When async video first started showing up in hiring workflows, most recruiters treated it as a novelty — something for high-volume hourly roles, maybe, or for companies too small to staff a proper phone screen operation. That framing has not aged well. We now routinely see recruiting teams at Series B software companies using async video as their primary first-contact screen for engineering, product, and sales roles. The question has shifted from "should we try this?" to "how do we do it well?"
The short answer is that async video screening done well requires more careful design than a live phone call — specifically in how you write prompts, how you frame the process for candidates, and how you score responses without the conversational cues that a live screen provides. Get those three things right and you end up with more consistent data than a phone screen at roughly 40% of the recruiter time per candidate.
Writing Prompts That Produce Useful Responses
The most common failure mode in async video screening is the vague prompt. "Tell us about yourself" and "Why are you interested in this role?" produce answers that are almost impossible to score comparably across candidates. They also fail to elicit the operational detail that distinguishes candidates with real experience from candidates who describe experience well.
Useful async prompts share three properties. They are specific to the role, they ask for evidence not opinion, and they have a natural length ceiling — not by imposing a timer, but by asking a question narrow enough that a two-minute answer covers it adequately.
Compare these two prompts for a growth marketing role:
- Weak: "Tell us about your experience in B2B marketing."
- Strong: "Walk us through a specific campaign you ran that produced a measurable outcome. What was the hypothesis, what did you ship, and what did the results show?"
The second prompt has a built-in structure. A candidate with real experience will fill it with detail — they will name the channel, approximate the result, and explain what they concluded. A candidate inflating their experience will describe a campaign in general terms without the specifics that come from having actually run one. That gap is visible and scorable.
We recommend three to five prompts per role. Fewer and you have too little signal to compare candidates. More and completion rates drop — our data from design-partner hiring flows shows that flows with more than five questions have candidate dropout rates roughly 18 percentage points higher than flows with three questions, even when total time is similar. Candidates make the decision to abandon early, usually after the third or fourth question if they have not yet committed to finishing.
Setting Candidate Expectations Before the Flow Starts
Async video has a psychological barrier that live phone screens do not: the sense of speaking into a void. On a phone screen, a recruiter reacts, adapts, follows up. On an async video flow, the candidate presses record and talks at a camera. If they have not been told what to expect, the experience can feel cold and mechanized.
The framing you send before the link matters more than most teams realize. We have seen candidate NPS scores shift by more than 20 points between flows that sent a brief two-sentence explanation and flows that included a 90-second "what to expect" paragraph from the recruiter. The paragraph does not need to be long. It needs to answer four questions the candidate is already asking:
- Why are we using async video instead of a call?
- Will a real person watch this?
- How long should each answer be?
- When will I hear back?
Answer those four questions and the completion rate goes up, the response quality goes up, and the candidate experience survey results improve. Skip them and you are asking candidates to trust a process they do not understand.
On the "will a real person watch this?" point — be honest. If a human recruiter is watching every video before advancing a candidate, say so. If an AI is doing the first pass and a human reviews flagged responses, say that too. Candidates can accept both answers. What they do not accept well is discovering the reality was different from what they assumed.
Scoring Responses Without Conversational Cues
The biggest operational challenge in async video is scoring consistency. On a live call, an interviewer can probe — "can you say more about that?" — and gather additional signal when an answer is ambiguous. On an async video, the recording is the complete artifact. Interviewers who are used to live conversations often score async responses too harshly when a candidate did not elaborate, without recognizing that the prompt did not invite elaboration.
A good scoring rubric for async video responses should be anchored to what the prompt explicitly asked for, not to what an ideal live response might have included. If your prompt did not ask for a quantified result, do not penalize a candidate for not providing one.
We use a three-level scoring structure for async responses at Proofglint: strong signal, neutral signal, weak signal. Strong signal means the response directly addresses the prompt with specific evidence. Neutral means the response addresses the prompt but lacks specificity or supporting detail. Weak means the response is off-topic, describes general tendencies rather than specific instances, or leaves a required component unanswered.
Each level should have written examples tied to the specific prompt. Generic rubric language like "demonstrates clear thinking" is not scorable — "cites a specific outcome with a number or named project" is. When you build rubrics this way, two different reviewers watching the same video within the same week should land within one level of each other on most responses. If they are not, the rubric needs tighter anchors, not better reviewers.
Video Versus Text: When Each Format Fits
Async screening does not require video. For many roles and question types, text responses are more appropriate — and often produce better signal.
Video works well when communication style is a core job criterion (client-facing roles, team leadership, sales), when you want to assess how a candidate explains technical concepts in real time, and when the company culture places a premium on presence and articulation. These are legitimate criteria and video lets candidates demonstrate them directly.
Text works better when the question asks for structured analysis ("walk us through how you'd prioritize these three features"), when written communication is the actual job requirement, and when you are screening in a time zone spread where recording a video adds friction candidates are not expecting. Text responses are also easier to review at volume — you can read faster than you can watch.
A hybrid flow works for many roles: one or two video questions on criteria that benefit from seeing how someone speaks, two or three text questions for analysis and judgment. We have seen teams run effective 15-minute hybrid flows that replaced 45-minute phone screens without candidates feeling like the process got shorter. The completeness of the data was what changed — not the time the candidate spent.
What the Numbers Say About Completion and Quality
In our experience across the design-partner companies we have worked with closely, async flows that are well-designed — specific prompts, clear expectations sent beforehand, three to five questions — see completion rates between 68% and 77%. That number surprises some teams; they expected higher dropout. The comparison point matters: the industry median completion rate for phone screen slots that get calendared but not kept is roughly 35% — one in three scheduled screens does not happen.
Async video also surfaces candidates who perform worse on live calls but produce strong signal in recorded responses. This includes candidates who need a moment to collect their thoughts before answering, candidates for whom English is a second language and who benefit from the lack of time pressure, and candidates whose roles required deep solo work rather than real-time verbal fluency. Those candidates are systematically disadvantaged by phone screens and systematically better served by async flows designed fairly.
None of this means async video is a universal replacement for live conversations. There are questions that require back-and-forth, candidates who perform better when they can sense the other person's reactions, and decisions where a hiring manager wants a gut-level read that async simply cannot provide. The honest framing is that async video replaces a specific type of phone screen — the qualifying screen that answers basic availability, communication, and fit questions — not all live conversations. Done well, it means the live conversations you do have are deeper because the qualifying work already happened.
