Your Next Applicant Is a Bot: Why Smarter Friction Beats More Filters

Candidates are applying to dozens of jobs in one sitting now, with AI doing most of the typing. Your inbox fills up, the quality drops, and the first thing most teams want to do is add more filters. I get that reflex, but I think it's the wrong move. The answer isn't more filters. It's smarter friction.

What's actually happening

The tools are just very good now, and they're cheap. A candidate can paste a job ad into ChatGPT, get a résumé that has every keyword from the description, and send it in a few minutes. Or they skip even that and buy an auto-apply tool that sends applications for them.

The numbers back this up. In June 2025, The New York Times reported that applications on LinkedIn had gone up more than 45% in a year, to about 11,000 applications per minute, and that AI tools are part of why (Semafor, eWeek, both summarizing the Times piece). Greenhouse's 2025 AI in Hiring survey of 1,200 U.S. job seekers found that about three in four use AI personally and 49% were sending more applications than the year before. 34% of recruiters said they spend up to half their week filtering spam and junk applications (Greenhouse, November 2025).

And in July 2026, Greenhouse's CEO told Fortune that tools to auto-apply to every job on their platform cost "like 20 bucks," and that the average job there gets around 254 applicants (Fortune, July 2026). He calls it the AI doom loop. Candidates use AI to apply to more jobs, companies use AI to filter them back out, and it keeps getting worse for both sides.

One thing I want to be fair about. Most candidates aren't running a fully automatic bot. In Greenhouse's October 2026 survey, 77% of U.S. job seekers had used AI in their search, but only 5% said they'd let an AI agent apply to jobs on its own (Greenhouse, October 2026). So "your next applicant is a bot" is a bit of a stretch. It's more like your next applicant is a person, and AI wrote most of what you're reading. For you, that ends up almost the same. Everything starts to look alike and you can't tell who actually wants this job.

Why the classic fix makes it worse

So what do most companies do when they get flooded? They add stuff. A mandatory cover letter. Make an account first. Twelve more fields. A list of knockout questions. Upload your résumé, and then type the whole résumé into the form again.

The thing is, every one of those steps is free for AI. A cover letter costs ChatGPT about four seconds. A long form is exactly the kind of thing an auto-apply tool was built for. The bot doesn't get tired at field 14. It doesn't care that it has to make an account.

You know who does care? The good candidate. The one who already has a job, sees your vacancy on their phone during lunch, and thinks "ok, maybe." They hit the account wall, or the cover letter, and they close the tab. Not because they're lazy, but because they weren't desperate in the first place. That's the person you wanted.

So the classic fix filters the wrong way around. It barely slows down the people mass applying, and it drives away the people who were on the fence. You end up with more junk as a share of what's left, not less.

Bad friction vs. good friction

I don't think friction is bad. A form with zero effort is also a problem, because then applying means nothing. The question is which kind of friction you add.

Bad friction is effort that anyone or anything can do, and that tells you nothing about the person. An account, a cover letter, retyping your work history, a 40-question form. AI does all of it perfectly, so it filters for patience, not for interest.

Good friction is one small, specific moment that a real person who actually wants this job can do well in a minute or two, and that's hard to fake across 200 applications at once. It's about the actual job, not jobs in general.

I want to be honest here too. Good friction is not bot-proof. Someone can paste your question into ChatGPT and get a decent answer. But two things happen. A question about a real moment from your job gets a generic answer from AI that you'll spot in five seconds once you've read a few real ones. And someone who sends 200 applications in one evening isn't going to stop and record 200 voice notes. The point isn't to make cheating impossible. The point is to make interest visible.

There's a nice signal in the Fortune piece about this. Greenhouse lets candidates flag one role a month as their "dream job," and the CEO says those candidates get hired at roughly five times the rate of everyone else (Fortune, July 2026). That's his claim and not an independent study, but the idea makes total sense to me. When something is scarce, choosing it means something. Good friction does the same thing on a smaller scale.

7 kinds of good friction

These are the formats I keep coming back to. You only need one of them per job, two max.

1. The real situation question. Take a moment that actually happens in the job and ask what they'd do. "A truck shows up two hours late and the next one is already at the dock. What do you do first?" You learn way more from that than from "Are you stress resistant?"

2. Pick one of two. Give two real situations from the job and ask which one they'd rather have. "A quiet night shift with three patients who need a lot of care, or a busy day shift with ten who need less?" No right answer, which is exactly why it's hard to fake. It also helps you place people.

3. The 30-second voice note. "Record 30 seconds: why this job, why now?" Optional is fine, it still sorts people. Anyone can record it on a phone, but nobody mass applying is going to record 200 of them.

4. The "what caught your eye" question. "What in this job ad made you click apply?" If your ad is specific, real candidates point to something specific. If your ad is generic, well, that's a different problem, and the job ad scorecard is a good place to start.

5. The first-week question. "If you started Monday, what's the first thing you'd want to learn?" Good people already think about this. AI writes "the company culture and processes."

6. The honest dealbreaker. "What would make you say no to this job?" Weird question for a form, I know. But it gets you honest answers about hours, travel or salary before the first call, instead of after the third.

7. The tiny work sample. One small thing, not a take-home test. A salesperson writes the first line of a cold email to one specific type of customer. A developer explains the last bug they were proud of fixing in three sentences. Keep it under two minutes or it turns into bad friction again.

If you want good input for questions 1 and 2, ask someone who does the job right now what their day looks like. The "day in the job" tutorial has the six questions I use for that.

How to keep it under 5 minutes anyway

This is where most teams mess it up. They add the good friction and keep all the bad friction too. Then you have a long form plus a voice note and nobody finishes it.

So the trade is: for every good-friction question you add, you remove at least one piece of bad friction. Kill the account. Kill the cover letter. Kill every field you don't need before the first conversation. The under-5-minutes tutorial walks through exactly how to do that with a timer and your own phone.

What you want to end up with is name, phone or email, a résumé or LinkedIn link (optional if you can live with it), and one good-friction question. That's it. Test it on your phone, and if it takes you more than five minutes, something has to go.

One thing I don't know for you is what this costs in completion rate, because that depends a lot on the job and the channel. My advice is to measure it yourself. Look at how many people finish the form and how many of those you'd actually want to call, not just total applicants. Cost per qualified applicant is the number that matters here.

The sauce: 20 good-friction questions per job type

Pick one or two per vacancy. Change the details to your real job, because the more specific it is to your situation, the better it works. These are starting points, not a quiz to copy word for word.

1. A truck arrives 2 hours late and the next one is already waiting. What do you do first?
2. Would you rather do a fast, physical shift or a slower one where you work with the scanner and system all day? Why?
3. What's the heaviest or most hectic shift you've done? What made it work, or not?
4. Early shift (6:00) or late shift (14:00)? What fits your life right now?
5. You see a coworker stacking a pallet in a way that isn't safe. What do you do?
6. Which part of this job do you think you'd be fastest at?
7. What do you need from a team lead on your first day?
8. A customer order is missing 3 items and the truck leaves in 20 minutes. Walk us through it.
9. Have you driven a forklift or reach truck? If not, would you want to learn?
10. What made you click on this job?
11. What would make you say no to this job? (hours, travel, pay, something else)
12. How do you get to work? How long can your commute be?
13. Record 30 seconds: why this job, why now? (optional)
14. What did you like most about your last job? What did you like least?
15. Picking accuracy or picking speed: which one are you better at today?
16. If you started Monday, what would you want to know first?
17. Do you prefer working alone with headphones in, or in a team on the floor?
18. What's something you'd improve in how a warehouse usually works?
19. How would your last team lead describe you in one sentence?
20. Is there anything that makes weekend shifts hard for you?

1. A quiet night shift with 3 patients who need a lot of care, or a busy day shift with 10 who need less? Why?
2. A family member is upset and wants answers you can't give yet. What do you say?
3. What's a moment in care work that made you sure this is your job?
4. Which ward or type of care would you pick if you could choose freely?
5. You're behind on your round and a colleague asks for help. What do you do?
6. What do you need from a team to feel good after a hard shift?
7. What made you click on this job?
8. How many hours a week do you want to work, honestly?
9. Which shifts work for you and which ones really don't?
10. What would make you say no to this job?
11. Record 30 seconds: why this job, why now? (optional)
12. What's something your current or last workplace does well that we should copy?
13. What's something you'd never want to go back to?
14. If you started Monday, what's the first thing you'd want to learn about this team?
15. How do you handle a shift where everything goes wrong?
16. Do you want to grow into a specialty? Which one?
17. What does a good handover look like to you?
18. What's a small thing that makes a patient's day better?
19. How would your colleagues describe you in one sentence?
20. How far can you travel for work, and how do you get there?

1. Write the first line of a cold email to [type of customer]. One line only.
2. Would you rather close 10 small deals a month or 1 big one a quarter? Why?
3. What's a deal you lost that taught you something?
4. A customer says "too expensive" in the first minute. What's your next sentence?
5. What do you think is the hardest part of selling [your product]?
6. What made you click on this job?
7. Hunting new customers or growing existing ones: what gives you more energy?
8. What's your current target, roughly, and how close are you this year? (optional)
9. What would make you say no to this job? (commission, territory, travel, something else)
10. Record 30 seconds pitching yourself for this job. (optional)
11. What does a good week look like for you in numbers?
12. What's a tool or habit that makes you better at sales?
13. If you started Monday, who would you want to talk to first?
14. What's something a manager did that made you sell more?
15. What's something a manager did that made you want to leave?
16. Phone, email, LinkedIn or in person: where are you strongest?
17. What do you know about our customers already?
18. What's the best reason someone has ever given you for buying?
19. How do you handle a month where nothing closes?
20. What would your best customer say about you?

1. In 3 sentences: the last bug you were proud of fixing.
2. Would you rather build something new from scratch or make a messy system better? Why?
3. Which part of our stack would you want to work on first?
4. A release breaks something on Friday at 16:00. What do you do?
5. What's a technical decision you'd make differently now?
6. What made you click on this job?
7. Do you prefer clear tickets or vague problems you figure out yourself?
8. What would make you say no to this job? (remote policy, stack, pay, something else)
9. How many meetings a week is too many for you?
10. Link one thing you built that you like. (optional, anything counts)
11. What do you want to learn in the next year?
12. If you started Monday, what would you want to read or see first?
13. What does good code review look like to you?
14. What's a tool you can't work without?
15. Remote, hybrid or office: what actually works for you and why?
16. Tell us about a time you disagreed with a product decision.
17. What's the most useful thing a senior developer ever told you?
18. How do you explain a technical problem to someone who isn't technical?
19. What kind of team makes you do your best work?
20. Record 30 seconds: why this team, why now? (optional)

1. A customer is angry about something that isn't your fault. Write your first reply in 2 sentences.
2. Phone all day or chat and email all day? Why?
3. What's a customer conversation you're still proud of?
4. You don't know the answer and the customer is waiting. What do you say?
5. What made you click on this job?
6. Which shifts work for you and which don't?
7. What would make you say no to this job?
8. What's the most annoying thing a company has ever done to you as a customer?
9. How do you reset after a really tough call?
10. Record 30 seconds: why this job, why now? (optional)
11. What does great service look like to you, in one example?
12. If you started Monday, what would you want to learn first?
13. Do you prefer fixing things fast or explaining them really well?
14. What's something your last team did well that we should copy?
15. How do you stay friendly on call number 60 of the day?
16. Which languages can you help customers in?
17. What tools or systems have you worked with? (fine if none)
18. A customer asks for something against the rules. What do you do?
19. What would your favorite customer say about you?
20. Where do you see yourself in a year: same job, team lead, something else?

And this is how I'd put them in a form. Works the same in your ATS, a landing page form or a Meta lead form.

1. Pick ONE question (two max) that matches the job's hardest or most typical moment.
2. Put it AFTER name and contact details. If someone drops off at the question,
   you still have their details and can follow up.
3. Use a short text field. Add a hint like "2–3 sentences is perfect" so people
   don't think you want an essay.
4. Offer a voice note or a short video as an alternative to typing, if your
   form tool allows it. Mark it optional.
5. Never use it as an automatic knockout. Read the answers yourself, or at least
   look at them before you reject anyone.
6. For every question you add, delete at least one field that isn't needed
   before the first conversation (account, cover letter, retyping the résumé).
7. Time yourself applying on your phone. Over 5 minutes? Cut something.
8. Use the answer in your first call: "You wrote X, tell me more." That's the
   moment candidates notice someone actually read it.
9. Change the question every few months, or per channel, so ready-made answers
   don't spread.
10. Track finished applications AND how many you'd actually call, before and after.

Frequently asked questions

Can't AI just answer the good-friction question too?

Yes, it can. Good friction isn't about making AI useless. A question about a real moment from your job gets a generic answer from AI, and once you've read a few real answers next to it, you see the difference fast. And someone mass applying to 200 jobs won't record 200 voice notes. You're not trying to catch everyone, you're trying to make real interest easy to see.

Won't an extra question lower my number of applicants?

It might a bit, but you're also removing bad friction at the same time, so the form gets shorter overall. Fewer junk applications isn't a loss. Watch finished applications and qualified applicants together, not just the total.

Should I use AI detection tools to filter out AI-written applications?

I wouldn't make that your main filter. Lots of real, good candidates use AI to polish what they write, and Greenhouse found 60% of U.S. job seekers have changed or thought about changing how they write so they don't get mistaken for AI. Ask a better question instead of trying to guess who typed the answer.

How many good-friction questions should one form have?

One. Two if the job really needs it. The moment it feels like a test, you're back to bad friction.

Sources

• Semafor, "Recruiters swamped with AI-generated job applications" (June 24, 2025): summary of The New York Times reporting, LinkedIn applications up 45% to about 11,000 per minute.
• eWeek, "Job Seekers – Some Using AI – Flood LinkedIn With 11,000 Applications a Minute" (June 25, 2025): same New York Times figures and context.
• Greenhouse, AI in Hiring survey press release (November 19, 2025): about 3 in 4 U.S. job seekers use AI, 49% sending more applications, 34% of recruiters spending up to half their week on spam applications.
• Fortune, interview with Greenhouse CEO Daniel Chait (July 27, 2026): auto-apply tools for about $20, around 254 applicants per job, the "AI doom loop," the "dream job" flag and its claimed hiring rate.
• Greenhouse, job seeker survey press release (October 20, 2026): 77% of U.S. job seekers used AI in their search, only 5% would let an AI agent apply on its own, 60% changed or considered changing how they write.