AI Interview Rules Are Vague. Ask These 5 Questions, One Per Stage

Say you have a take-home due Friday and a live loop the week after, and nobody has told you whether Claude is allowed in either one. On Oct. 13, CNBC Make It reported Greenhouse CEO Daniel Chait's advice: ask about the rules for AI use both during the interview process and on the job, because those rules are "often not published." Asking "Is AI okay?" once won't settle it. You need a separate question at each stage.

Why are employer AI policies for candidates so vague?

Most employers don't publish a candidate AI policy, and candidates who have been through AI evaluation say so. Greenhouse's 2026 candidate report found that 80% of candidates with relevant experience (people who had already been through AI evaluation) describe employer AI rules as vague.

That 80% covers a subset, not every job seeker. Greenhouse surveyed 2,950 candidates, 1,200 of them in the US, with the rest in the UK, Germany, Australia and Ireland. Greenhouse's press release is dated May 1, 2026. The disclosure numbers in it are worse than the headline:

• 70% were never clearly told upfront that AI would be evaluating them.
• 21% (about one in five) only found out once the interview had started.
• 18% of US candidates said employers had clear AI policies, according to Greenhouse's release.
• 38% have pulled out of a hiring process at least once because it included an AI interview, and another 12% would if required.

In practice, if the rule isn't written down where you can see it, you have to ask for it.

Your recruiter probably can't answer the question either

The first person you talk to usually didn't write the AI policy and may not know it. In Refolk's index of professional profiles, about 123,000 people in the US hold Recruiter, Technical Recruiter or Talent Acquisition Partner titles. Only 495 of them mention "AI" in their headline.

A headline keyword is a rough signal. Some of those 495 recruit for AI companies (FieldAI and Inductive Automation are among the top employers in that set) and aren't describing their own tool use. Even so, the direction is clear. The people running your screen mostly aren't the people deciding whether Cursor is allowed in the coding round.

This changes how you ask. If you ask out loud on a call, you'll get a guess. If you ask in writing, the recruiter has to forward it to someone who can commit to an answer, and you end up with a record.

"AI allowed" almost always means one named round

When a company does allow AI, it usually means one specific round, not the whole process. A yes on the take-home tells you nothing about the live loop.

Look at the companies that have published something:

• Canva posted "Yes, you can use AI in our interviews" on its engineering blog in June 2025. It expects backend, machine learning and frontend engineering candidates to use tools like Copilot, Cursor and Claude in technical interviews. That covers engineering roles, not every role.
• Meta pairs its AI-enabled coding round with at least one traditional, AI-free coding session.
• Anthropic lets applicants refine resumes and cover letters with Claude after writing their own first draft. Its line for live interviews is "This is all you - no AI assistance unless we indicate otherwise."

A third-party tally from Ophy AI, which read the official candidate AI guidance of 32 employers on Sept. 29, 2026, found that 24 restrict AI during live interviews and 10 describe at least one round or task where AI is allowed or expected. 24 plus 10 is more than 32, so some employers do both: AI in one round, banned in another.

The reasoning makes sense once you see it. A take-home is meant to look like the real job. CNBC put it this way: recruiters may frown on real-time AI in a screen "but they might encourage candidates to use AI to complete a take-home assessment to mimic doing the actual job." The live round checks the part you can't hand off, which is your own reasoning. Each stage tests something different, so each stage has its own rule.

Skipping AI costs little. Using it wrongly can cost the offer.

When the rules aren't clear, skipping AI is the safe choice. Not using a tool you were allowed to use costs you almost nothing. Using one you weren't allowed to use can end your candidacy.

Amazon is the clearest example. Its guidance, as reported by Business Insider, says failing to comply "may result in disqualification." It doesn't say you lose points. It says you're out.

Candidates and employers also disagree about where the line is. In Greenhouse's 2025 report, only 19% of US candidates said using AI in a live interview crosses the line, so most candidates think it's fine. Meanwhile 24 of the 32 employers Ophy AI tracked restrict it. People get caught in that gap. They aren't trying to cheat. They're assuming their own norm is the employer's norm.

So asking is cheap insurance, and it doesn't make you look weak. The question takes one sentence. Guessing wrong can cost you the job.

The five questions, one for each stage

Ask one specific question before each stage. The table below is the short version. The sections after it give you the wording and explain how to read each answer.

Application: the resume and cover letter

Ask: "Is there any guidance on using AI tools for the resume, cover letter or application questions?"

You often don't need to ask this one, because the safe default is already clear. Anthropic's standard (your own first draft, then AI refinement) matches what most of the 32 employers Ophy AI tracked say: polishing is fine as long as every fact is yours. If a company spells out something stricter, follow it.

Recruiter screen

Ask: "Will any part of this process be evaluated by AI, and is real-time AI assistance okay on this call?"

The first half addresses the 21% who only found out once the interview started. If the answer is that an AI tool evaluates you, prepare for that format. For the second half, expect a no, since CNBC reports recruiters may frown on real-time AI in a screen. Asking anyway shows you respect the process.

Take-home

Ask: "Is AI expected, allowed or off-limits on the take-home, and would you like me to note how I used it?"

This is the stage where recruiters may actively encourage AI. A good answer names the rule and how they want it disclosed. "Use your judgment" is not a policy. Assume they'll evaluate the result as if you used AI, and include a short note explaining what you used AI for and what you wrote yourself.

Live loop

Ask: "Are any of the onsite or virtual rounds AI-enabled, and which ones should I treat as AI-free?"

Ask about each round, because the answer is often split, as with Meta's AI round plus an AI-free one. If the reply is "it depends on the interviewer," treat every round as AI-free.

On the job

Ask: "How does the team actually use AI day to day, and is there an internal policy I'd be working under?"

Chait specifically said to ask about the job itself as well as the hiring process. This tells you more about your daily work than anything else in the loop. A team that expects Copilot in interviews but bans it in production is telling you something about how it makes decisions.

How to ask without hurting your chances

Ask in writing, ask before the stage starts, and phrase it as preparation rather than permission. Done that way, it reads as professional.

A few rules that hold up:

1. Send it by email right after the stage is scheduled, not halfway through a call. The recruiter can forward it, and you have the answer on record.
2. Frame it as prep. "I want to prepare the way you'd like me to" comes across as careful. "Can I use ChatGPT?" can sound like you need it.
3. Ask for each stage. One blanket answer won't cover a split policy.
4. Have a fallback. If nobody answers, go AI-free in anything live and disclose any AI use in written work.

A two-line version that works:

Thanks for scheduling the take-home. Before I start, could you confirm whether AI tools are expected, allowed or off-limits for it, and whether the live rounds differ?

Where you live can turn this into a legal question. Illinois HB 3773 requires employers to notify applicants when AI is used to make or assist in employment decisions. The implementing rules were temporarily postponed on June 2, 2026, but the law is still in force. In Illinois, "Will AI evaluate me?" asks about a legal duty the employer has.

Match your resume and cover letter to the answer

Write your own facts first, use AI for polish only where it's allowed, and leave out any AI skill you can't demonstrate live. The policy answer you get should change what your resume emphasizes.

Inflated claims are the biggest risk. In Greenhouse's 2025 report, 32% of candidates admitted to claiming AI skills they don't actually have. That's why an "AI tools" line on a resume tends to get probed in the interview. Change the emphasis depending on the answer:

• AI-expected employers (Canva-style): Name the tools and what you shipped with them. "Cut test-writing time by pairing with Cursor on the payments service" is something you can show in a live AI round. "Proficient in AI" isn't.
• AI-restricted employers (Amazon-style): Keep bullets focused on outcomes and your own reasoning. Make sure you can defend every line out loud with no tools, because that's exactly what the loop will test.
• Split policies (Meta-style): Do both. Show tool fluency for the AI round and keep bullets you can explain from scratch for the AI-free one.

That's a lot of rewriting when you're applying to ten companies with ten different rules. Refolk handles it: paste the posting and get your resume back, rewritten for it and built from your own work history. Nothing gets made up, so the facts stay yours, which is the standard Anthropic and most published guidance hold you to. If an employer asks for a first draft written by you, write it, then let the tool handle the per-posting tailoring.

The cover letter works the same way. Most employers that address it allow AI polish on your own content. Refolk drafts the cover letter from that same history and scores how well you actually fit the role. That score is a useful check before a live round, because the parts where you fit weakest are where an interviewer will push and where padding is most tempting.

FAQ

Can I use AI in a job interview?

Usually not in live rounds, unless you've been told otherwise in writing. Of 32 employers whose candidate guidance Ophy AI reviewed on Sept. 29, 2026, 24 restrict AI during live interviews. Some, like Canva for engineering roles and Meta in one paired round, allow or expect it in specific rounds. Ask about each round, and if nobody answers, assume no.

Is it ok to use AI for a take-home assessment?

Often yes. CNBC's reporting on Greenhouse CEO Daniel Chait's advice notes that recruiters may encourage AI on take-homes because it mirrors the actual job. "Often" isn't "always," though. Confirm by email before you start, and include a short note on what AI did and what you did yourself.

Is an AI cover letter allowed?

Generally yes, as long as you use AI to polish your own content. Anthropic's published guidance asks applicants to write their own first drafts of resumes, cover letters and application answers, then refine them with Claude. The real risk is AI adding claims that aren't true, especially about AI skills, which 32% of candidates admit to inflating.

Will asking about AI policy make me look like I can't do the work?

Not if you ask in writing and frame it as preparation. With 80% of AI-evaluated candidates calling employer AI rules vague, a clear, specific question signals that you take the process seriously. Using AI where it's banned, as at Amazon where it "may result in disqualification," is the real risk.