Greenhouse just shipped a small Voice AI change that a lot of recruiters asked for. Each candidate invite to an AI voice interview can now have its own deadline. Before this, every candidate in an interview worked against one shared end date (Greenhouse).
That sounds like a settings tweak. It is really a question about fairness and drop-off. How long a person gets to finish a screen changes who finishes it. So if you use Voice AI, or any async AI screen, this is a good week to decide your deadline rules on purpose.
What changed
Greenhouse now gives you two options for each Voice AI interview:
- Global end date. One expiry date for the whole interview, the same as before.
- Per-invite expiry. Each invite gets its own timeline, counted from when that candidate was invited.
A few details matter. Existing interviews stay on a global end date. If you switch an interview to per-invite, people you already invited keep their old expiry. Only new invites use the new rule.
Greenhouse explained the problem in its launch post on LinkedIn: "Your req closing Friday and your req open all quarter no longer have to share a deadline."
The same update added a few other things worth knowing:
- Candidates can speed up or slow down the AI interviewer during the call.
- A stronger connection check runs before the interview starts.
- If a candidate is asked a question but gives too little to score, it now scores 0 instead of N/A. Questions the interview never reached stay N/A.
- Greenhouse Recruiting customers get a Voice AI interview status field they can filter on.
Why the deadline matters more than it looks
With one global date, the person invited on day one gets two weeks. The person invited on day twelve gets two days. Same role, same screen, very different chance to show up well.
The late invite is often someone you found later. A referral. A sourced candidate. Someone who applied after a repost. These are not weaker candidates. They just arrived later in your process. A shared deadline quietly punishes them for timing.
Per-invite expiry fixes that. Everyone gets the same window, no matter when they entered the pipeline.
How to pick a window
There is no magic number, and Greenhouse does not prescribe one. Here is a simple way to think about it.
Start with the candidate's week. Most people doing a screen already have a job. They will do it at night or on a weekend. A window that does not include at least one weekend is tight for them.
Then look at the role's clock. If the hiring manager wants a shortlist by Friday, a long window will not help. In that case, a global end date may still be the right call. Just tell candidates the real date in the invite, in plain words.
Then write it down. Pick a default for each type of role. For example (hypothetical):
| Role type | Expiry rule | Reason |
|---|---|---|
| High-volume hourly | Per-invite, short window | Fast hiring, many applicants |
| Salaried, open for weeks | Per-invite, include a weekend | People screen outside work hours |
| Campaign or event hiring | Global end date | Everyone is invited at once |
The numbers are yours to set. The point is that the rule should be the same for everyone in that row, and someone should be able to explain why.
Watch the new zero score
The scoring change is fair on paper. If someone says "I don't have an example for that," a 0 tells the truth better than a blank.
But it raises the stakes for the questions you write. A vague question gets vague answers, and vague answers can now score 0. Before you launch a screen, read each question out loud and ask:
- Could a strong person in this role answer it from real work?
- Does it need context they won't have yet, like your tools or your team?
- Is it really two questions in one?
Greenhouse also added "closed questions." These are short factual checks, like a work authorization date, that skip follow-ups. Use those for yes or no facts. Keep the open questions for things that really show skill.
Use the pace control as a signal, not a score
Candidates can now slow the AI interviewer down. Some will. People who are nervous, who are not native English speakers, or who just think before they talk may want more time.
Make sure nobody on your team reads that as a weakness. It should not show up in anyone's notes. If your scorecard has a line for "communication," define it in terms of what was said, not how fast.
Track who drops off
The new status field in Greenhouse Recruiting lets you filter by interview status. Use it.
Once a month, look at three groups:
- People who were invited and finished.
- People who were invited and never started.
- People who started and did not finish.
If group 2 or 3 grows after you change a deadline, that tells you something. Talk to a few of them if you can. Ask what got in the way. Was it the deadline, the tech, or the format?
Also check who opts out of the AI screen. Greenhouse lets you segment by opt-out too. Make sure people who opt out have a real human path that moves at the same speed. If opting out means waiting two extra weeks, it is not really a choice.
A quick checklist before your next screen
- Pick global or per-invite expiry for each live Voice AI interview, and note why.
- Put the actual deadline in the candidate invite.
- Reread every question with the new 0 score in mind.
- Move yes or no facts into closed questions.
- Tell your team that pace changes are not a signal.
- Set a monthly check on finish rates and opt-outs.
- Check your local rules on AI in hiring. Some places require notice or a human option.
The bigger point
AI screens save time when they are set up with care. Most of that care is not in the model. It is in boring choices like deadlines, question wording, and what happens when someone says no thanks.
Greenhouse gave you more control this month. The useful part is deciding how you will use it, writing that down, and checking it again in a few weeks.



