AI Interview Rules Compared: Retakes, Practice and Recordings

The six platforms on this page give four different answers on second attempts. Mercor publishes a limit: up to three retakes per interview. micro1 publishes no retake rule. Alignerr offers no retake for a better result, and its help center contradicts itself on technical failures. On HireVue, Spark Hire and Willo, the employer sets the attempts for each question. The gap is as wide on your recording: Alignerr’s terms let it use your interview to train AI, micro1 reserves the right to train its own models on de-identified interview data, and Willo says it does not use candidate data to train AI models.

Every cell below comes from our page on that platform, and each of those pages was built from the platform’s own public documentation. Five were checked on 2026-10-06. The HireVue page was checked on 2026-09-23, and its row on recording attempts comes from HireVue’s 2026 Explainability Statement, checked on 2026-10-06. Offermic is not affiliated with any of the six companies. The platform name in each row links to the full page, which has the detail and the sources.

Three labels recur in the tables:

• Not published: the platform’s public documentation, as checked for our page, does not state it.
• Employer setting: the company that invited you chooses. Look in your invitation or on the question screen.
• Not checked: we have not verified this point for that platform, so the table makes no claim either way.

Format, length and practice mode

The first three rows are conversations with an AI interviewer. The last three are mainly one-way recordings: you answer preset questions with nobody on the other end.

Platform	Format	Published length	Practice mode
Mercor	Video and audio conversation with an AI interviewer	About 20 minutes for most interviews	Free Practice Interview, unlimited, not shared with companies
micro1	Live spoken conversation with Zara, an AI recruiter	20 to 40 minutes, about 7 minutes per skill	Free Interview Prep tool: 18-minute practice interviews by role, with feedback
Alignerr	Recorded video conversation with Zara, an AI interviewer. Only some jobs require it	Not published. A page Alignerr has removed said 15 minutes	Not checked
HireVue	Most often a one-way recording. Some employers add games or skills tests	Employer setting. HireVue’s practice example: 30 seconds to prepare, up to 3 minutes per answer	Practice questions that are not part of your submission
Spark Hire	One-way recording	Employer setting: 30 seconds to 5 minutes per answer	A practice video before the first question, not saved or shared with the company
Willo	One-way. Answers by video, audio, text, file upload or multiple choice	Employer setting: 1 to 5 minutes per video or audio answer. 12 hours to submit after your first answer	A sample practice question. Nothing is saved, and you can repeat it

micro1 and Alignerr both call their interviewer Zara. Neither company’s documentation mentions the other, so don’t carry one platform’s rules over to the other. The 20 to 40 minutes is micro1’s figure, and a second micro1 guide describes a short interview of about 20 minutes.

Spark Hire’s own pages disagree on the answer ceiling: its two settings articles say 5 minutes, and two other pages say 3. Willo fixes no overall length. It advises employers to ask 3 to 5 questions and tells them candidates take 20 minutes on average. Mercor does not say whether its Practice Interview is scored.

Second attempts and rules on AI tools

Platform	Second attempts	Who decides	AI tools during the interview
Mercor	Up to 3 retakes per interview, counted across every application that uses it	Mercor sets the limit. You start a retake yourself	Do not use LLMs to compose answers. Grammar and wording checks are allowed
micro1	Not published: no retake limit and no waiting period	Not published	Answers must be your original work, without third-party assistance unless explicitly allowed
Alignerr	None for a better result. After a technical failure, its help center contradicts itself	Alignerr. One help article says to request it from support	AI tools, external resources and outside assistance are strictly prohibited
HireVue	Recording attempts per question: 1 to 5, or unlimited. HireVue recommends three in total	Employer setting	Not checked
Spark Hire	Takes per question: 1 to 5, or unlimited. A think-time countdown always means one take	Employer setting	The consent step discourages AI use. An AI-content flag can label an answer “Possible misconduct”
Willo	Retakes per video or audio question. The number can be zero	Employer setting	Candidate rule: not checked. Real Talk labels answers that sound scripted or AI-generated

Mercor. A retake replaces your earlier attempt, according to the Mercor page that covers interviews by name. One line elsewhere in its docs points the other way, and our Mercor page sets out both.

HireVue. The range comes from HireVue’s 2026 Explainability Statement. Read the instructions on your invitation for what your employer chose.

Spark Hire and Willo. Both show the allowance on the question screen. On Spark Hire a submitted answer can’t be redone, even if takes remained. Willo’s Reveal and Respond questions give one attempt. Neither platform documents a way to redo a whole interview after you submit it.

The AI flags. Spark Hire’s flag is on by default for companies that use AI Video Review, and Spark Hire says it does not block submissions or change scores. Willo says Real Talk is automatically enabled, marks its AI features as available only on selected plans, and calls the labels “signals, not verdicts”.

Paper notes. None of our six pages records a platform rule on them. The nearest facts: Mercor’s setup list includes pen and paper, and Alignerr’s examples of banned help include separate monitors with help resources.

AI scoring, your recording and your result

Platform	What it says about AI scoring	AI training on your recording	Result or feedback
Mercor	The AI interviewer evaluates your performance for the listing. Pass score: not published	The interview FAQ says your interview data won’t be used to train AI models. The data policy lists training Mercor’s internal recommendation models	Usually 2 to 4 weeks to hear. Support gives no scores or individual feedback
micro1	Pass mark: not published. For a position, a human evaluator reviews the automated assessment before a final decision	Reserves the right to train its own models on de-identified interview data. Links an opt-out form	Real-time result. Feedback on request if you are not certified
Alignerr	Pass mark and rubric: not published. A page Alignerr has removed said AI tools and people score interviews together	Its terms allow using the interview to train or improve AI	Its FAQ says no individual results or feedback
HireVue	Depends on the employer’s setup. Scoring models: not published. No facial analysis since 2020	Not checked	From the employer, on its own timeline
Spark Hire	AI Video Review is an employer setting, off by default. It rates English answers from the transcript	Not checked	From the employer. Questions about feedback and timing go to the company that invited you
Willo	On selected plans, Insights scores your transcript against the employer’s criteria. Insights does not reject automatically	Says it does not use candidate data to train AI models	From the employer. Accept or Reject sends you nothing by default

Mercor. The data policy rules out training external AI models and third-party LLMs. Human reviewers may review assessments, and Mercor says project teams make the final decisions.

micro1. Its API documentation shows the report an employer using the platform receives: a rating per skill, a soft-skill rating and a proctoring score. micro1 doesn’t say its own certification uses the same fields, or whether a person reviews the real-time result. It says it does not share your data with AI model companies to train theirs.

Alignerr. The terms go beyond training: they grant a perpetual, irrevocable right to use the interview and your name, audio, video and likeness for any purpose. The answer about results sits under the FAQ’s Human Interviews heading, and Alignerr doesn’t say separately that it covers the Zara interview.

HireVue. HireVue says it never uses the video recording to identify you, and that people make the hiring decisions.

Spark Hire. Spark Hire says final hiring decisions are always made by humans. Neither of its help centers describes a way for a candidate to see the AI rating or opt out. A request to delete your recordings is forwarded to each employer, which decides.

Willo. A separate feature, Benchmark, can reject automatically on multiple-choice scores. Willo says Insights uses no facial recognition, voice pattern scoring or biometric data. The employer is the data controller.

What follows from the comparison

This section is our view.

Prepare for one attempt on every platform. Mercor is the only one of the six that promises every candidate a second attempt, and a retake there replaces the earlier one. micro1 promises nothing, Alignerr offers nothing for a better result, and an employer on HireVue, Spark Hire or Willo can allow a single attempt. On Spark Hire and Willo you may only learn the allowance when the question appears.

Check four things before you start.

1. The attempts allowed. Look in your invitation, then on the first question screen.
2. Whether a countdown is involved. On Spark Hire, limited think time means one take, and recording starts by itself when the countdown ends. Willo’s Reveal and Respond works the same way.
3. The terms on your recording, if AI training matters to you. On Alignerr you accept them at sign-up. micro1’s privacy policy links an opt-out form.
4. Whether AI review is switched on. On HireVue, Spark Hire and Willo that is the employer’s choice, so ask the employer.

Use the practice mode. Five of the six pages document one. Run it on the device, browser and microphone you’ll use for the real interview.

Say it in your own words. Mercor, micro1 and Alignerr each have a written rule restricting outside help, and Spark Hire and Willo both label answers that sound scripted or AI-generated. Practice out loud beforehand with a friend, a phone recording or a mock interview app (Offermic, the app we make, is one), and use tools like that before the interview, never during it. Our guide on how to pass an AI interview covers what carries across platforms.

Don’t count on feedback. micro1 documents feedback on request, and Mercor documents it for some assessments only. On the other four, expect none from the platform.

These rules will change

The dates near the top are when each platform’s documentation was last checked. Help centers get edited without notice: Alignerr removed its interviews page, the one that gave a 15-minute length, between April and May 2026. If your invitation or question screen says something different from a cell here, go by what is in front of you.