In this article
Before you start
Why interviewers ask this
The question separates a candidate who has read the rules of each stage from one who assumes a single rule fits everything. Some employers now write those rules down, Anthropic among them, so the first thing a candidate shows is care about what is allowed where.
A second form of the question tests your own habits. A candidate reported that an interview for an AI-lab role asked how they use AI tools in their own workflow, with concrete daily usage, how they check the output and a story of a failure. A weak answer says "I use it for everything" or "I never use it." A strong answer names the tasks, the checks and one case where the tool was wrong. This page covers both: what the rules are, and how to talk about your own use.
The stage rule from first principles
- Stage rule
- A stage rule is an employer's statement of whether AI assistance is allowed in one step of hiring, such as the application, a take-home, preparation or a live interview, and it can differ from one step to the next.
- AI-enabled round
- An AI-enabled round is an interview in which the company explicitly lets the candidate use an AI assistant and judges how well the candidate works with it, not only the final result.
In plain words
Think of three kinds of school work. Coursework lets you use books and help. A closed-book exam does not. Revision for the exam is encouraged with any tool you like. Anthropic's published hiring guidance looks like that mix, with the company saying which kind each part is.
The analogy breaks in one place. A school has one rulebook for everyone, while each company writes its own, so you cannot carry a rule from one employer to another.
In one precise sentence: a stage rule says, for each step of one company's hiring, whether AI is allowed, and Anthropic's guidance says the company will state it. Reported practice differs between companies, and only one company's published rules are cited here.
Step 1: In Anthropic's live interviews, AI is allowed only when announced
A live interview is the round where you talk or code with an interviewer in real time. In its guidance, Anthropic writes "This is all you–no AI assistance unless we indicate otherwise." The default is no AI, and the exception must be announced.
Input: an invitation to a live round. Output: a rule you can act on. Anthropic adds that it will be clear when AI is allowed, with an example such as "You may use Claude for this coding challenge." An AI-enabled coding round is the case where the company says yes.
One candidate reported that Meta added an AI-enabled coding round to an engineering manager full loop, with a simple puzzle-game sample question, and that it was the only coding round for that role. This is a candidate report, not a company statement, and no public Meta source is cited here; see the Meta engineering manager loop. In Anthropic's process, no announcement means no AI; for any other employer, ask.
Step 2: In Anthropic's take-homes, the same default applies
A take-home assignment is a task you complete on your own time and submit for review, such as a coding exercise or a written case. Anthropic's guidance says to complete these "without Claude unless we indicate otherwise." In that guidance the default is again no AI.
Input: a take-home brief. Output: a submission you can fully explain. The brief is where the company states its exception, so read it for a line that allows AI before you open any tool. Anthropic lists having Claude write your code when it has not explicitly allowed it as not allowed.
A practical reason supports the default: a reviewer reading your submission often follows up in a later round, and you must be able to defend every line. For what a strong submission looks like, see the take-home example, and for timed screens see the online assessment guide. Take-homes and live rounds share one rule, so the first two steps give you the main habit.
Step 3: Anthropic's application allows AI to refine, not to invent
The application is the curriculum vitae (CV), cover note and any written answers. Anthropic asks candidates to "create your first draft yourself, then use Claude to refine it." Its guidance also says to "Use AI to refine your ideas, not replace them."
Input: your own draft. Output: a clearer version of the same facts. The line the guidance draws is between polishing and inventing. It lists using Claude to create experiences as not allowed, so an assistant may tighten a sentence but may not add a project you never did.
This corrects a common belief that any AI use on an application is wrong. Under this published rule, refining your own draft is welcome, and fabricating is the violation. A good habit is to keep the draft that came before the assistant, so you can show the work is yours. Application steps vary, so check the form's own instructions too.
Step 4: Anthropic encourages AI use in preparation
Preparation is everything you do before the round: studying, rehearsing answers, reviewing questions. Anthropic lists interview preparation with Claude as encouraged. This is the one stage in its guidance where the answer is a plain yes.
Input: the questions you expect. Output: answers you can say in your own words. Use an assistant to explain a concept you are rusty on, to quiz you, or to give feedback on a practised answer. Then say the answer aloud without it, because the round tests you, not the tool.
Anthropic also says it uses Claude every day and looks for candidates who excel at collaborating with AI. One candidate reported being asked about their own AI workflow in an AI-lab interview. For the wider picture of skills a round tests, see interview skills.
Step-by-step description
- No AI unless the company says otherwise.
- No AI unless the company says otherwise.
- Draft it yourself, then refine.
- AI use is encouraged.
- Live interview: no AI unless the company says otherwise.
- Take-home: no AI unless the company says otherwise.
- Application: draft it yourself, then refine it.
- Preparation: AI use is encouraged.
In short: live interview, no AI unless the company says otherwise. Take-home, no AI unless the company says otherwise. Application, draft it yourself and then refine it. Preparation, AI use is encouraged.
What else candidates report has changed
Beyond published AI rules, candidates describe other shifts. These are reports by candidates, not verified company statements, and they vary by employer.
- Reported by candidates: in engineering manager loops, every company asks about AI familiarity, and one one-to-one conversation was run by an AI rather than a human.
- Reported by candidates: a recruiter passed back that an answer to "how do you fix a bug in production" was "too generic".
- Reported by candidates: non-technical recruiters check answers against technical trivia and keyword lists.
- Reported by candidates: application essays and life-story questions, with little room to ask for clarity.
- Reported by an interviewer: an opening segment called "teach me something".
- Reported by candidates: open-ended openers about a favourite thing you have done.
A strong answer to any of these is specific and first-hand, so prepare real examples with concrete decisions.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →The Staff-level answer
Say it as a short rule, then show you apply it. If the rule is stated, follow it exactly. If it is not stated, ask the recruiter in writing before the round.
Interviews now involve AI. How do you handle that as a candidate?
I treat it as a rule per stage. I check what the company has said for each step. For preparation I use AI freely to quiz myself and tighten my answers, then I practise without it. For the application I write the draft and use AI only to refine it. In take-homes and live rounds I use none unless they say so, and I ask if the invitation is silent. In my own work, I use AI for first drafts and boilerplate, I check every output by running or reviewing it, and I can tell you a case where it was wrong.
What makes this Staff-level is that it states the rule for every stage, names the verification habit, and offers a failure story instead of claiming the tool is always right.
Text description of the diagram
- If the rule is stated, follow it exactly.
- If it is not stated, ask the recruiter in writing before the round.
The rubric
| Dimension | Weak answer | Strong answer | Staff answer |
|---|---|---|---|
| Correctness | "AI is banned everywhere." | AI is allowed in some stages, banned in others | Names the stages and the stated default for each |
| Depth | Mentions only live rounds | Covers application, take-home, live and preparation | Separates refining a draft from inventing content |
| Trade-offs | None | Notes the risk of unannounced AI use | Explains why you ask in writing when the rule is silent |
| Communication | Vague and defensive | Calm, structured, under a minute | Ends on a concrete example from your own work |
| Follow-up readiness | Cannot explain own AI use | Gives a daily-use example | Adds how you check output and one failure |
Three variants of this question
Can I use AI to write my application?
You can use AI to refine a draft you wrote yourself. Anthropic's guidance asks for your own first draft and says AI should refine your ideas, not replace them, and it lists creating experiences with AI as not allowed.
Can I use AI in a take-home or online assessment?
Not unless the company says so. Anthropic's rule is to complete these without AI unless it indicates otherwise. If the brief is silent, ask the recruiter in writing before you start.
How do you use AI tools in your own workflow?
A candidate reported this question in an AI-lab interview. Answer with the tasks you use it for, how you verify the output, and one real failure, in that order, so the interviewer hears judgement rather than enthusiasm.
Common mistakes
- "Companies ban AI everywhere." (wrong) The published rule varies by stage, and some rounds explicitly allow AI. (right) "It depends on the stage, and I check what they have stated."
- "Any AI use on an application is wrong." (wrong) Refining your own draft is encouraged in the published guidance. (right) "I write the draft, and AI only refines it; I never add experiences I do not have."
- "If the invitation is silent, I can assume AI is fine." (wrong) The default in the cited guidance is no AI unless stated. (right) "If it is silent, I ask in writing first."
- "I use AI for everything, and it is always right." (wrong) This shows no verification habit. (right) "I check each output and can describe a time it was wrong."
Follow-up questions
How have tech interviews changed in 2026?
Two kinds of change are visible. Some employers publish AI rules for each stage, and Anthropic's guidance (updated July 2025) is a public example. Candidates also report AI-enabled coding rounds, questions about their own AI use, generic-answer feedback and more open-ended openers; these reports are listed above and vary by company.
Why do interviewers ask about your favourite past project?
A candidate reported this as an open-ended opener. The reason given is unverified, but the question rewards a specific, detailed story that only the person who did the work can tell, so prepare one project with concrete decisions and results.
Is an AI-enabled coding round different from a normal one?
Yes in what is judged: the company lets you use an assistant and looks at how you work with it, not only at the final code. The AI-assisted coding interview page covers the format.
What if the invitation does not say whether AI is allowed?
Ask the recruiter in writing and follow the answer. Anthropic's guidance says it will be clear when AI is allowed, and a polite question before the round costs nothing. Role-specific details sit on the employer's own pages, such as Careers at Anthropic. The Anthropic interview process page shows how that company structures its rounds.
Does AI make the hiring decision?
Anthropic's guidance states: "We don't use your data to train Claude or let Claude make hiring decisions." That statement is about Anthropic only, so do not assume the same for other employers.
- Some employers, Anthropic among them, publish AI rules stage by stage; check each employer's own rules.
- In the cited guidance, live interviews and take-homes allow no AI unless the company says otherwise.
- On an application, refine a draft you wrote yourself; never invent experiences.
- Preparation is where AI use is encouraged, but practise the answer aloud without it.
- If the rule is silent, ask the recruiter in writing before the round.
Sources
- How to collaborate with Claude during our hiring process (Guidance on Candidates' AI Usage) — Anthropic, 10 Jul 2025
- Careers at Anthropic — Anthropic, 26 Oct 2026
