Forty resumes landed by Friday. The business still needs running, so you open ChatGPT and wonder whether you can screen job applicants with AI and skip the part where you read every single one. You can. The call itself stays with you.
The short version
To screen job applicants with AI and keep the human touch, let AI summarize each application against a must-have list you wrote, and never let it rank, score, or reject anyone. You read the summaries, you pick who moves forward, and a real person replies to every applicant. AI saves you the reading. You keep the decision.
What AI Does Well When You Screen Job Applicants With AI
AI is excellent at the boring middle of hiring. Give it a stack of resumes and a short list of requirements, and it can tell you in seconds which requirements each person clearly shows, which are missing, and what it can’t tell from the page. An hour of squinting at PDFs becomes five minutes of reading.
It also drafts the screening questions you send to applicants, writes the first polite reply to everyone who applied, and flags the application where someone forgot to attach the thing you asked for. That’s hours back every hiring round, and none of it requires AI to judge a single human being.

The quality of the output depends on the list you hand it. A list built from the real tasks of the role gives you useful summaries. If you haven’t written those tasks down yet, start with how to write a job posting that gets applicants, because the posting and the must-have list should say the same thing.
How Do You Write the Must-Have List AI Screens Against?
Start with the work itself. Write down the four or five things someone has to be able to do on day one, in plain language, such as “answers client calls and logs each one in our scheduling software” or “drives to job sites by 7 a.m. with their own tools.” Skip the adjectives. “Hardworking” and “team player” give AI nothing to check, so it fills the gap with its own guess about what sounds impressive.
Then split the list in two. Must-haves are the things a person cannot start without. Nice-to-haves are the things you can teach. AI should only ever screen against the first group. Treating a nice-to-have like a must-have is how a good applicant gets summarized out of your pile before you ever see them.
What AI Should Never Decide
Rejecting a person is where AI screening goes wrong. University of Washington researchers ran more than 550 real resumes through three large language models, changing only the names, and the models favored white-associated names 85% of the time, according to their 2024 findings. The resumes were identical. The names changed the ranking.
A small business owner doesn’t have a legal team to catch that, so the rule is simple. AI summarizes. You decide. Tell it to skip ranking and scoring, and to ignore anything it could guess about a person’s age, race, gender, or background from a name or a photo. Then read the whole pile yourself, or at least a real sample of the ones AI called weak, because the best hire is sometimes sitting in the pile it undersold.
Rules on automated hiring tools also vary by city and state, so check yours before you switch one on. A bad hire already costs a small business plenty, and what a bad hire actually costs is a good reminder that a shortcut that picks wrong is no shortcut.
A Screening Process That Keeps the Human Touch
This order works. It takes about as long as the old way took to open the first resume.
| Step | Who does it |
|---|---|
| Write the must-haves from the real tasks of the role | You |
| Summarize each application against that list, no ranking | AI |
| Read the summaries and a sample of the low ones, then choose who moves forward | You |
| Draft a first reply for every applicant | AI drafts, you edit and send |
| Hold a ten-minute call before any full interview | You |
Step two is where the prompt matters. This is the one I’d paste in: “Act as my hiring thought partner. Here are my five must-haves for this role. For each application I paste below, tell me which must-haves it clearly shows, which are missing, and what you can’t tell. Do not rank, score, or reject anyone. Do not guess anything about a person’s age, race, gender, or background. Tell me where you disagree with my list.” That last line matters, because ChatGPT is trained to agree with you, right? Tell it not to.
Step four is where most owners lose the human touch without noticing. Every applicant, including the ones you pass on, deserves a reply from a real person within a couple of days. AI can draft it in your voice in ten seconds. You read it, change one line so it sounds like you, and send it. Applicants remember who answered them, and that reputation travels further than you think.
The ten-minute call in step five is the cheapest filter you own, and for a VA or office manager role you can add the last-three-things-you-did-with-AI question to it. Ask the person to walk you through the last time they did the main task of the role, and listen for specifics. You learn more in ten minutes of talking than in ten resumes of reading, and the applicant gets a first impression of a business run by someone who picks up the phone. Keep the questions identical for everyone so you can compare answers fairly afterward.
See What ChatGPT Says About Your Business
Right now, someone in your city is asking ChatGPT who’s the best at what you do. I’ll go ask that exact question myself… and hand you back the real answer, word for word, with live screenshots. Not a guess. Not a summary. The actual sentence AI is already saying about you, or about someone else instead of you. Before your next customer reads it first. Custom to your business. Free.
No cost. Zero credit card. No pushy sequence.
Should You Tell Applicants You Use AI?
Yes. Add one sentence to the posting: “We use AI to help summarize applications, and a person reads every one.” It costs you nothing, it builds trust before the first call, and in some places it may be required. It also keeps you honest, because you can’t write that sentence unless it’s true.
What Should You Never Paste Into an AI Tool?
Resumes carry home addresses, phone numbers, and sometimes more than the applicant meant to share. Before you paste a stack of them into a chatbot, check the privacy settings of the tool you’re using and find out whether your inputs can be used for training. Remove anything you don’t need, like addresses, birth dates, and ID numbers. The summary only needs the work history and the skills, so that’s all the tool should see.
Keep a simple habit as well. Delete the chat once you’ve finished the round, and keep your notes in your own files. Applicants handed you their information for one purpose, and a short, tidy process respects that.
Where AI Screening Goes Wrong
Three mistakes show up again and again. Owners paste in a vague posting and get vague summaries back. They let the tool rank candidates and treat the ranking as the answer. And they stop reading, because the summaries look so tidy. The fix for all three is the same: keep your hands on the list, the decision, and the reply. The same discipline that makes SOPs written with ChatGPT usable applies here. AI drafts from the detail you give it, and you check the result against what you know.
Run one hiring round this way and track two numbers: how many applications came in, and how many were worth a ten-minute call. If the second number is a healthy share of the first, your must-have list is doing its job. If almost everyone made the cut, the list is too loose, so tighten it before the next round and run the same steps again.
If you're still working out where AI fits in the rest of your business, where to actually start with AI as a small business owner is the right next read.
Frequently Asked Questions About Screening Job Applicants With AI
Can I use ChatGPT to screen resumes?
Yes, as a summarizer. Give it your written must-have list and ask which requirements each resume clearly shows, which are missing, and what it can't tell. Don't ask it to rank, score, or reject anyone, and read the results yourself before you decide who moves forward.
Is it legal to use AI to screen job applicants?
Rules on automated hiring tools vary by city and state, and some places require notice or audits. Check the rules where you hire before you switch a tool on, and keep a real person making every decision.
Can AI be biased when screening applicants?
Yes. A 2024 University of Washington study ran more than 550 resumes through three large language models, changing only the names, and the models favored white-associated names 85% of the time. That is why a person should make every decision and read the pile.
How do I keep the human touch when I screen applicants with AI?
Reply to every applicant with a message a person read and edited, hold a short phone call before the full interview, and tell applicants a person reads every application. AI handles the reading and the first draft. You handle the relationship.
What should I tell ChatGPT when I screen job applicants?
Give it your must-have list and tell it to summarize each application against that list without ranking, scoring, or rejecting anyone. Add that it should not guess a person's age, race, gender, or background, and ask it to tell you where it disagrees with your list.
Should I tell applicants I use AI to screen them?
Yes. A single sentence in the job posting saying AI helps summarize applications and a person reads every one builds trust and keeps you honest. In some places, notice may also be required.
Want a free chat with Bella?
Thirty minutes to see how you can improve. No pitch, no homework.
About the Author
Bella Vasta built and sold a service company before becoming an AI implementation specialist, keynote speaker, and host of the Bella In Your Business and AI For The Busy Human podcasts. She's the founder of Jump Consulting and Marketing With Bella, where she teaches small business owners nationwide how to use AI without the overwhelm. Free AI visibility check: bellavasta.com/audit. Speaking inquiries: bellavasta.com/speak.
Always keep jumping.