "Hi [first name], it's [your name] from [agency]. Is now still a good time for about ten minutes? I'll ask a few questions about the role, and you can ask me anything at the end. I'm taking notes so I can be fair to everyone I speak to." The call sheet should leave room for the interviewer's own notes and end with what happens next and when. Tell applicants honestly if you are using any transcription tool, and never use one without their agreement.

Human interviews and reference checks

The interview is a human event. AI can prepare the interviewer by turning the screening answers into a one-page brief, with the questions to ask, the follow-ups the guide flagged and a reminder to ask every applicant the same core set. It should not sit in the interview scoring voices or faces. Applicants should be told if any tool records or transcribes the conversation, and should be allowed to decline it.

Reference checks matter even more in care than in most trades, because the applicant will work alone in people's homes. Ask for references that include recent care work where there is any, speak to referees rather than relying on written forms alone, and ask direct questions: would you employ them again, and was there any reason that you would not? Write down the answers. AI can draft the reference-request email and a question list; it cannot tell you whether a referee is hesitating. If something in a reference worries you, a conversation with your registered manager and, where needed, an adviser comes before any decision.

A reference email, drafted and then fixed

An illustrative first draft from AI might say: "We are considering [first name] for a role as a care assistant. Could you confirm their dates of employment, role and your view of their suitability?" What you add: a clear question about whether there were any concerns about conduct or safeguarding, a deadline, a phone number for a call, and a line stating that [first name] has consented to this request. The first draft was polite but would have let a lukewarm referee off the hook.

What to measure, so you know it worked

Pick four numbers and look at them monthly. Time from application to first contact, which should fall as templates and scheduling improve. Share of applicants who reach interview, which tells you whether the ad attracts the right people. Share of offers that start, which tests your follow-up and pre-start process. And first-90-day leavers, which tests whether you hired the right people and set them up well; caregiver retention check-ins are the next step for that.

Also keep a simple log of what AI drafted and who approved it. If an applicant ever asks why they were not progressed, you want a human reason on record, written in plain words, not "the tool said so".

Seven ways AI-assisted hiring goes wrong in a care agency

1. Pasting applicant names, addresses or dates of birth into a general AI chat. Remove identifiers first, or use a tool your data-protection review has approved.
2. Letting AI invent benefits. If the draft promises a bonus or training you do not offer, applicants will find out in week one.
3. Asking different applicants different questions, however well meant.
4. Treating a flagged phrase as proof of bias. The check finds candidates for review; you decide.
5. Over-messaging old applicants. One message, one reminder at most, then stop.
6. Skipping the human review of summaries for borderline applicants.
7. Forgetting that recruitment records are personal data. Set a retention period for unsuccessful applicants and delete on schedule.

For a wider view of tools and where they fit in a small team, AI recruiting tools for small businesses compares what is worth paying for, and recruiting a small team with AI from job ad to shortlist walks through a general version of the same flow. If you want to map your own hiring steps with someone, that is what my AI implementation consultation covers.

Recruiting questions home care owners ask

Can AI rank or reject caregiver applicants for me?

It is safer not to. Ranking and automatic rejection can carry hidden bias and may breach employment or data-protection rules where you operate. Use AI to draft ads, write consistent questions and summarise answers, then let a person read every application and decide who gets an interview.

What should a caregiver job ad include?

State the pay range or rate, the shift pattern, the area covered, whether driving is needed, what training and support you give, and the checks required before starting. Applicants decide quickly, so put the pay and hours in the first two lines rather than at the bottom.

How do I check an AI-written job ad for biased wording?

Read it for age, gender, physical and cultural cues such as young, energetic or native speaker. Ask the AI to list any such phrases, then confirm yourself. Keep requirements tied to the work: a driving licence if driving is needed, not a general preference for someone fit.

Is it fine to message old applicants with AI-drafted texts?

Yes, if they gave consent to be contacted and you follow your marketing and data-protection rules. Use their first name, say why you are getting in touch, offer an easy opt-out, and keep a record of who was contacted. Send a person's reply to a human, not back to the AI.

Further reads

• Home Care Agency AI: A 90-Day Plan for the Owner — Fit recruitment inside a 90-day AI plan for the whole agency.
• AI in Healthcare Staffing: What a Small Agency Can Automate — See what a small staffing agency can automate.
• Home Care Agency Employee Handbook: Draft With AI, Then Review — Give new hires a handbook that matches what the ad promised.
• Real Estate Agent Recruiting With AI: Outreach and Onboarding Plan — A different recruiting problem with a similar outreach loop.
• Home Care Agency Software Pricing: What the AI Features Add — Check whether your care software already has recruitment features.
• AI for Home Health Agencies: Faster Replies to Hospital Referrals — A referral intake checklist, ready-to-send missing-information requests and a response-time log, with AI under a signed business associate agreement.
• Domiciliary Care Rota Template: When Software or AI Fills the Gaps — A domiciliary care rota template with travel time and continuity rules, plus how AI can suggest gap cover while a coordinator makes the call.
• Domiciliary Care Spot Check Template: AI Summaries and Action Lists — A spot check form for home care, a fair scoring approach, and AI prompts that turn completed forms into summaries and action lists.
• Home Care Agency Marketing Plan: Build One With AI and Local Data — A one-page plan built from your own local data: search terms, referral partners, family messages and a budget, with an AI prompt and a reality check.
• Home Care Agency Welcome Packet: Build the Admission Pack With AI — Build a home care welcome packet section by section: what to include, the prompts that draft it, and the manager sign-off that makes it safe to send.
• Non-Medical Home Care Leads: An Enquiry-Reply Routine Using AI — A practical routine for answering home care enquiries within minutes: reply templates, after-hours cover and a response-time test.
• AI for Nanny Agencies: Family Briefs, Candidate Shortlists and Check-Ins — A nanny agency's guide to AI for family briefs, shortlist notes and check-in messages, with bias and data-protection safeguards.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general recruitment and equal-opportunity practice; applicant-tracking and care-software vendor documentation; your own regulator's and employment-law adviser's current guidance.