AI Interview Practice: A 7-Day Plan Using ChatGPT, Gemini, or Claude

AI interview practice means using a chat assistant as a mock interviewer: you give it the job posting and a short summary of your background, it asks one question at a time, you answer out loud, and it critiques your structure and specifics using a rubric you set. ChatGPT, Gemini, and Claude can all do this with plain text prompts, and the quality of the practice depends far more on your instructions than on which one you open.

The most common mistake I see in advice on this topic is asking the AI to write your answers. The result reads smoothly on screen and collapses the moment an interviewer asks, “What specifically did you do?” In this guide the AI asks and scores, and you do the talking. Feature availability and usage caps change often, so everything about specific tools below reflects the vendors’ public help pages as of October 2026; confirm in the app you actually use.

For the interview fundamentals that no chatbot replaces, such as what to wear, how to handle salary questions, and which questions to ask the employer, see our job interview prep guide alongside this one.

What is AI interview practice, and what can it actually do?

AI interview practice is a repetition tool. It is good at three things: turning a job posting into likely questions, spotting vague or structureless answers, and asking the same question again without getting bored or judgmental.

It is not good at knowing your real interviewer, predicting how a specific hiring manager will score you, or telling you true facts about a company. Its feedback on voice is usually feedback on the words it heard, not on your tone, pace, or eye contact. Here is a sensible division of labor.

Task	Hand it to the AI	Keep it for yourself
Generating questions	Likely questions and follow-ups from the posting	Researching recent company news
Checking answers	Missing STAR parts, vague claims, length	Choosing which stories to tell
Repetition	Re-asking, raising the difficulty	Actually speaking out loud
Scoring	Consistent rubric across attempts	Not over-trusting the number

The U.S. Department of Labor’s CareerOneStop site gives the same core advice without any software: review the posting, match your qualifications to each requirement, and prepare stories about your experience. AI just lets you rehearse that advice with an opponent that talks back.

Which tool should I use: ChatGPT, Gemini, or Claude?

All three can run a text-based mock interview, so start with whichever app you already have open. The real differences are voice features and free-plan limits, and those are the parts that change most.

	ChatGPT	Gemini	Claude
Text mock interview	Yes	Yes	Yes
Voice conversation	Voice mode in the app (OpenAI help)	Gemini Live in the mobile app (Google help)	Voice mode in the iOS and Android apps, labeled beta (Anthropic help)
Free-plan voice limits	Limited access measured over a rolling 24-hour window; older pages list different numbers	Varies by plan and region	Counts against your normal usage limits
Good fit	Continuing a long practice thread	Quick spoken sessions on a phone	Pasting long postings and resumes for detailed critique

A caution on the free-plan row. Blog posts disagree wildly: some say free ChatGPT voice is about 15 minutes a day, an older OpenAI page mentioned two hours, and the newest help page I found gives no free-tier number at all, only that access is limited and may change. I would not plan around any figure. Start a voice session, and when the app tells you the limit is reached, switch to typed or dictated answers.

If you want a deeper comparison of how the three differ in writing style and reasoning, our ChatGPT vs Claude vs Gemini comparison covers it. A practical trick: run the same mock question through two of them and compare the feedback. Differences in what each one catches are a useful reminder that none of them is the interviewer.

How do I set up a mock interview step by step?

Set aside 30 minutes the first time. The setup is what separates a useful session from a chat that tells you everything is great.

Step 1: Paste the entire job posting. Include responsibilities, required qualifications, and preferred qualifications. Do not summarize it. The questions come from the posting’s own wording.

Step 2: Add a short experience summary. Three or four roles or projects with the outcomes and numbers you can defend. Leave out your name, address, phone, and email.

Step 3: Set the role and the rules. Tell the AI who it is (a hiring manager for this team), and put the rules in writing: one question at a time, wait for your answer, no hints before you answer, no praise in the middle.

Step 4: Answer out loud. Use voice mode, or dictate, or at minimum speak to the screen and then paste your transcript. Typed answers are tidier than spoken ones and hide the exact problems you need to fix.

Step 5: Get the critique, then redo the answer. The second attempt at the same question is where improvement happens. Moving on to a new question after one pass is a tour, not practice.

Why do the results feel too easy?

Chatbots lean agreeable. If you do not set a strict scoring rule, you will hear “great answer” on nearly everything. The prompts below build in a harsh rubric on purpose: high scores are only available when you name your own role, give a number, and state a result. If your first session scores everything above eight, tighten the rule rather than celebrating.

What prompts can I copy and paste?

Replace the bracketed parts. These work in plain text in any of the three apps.

Prompt 1: Turn the posting into questions

Read the job posting below and write the interview questions I am most
likely to get, three each for: role knowledge, past experience
(behavioral), situational judgment, and motivation for this job.
Next to each question, add one line on what the interviewer is trying to
learn. Do not guess facts about the company that are not in the posting;
mark those "verify".

[paste job posting]

The last instruction guards against invented company details before they enter your notes.

Prompt 2: Run the mock interview

You are a hiring manager on the [team] at a [type of company]. Interview
me for the [job title] role using the posting and my background below.
Rules:
1. Ask exactly one question at a time.
2. Wait for my answer. Give no hints or model answers before I respond.
3. If my answer has a gap, ask one realistic follow-up, like a real
   interviewer would.
4. Ask eight questions in total and tell me when we are done.
Start with "Tell me about yourself."

[paste short background summary]

Prompt 3: Score the answer with STAR

Evaluate the answer I just gave on:
- Which STAR parts (Situation, Task, Action, Result) are missing
- Whether it is clear what I did versus what the team did
- Whether there is a number or concrete example
- How long it would take to say out loud, and what to cut
Score it out of 10. Give 8 or higher only if my personal role and a
specific result are both clear. Tell me what to fix before anything
positive. Do not write a model answer; tell me the direction of the fix.

Prompt 4: Drill your weak spots

From this conversation, list the three weaknesses that keep coming up in
my answers. Then ask me five new questions aimed at those weaknesses, one
at a time.

Do not delete the line about not writing a model answer. Without it, the assistant will happily hand you a polished script, you will memorize it, and the practice is over.

What does a 7-day practice plan look like?

Thirty to forty-five minutes a day is enough. If your interview is tomorrow, compress the plan from day three onward.

Day	Focus	How to use the AI	Target
1	Analyze the posting	Prompt 1	Five keywords from the role
2	Build your story bank	Show drafts, ask what is missing	Four core stories in STAR form
3	Introduction and motivation	Write it, then say it three times	Under one minute each
4	Behavioral questions	Prompts 2 and 3	Each answer about two minutes
5	Role knowledge and situational questions	Increase follow-up difficulty	Say “I don’t know, here is how I would find out” calmly
6	Full simulation	Do not stop for feedback until the end	Forty uninterrupted minutes
7	Fix weak spots, prepare your questions	Prompt 4	Three questions for the interviewer

Day six is the one people skip, and it matters most. A real interview does not pause to grade you. Hold your feedback until the end so you practice recovering from a stumble instead of restarting. If your materials are the weak link, fix them first with our resume and cover letter guide or the step-by-step resume guide, then come back.

For virtual interviews, which are common in the U.S., CareerOneStop also recommends testing your camera, microphone, and the employer’s video platform and choosing a quiet, well-lit spot; its virtual interview page lists the basics. A practice session with your laptop camera on is a free dress rehearsal for this.

What should I never share with an AI chatbot?

Assume anything you paste could be stored, reviewed under the service’s policies, or used to improve the product depending on your settings. Those settings differ by app and plan, and their names move, so open the privacy or data controls in the app and look before your first session.

Category	Fine to include	Remove or generalize
Identity	Job title, years of experience, degree level	SSN, home address, phone, personal email
Past work	Duties, outcomes you could say publicly	Your current employer’s nonpublic numbers, client names, internal documents
The new company	The job posting, public press releases	Anything under an NDA, take-home assessments you were told to keep private
Your answers	Draft stories	Screenshots of internal chats or emails

Current-employer confidentiality is the one to take seriously. If a story depends on a real client or number, anonymize it: “a large retail client” and “a seven-figure budget” work just as well for practice, and they are also how you should tell the story in the real interview. Some assistants keep memory across chats; our explainer on ChatGPT’s memory feature shows how that works and how to clear it.

What are the limits of practicing with AI?

There are three that matter.

First, the AI does not know your interviewer. It rewards a generically good answer. A real panel might care about one narrow skill, a particular tool, or a culture signal that no posting spells out.

Second, it can invent facts. A mock interviewer may say “your company recently expanded into X” with total confidence. Check anything about the employer on its own website and careers page, and for public companies in SEC filings, before you repeat it. Putting a wrong fact about the company in an interview answer costs you more than leaving it out.

Third, scores are loose. Ask for the same evaluation twice and you may get different numbers. Look for repeated comments, not the digit.

A composite failure case

This is a composite of a common pattern, not a single real person. A candidate asks an assistant to “write my answer to the conflict question,” memorizes the clean result, and delivers it well. The interviewer asks, “What did you say to the other person in that meeting?” The memorized story has no meeting in it. The candidate pauses too long and then rephrases the same summary. Practicing with an AI that kept asking follow-up questions until the story had real detail would have prevented it.

For anyone with free access to career services, many American Job Centers and school career offices run mock interviews and prep classes with people. Add one of those before a high-stakes interview.

What mistakes should I avoid?

• Having the AI write your answers. You will sound rehearsed in a way that breaks under follow-ups.
• Practicing only by typing. The interview happens in speech. Speak at least once a day.
• Accepting praise. Without a strict rubric, assistants flatter. Raise the bar in the prompt.
• Letting it ask many questions at once. You lose pacing and pressure.
• Trusting company claims. Verify on official pages.
• Running one endless chat. Long threads drift and the rules get forgotten. After 30 to 40 minutes, open a new chat and paste Prompt 2 again.
• Pasting private data. Use the table above as a filter.
• Skipping the human run. One live mock interview with a friend or career counselor catches what the AI cannot: your eye contact, your pace, your nerves.

If you also want to speed up the rest of your job search, our look at AI tools for working professionals covers where an assistant helps and where it gets in the way.

---

This article is general information based on vendor help pages and public sources as of October 2026. Features, usage limits, supported languages, and data settings vary by plan, region, and over time, so check the current app and the vendor’s help center. It does not guarantee any hiring outcome. Do not enter personal identifiers or confidential employer information into any AI service.