What AI interview help is allowed?
Help that runs before the call or after it is allowed by every employer policy quoted on this page. Research, practice questions, mock interviews, resume polishing and reviewing a recording afterwards all sit there. Help that runs during the live interview is banned by all three policies unless the interviewer gives permission first.
The framework comes from the page on whether using AI in an interview is cheating. This page applies it to the kinds of tools you can buy or use for free. Tool names are examples, quoted from their own sites on October 22, 2026. Prept makes one of the products in the after column, and its row lists its limits.
What do employer policies say about AI before, during and after?
They agree on timing. Anthropic, American Express and Greenhouse all welcome AI for preparation, all ask that the live interview be yours alone unless someone says otherwise, and all warn against using AI to claim experience you don't have. Their wording differs, so read the policy of the company you're interviewing with before you pick a tool.
Anthropic's candidate guidance, last updated July 10, 2025, says to "use Claude to research Anthropic, practice your answers, and prepare questions for us." For live interviews it asks for "no AI assistance unless we indicate otherwise." For applications it says to "create your first draft yourself, then use Claude to refine it."
American Express's AI guidance for candidates says AI "can be used to practice frequently asked interview questions and interview tips." It bars AI use "during a virtual or in person interview unless explicitly authorized by the interviewer." It also allows AI to refine work "that you fully understand and can independently reproduce and explain."
Greenhouse's guidelines encourage preparation with AI and ask candidates not to "use AI to supplement your performance in a live interview unless otherwise instructed or agreed upon." Greenhouse says AI used outside its guidelines "without disclosure or citation" may get a candidacy disqualified. More published rules are collected on the page about employer AI interview policies.
Which kinds of AI tools sit on which side of the line?
Sort tools by when they run and whose words come out of them. Chatbots, practice platforms and resume tools run before the call, and live answer assistants run during it. Transcription and debrief tools run after it, from a recording made with consent. The table sets each category against the three policies above, with examples quoted from their own sites.
| Category | When it runs | Example, in its own words (checked October 22, 2026) | Allowed by the three policies | Where it helps |
|---|---|---|---|---|
| General chatbot for research and practice | Before | Anthropic's guidance tells candidates to use Claude to "practice your answers" | Yes | Company research, practice questions, clearer wording |
| Mock interview platform | Before | Big Interview: "All-in-one, AI-powered interview platform to prep, practice, and get hired." | Yes | Rehearsing out loud with feedback on delivery |
| AI resume writer or tailor | Before | No single tool named | Yes, if every claim is yours | Formatting, and matching a posting's wording to facts you have |
| Live answer assistant | During | Final Round AI: "an AI-powered real-time interview assistant on your screen" | No, unless the interviewer authorizes it | Practice sessions where nobody is assessing you |
| Recording and transcription | Recorded during, read after | No single tool named | Not addressed by the three policies; needs everyone's consent | Your exact words, for review |
| Post-call debrief | After | Prept: reads a recorded round for structure, evidence and numbers | Not addressed; the policies cover the live call | Finding the one answer to fix |
Where are practice and prep tools useful?
They're useful for the parts of an interview you can rehearse: the common questions, the length of your answers, and how you sound. A mock interview platform gives you repetitions without booking a friend. Big Interview's homepage, checked on October 22, 2026, offers "instant AI insights on delivery, clarity, and more," feedback a mirror can't give you.
The limit is content. A practice tool can tell you an answer ran long, and it has no way to know whether the number in it is true. That check is yours. It matters most on the resume, and the page on why AI resume tools invent numbers shows how a made-up figure gets in.
Why are live answer tools on the wrong side of the line?
They produce words for you while the interviewer is assessing you, which is the one use all three policies ban without permission. Final Round AI's Interview CoPilot page, checked on October 22, 2026, says the tool "hears the question, matches it against your Goal, and writes an answer you can say out loud on your display only."
The same page sells a Stealth Mode it says is "designed to keep the assistant hidden from screen shares and recordings." A feature built to stay hidden from the other person exists for help they didn't agree to. That is the problem, more than the AI.
To be fair to the product, Final Round AI also advertises practice and review, with the line "Practice before. Debrief after." Those features sit on the allowed side, and someone who wants mock interviews and a review in one app may find them useful. The live mode is the part to leave alone. What happens to candidates found using one is on the page about being caught using an AI copilot.
Is recording the call for a later debrief allowed?
Recording gives you no answers during the call, so the three policies above don't cover it. It raises a different rule instead: consent. Ask at the start of the call, in one sentence, and accept a no. Laws on recording vary by place, and the guide to whether it's legal to record a job interview covers the main ones.
The page on how to ask to record your interview has the exact sentence to say. If the interviewer says no, you still have your memory and the next hour. The free 10-minute debrief template gets the questions and the numbers you said onto one page before they fade.
What does Prept do, and where does it stop?
Prept is a post-call tool. It transcribes a round you recorded with consent and reads each answer for structure, evidence and numbers, trade-offs and the questions you asked. Its line is analysis after the call, never during it. It gives no live prompts of any kind, so it can't help you while the interview is happening.
It has other limits. It can't confirm that everyone on the call agreed to be recorded; it asks you to confirm each time, and the asking is your job. It reads what was said, so it can't see the interviewer's scorecard or tell you why they chose someone else. The audio goes to Google's paid Gemini API, whose terms say it isn't used for training, and the audio is deleted once the transcript is written.
On the resume side, every claim on a tailored resume is traced to a fact you confirmed, and a number you can't stand behind is left off rather than invented. The free tier covers one of each, including one interview hour, with no card. If that fits how you work, you can get your first debrief free, and the resume optimizer covers the resume side.
What about AI-written resumes and cover letters?
They sit on the allowed side as long as the content is yours. Anthropic asks for your own first draft, then refinement. American Express allows AI for "enhancing the formatting and structure of the resume" and rules out content that "embellishes responsibilities and contributions without personal input." Greenhouse says application materials "must represent your professional experience, qualifications and skills accurately."
The test is whether you could defend each line if an interviewer asked about it. Recruiters apply the same test, and the page on whether recruiters can tell AI wrote your resume covers what they check.
How do you check a tool before you use it?
Read the tool's own website and answer three questions. When does it run: before, during or after the interview? Do the words it produces come from you, or are they written for you to say? Would you be happy to tell the interviewer you used it, in those words, if they asked you directly at the end of the call?
If the answer to the last one is no, that settles it. The page on whether to tell the interviewer you used AI has the one sentence to say when they ask. The guide to AI interview tools goes through the main products one by one, with what each does and when it runs.