How to Raise AI Interview Completion Rates

48% of recruiters on Equip report that most candidate drop off in an AI interview happens before the first question, not in the middle of it.

To raise completion, find which of three points is losing candidates, then adjust the setting behind it: the deadline, the restart rule, device support, the invite or the monitoring. The six fixes below are in the order most teams should try them.

Where candidate drop off happens

"Completion rate" hides three different problems. Split it into stages before changing anything, because each stage has a different cause.

Stage How to measure it Usual causes
Invited, never started Starts divided by invites The email was not delivered or opened, the deadline was too short, or the invite gave no reason to bother
Opened, blocked at setup Candidates who clicked the link but have no attempt Unsupported browser, camera or microphone permissions, a desktop-only rule, or a monitoring step they did not expect
Started, never submitted Submissions divided by starts A dropped connection with no way back in, an interruption, or an interview that ran longer than the invite promised

If most of your loss sits in the first row, work on the invite and the deadline. If it sits in the second or third, work on devices and the restart rule. Teams that skip this step usually shorten the interview, which addresses the smallest of the three.

What research says about why candidates quit

Two findings are worth knowing before you change your setup.

People who quit, quit early. A study of 222,772 job seekers across 69 selection systems, published in the Journal of Applied Psychology (Hardy et al., 2017), found that most applicants who abandoned an online assessment did so in the first 20 minutes, and that longer assessments did not have higher quit rates. That study covered assessments, not interviews, but the pattern points the same way: the opening minutes decide completion, more than total length does.

The format changes who continues. In a 2026 working paper from Monash University and the University of Exeter (Avery et al.), more than 3,000 real applicants were randomly assigned to different next steps. Of those asked only to confirm interest, 85% continued. A live online interview kept 68%. A recorded one-way interview kept 40%, and the candidates who left included the most qualified. A follow-up experiment traced the gap to two beliefs: that a recorded interview is sent to everyone, and that it is a less fair way to be judged. Fear of the technology played no measurable role.

The second study tested recorded one-way interviews, not conversational ones, so format is one of the first things to compare in your own data.

AI interview completion rate by format, on Equip
73%
One-way AI interviews
88%
Conversational AI interviews
Share of invited candidates who submitted the interview. Aggregate, anonymised platform data, October 2025 to October 2026.

Use these as a reference point, not a target. Role, seniority and where the interview sits in your funnel all move the number. If you are still deciding between the two, the trade-offs are covered in one-way vs conversational AI interviews.

Six ways to raise AI interview completion rates

1. Set a deadline that includes an evening and a weekend

Candidates with jobs take screening interviews outside working hours. A window that opens on Monday morning and closes on Wednesday evening excludes the two times they are most likely to be free.

  • Keep the window open for at least five days, including one full weekend.
  • State the closing date and time in the invite, next to how long the interview takes.
  • Don't set an opening time unless you need one. A candidate who clicks straight away and is told to come back later often doesn't.

Most platforms give you two separate controls here: a start and end time for the interview itself, and an expiry on the invitation link. On Equip these are the availability window and the link expiry. Check that the two agree, so a link never outlives the window or closes before it.

2. Decide your restart policy before you send invites

A candidate on mobile data loses the connection eight minutes in. What happens next is a setting, and it is usually chosen by default.

Restarts allowed
A candidate who leaves before submitting can begin again. Only the submitted attempt is scored.
Fits: first-round screens, high-volume roles, candidates on phones or unreliable networks.
Restarts not allowed
A candidate who leaves mid-interview cannot return unless you reset the attempt for them.
Fits: later rounds and roles where seeing the questions once, then starting over, would be an unfair advantage.

On Equip's conversational interviews, you can turn restarts on in the interview settings. Brief network drops reconnect on their own either way. The setting matters for the candidate who closes the tab or loses the connection for longer.

Whichever you choose, say so in the invite. "You can restart if you get disconnected" removes a reason to put the interview off. "Take this in one sitting, on a stable connection" prevents a preventable loss.

3. Make sure the interview can be finished on a phone

Many candidates open the invite on a phone, and for some it is the only device they have. Three things commonly stop them at setup:

  • A desktop-only rule you didn't need. Coding tasks and some monitoring options require a computer. If your interview has neither, check that it isn't restricted to desktop anyway.
  • In-app browsers. A link tapped inside LinkedIn, Instagram or a messaging app often opens in a browser that can't use the camera or microphone. Tell candidates to open the link in Chrome or Safari.
  • Permissions. Camera and microphone prompts are easy to dismiss by accident on a small screen, and hard to find again.

The fastest test is to take your own interview on a phone, on mobile data, from the invite email. Anything that stops you will stop a candidate.

4. Tell candidates they were selected, and who reviews the result

As the field experiment above showed, candidates leave when they believe everyone got the same link and nobody will look at the result. Both beliefs can be corrected in two sentences of the invite, as long as the sentences are true.

  • Say that the interview follows a shortlist. "You are one of 40 applicants invited to this stage" reads very differently from a generic next-step email.
  • Name who reviews it. State that a recruiter reads every report and makes the decision.
  • Give a date for the outcome. Candidates finish more readily when they know what follows.

For the questions candidates raise after reading the invite, see how to handle candidate objections to AI interviews.

5. Keep only the monitoring the role needs

Every monitoring check adds a step before the first question: a permission, a photo, a second device to pair, a full-screen prompt. Each one is a place to stop, and some checks only work on a computer.

Match the monitoring to the stage. A first screen for a high-volume role rarely needs the same setup as a final technical round. The trade-offs for each check are set out in AI interview cheating detection. Whatever you keep, list it in the invite so the consent screen is not a surprise.

6. Ask for audio when you don't need video

In the same field experiment, recorded video interviews had completion six percentage points lower than audio-only ones with the same questions. If you are assessing what a candidate says, not how they present on camera, audio is enough.

Many tools let you choose. On Equip, a conversational interview can run with the camera requirement switched off, which makes it audio-only. The exception is any role where you rely on camera-based monitoring, since that needs video.

What to do about candidates who never start

Some no-shows are a delivery problem, and some are a decision. Treat them in that order.

  1. Check delivery first. A bounced or unopened email is not a candidate who declined. Confirm the address and resend.
  2. Follow up personally, once. An automated reminder is unlikely to be enough: in the same experiment, one sent five to seven days after the invite made no significant difference to completion. A short note from the recruiter that asks whether something is in the way does a different job: it surfaces a device problem or a scheduling conflict you can fix.
  3. Close out after the deadline. Where your tool charges per invitation, recover the cost of invites that were never used. On Equip, credits for an invited candidate who has not started can be reclaimed once the invitation is more than 24 hours old. Reclaiming cancels that candidate's link, so do it after the window closes, not while they still have time.

Reclaimed credits fund the next round of invites, which keeps a low start rate from turning into a cost problem while you fix the cause.

Frequently asked questions

What is a good AI interview completion rate?

There is no universal figure. Completion depends on the role, the stage and how selective the invite list was. Compare each interview with your own earlier ones in the same format and for similar roles, and track start rate and finish rate separately.

Does drop off just filter out candidates who were not serious?

Not reliably. In the field experiment cited earlier, completion among applicants with the most experience and the strongest CVs fell about as far as it did for everyone else. Some loss is healthy self-selection, but you cannot assume the people leaving are the ones you would have rejected.

Should you give a second chance to someone who dropped out midway?

If the cause was technical, yes. Ask what happened, reset the attempt, and send a new invite with a fresh deadline. If the same failure appears for several candidates, fix the setup before inviting anyone else.

For how the interview itself runs once a candidate is in, read how AI interviews work.

Take an AI interview on your phone.
Set the availability window and restart policy, invite yourself, and go through every screen your candidates will see. Equip starts at $1 per candidate.
Start for free See AI interviews

Read next