How to use AI during a live interview on Zoom, Teams or Meet

Search for how to use ChatGPT during a live interview and almost every result tells you how to prepare the night before: practise answers with a chatbot, research the company, write your stories. Useful, but you asked what to do while the call is happening, with the interviewer on screen and a question you did not expect.

This guide answers that question. It starts with whether you are allowed to use AI in that round at all, because the answer changes from company to company and sometimes from round to round. Then it covers the four ways people get AI help during a call, with the honest cost of each, and how to use a live helper without sounding like you are reading. The longest section is about testing, because in the reviews we read, a tool failing mid-call came up far more often than anyone being caught. The last part explains what our own app, PiriPiri AI, does and cannot do.

Is it allowed to use AI during a live interview?

It depends on the employer and on the round, and the rules are moving in both directions at once.

Some employers say no

Amazon is the clearest reported example. In early 2025 Business Insider saw internal recruiting guidance (summarised by ITPro) asking candidates not to use generative AI tools during interviews unless explicitly permitted. Other companies have answered the same worry by moving rounds into a room. In August 2025 the Wall Street Journal reported, as summarised by heise, that Google and Cisco were bringing back face-to-face interviews for some roles. The head of an IT recruiting firm told the paper that about one in three of its corporate clients were asking for in-person rounds again, up from about 5 percent the year before. Sundar Pichai said Google makes sure there is at least one in-person round “just to make sure the basics are there.”

Some employers now say yes, for certain rounds

In June 2025 Canva’s engineering team published a post titled “Yes, you can use AI in our interviews”. It replaced its computer science fundamentals screen for backend and frontend roles with an AI-assisted coding interview, where candidates can use tools such as Copilot, Cursor or Claude. The post says: “We want to see how well candidates collaborate with AI to solve problems.” What Canva grades is judgement: breaking down a vague requirement, spotting problems in AI-generated code and fixing them.

Meta has been moving the same way. In July 2025, 404 Media reported from internal messages that Meta was building a coding interview in which candidates have an AI assistant, describing it as “more representative of the developer environment that our future employees will work in.” Meta also said the format makes hidden AI use less effective, which tells you how it sees the alternative.

And some rounds ask about AI directly

Even where AI is not allowed during the call, you may be asked how you use it. In December 2024 LinkedIn’s chief operating officer, Dan Shapero, told Fortune that employers were increasingly asking candidates: “Tell me a story about how you used AI at the workplace or at home.” Have one real example ready. It is an easy question to ace if you have actually used the tools, and an awkward one if you have not.

How to find out the rule for your round

If the answer is no, respect it in that round. You can still use AI as much as you like to prepare, which is where it helps most anyway.

Is it legal to have AI listen to an interview call?

A tool that transcribes a call captures what the other person says, so recording and consent laws may apply, depending on where you and the interviewer are. In the United States, federal law allows recording when one person on the call consents, but a group of states, including California, require everyone on the call to agree (see the Digital Media Law Project’s summary, which is no longer updated, so check current rules). Other countries have their own rules, often under data protection law.

Notice how the meeting apps handle this for their own features. Zoom’s help centre says “Zoom will always notify meeting participants that a meeting is being recorded.” Microsoft’s Teams documentation says “All participants see a notification that the meeting is being transcribed.” Google says Meet “informs all meeting participants” when its note-taking feature is on. A desktop helper like ours sends no such notice, so the consent question falls on you. This is not legal advice. Look up the rules where you are and where the interviewer is, and if you are unsure, do not record or transcribe that call.

What are your options for AI help during a live call?

People who use AI during a call do it in one of four ways. Each has a real cost, and the right one depends on what the round allows and how you work.

1. Typing into ChatGPT in another window

It is the obvious first try, because it is free and already on your laptop. The interviewer asks a question, you type a short version of it into ChatGPT, and you read the answer.

The problems show up in the first minute:

• It is slow. You have to listen, type, wait for the reply and then read it. By then you have been quiet for several seconds, which is the long pause interviewers notice.
• The keys are audible. Laptop microphones pick up typing. Typing while someone asks you a question is easy to hear.
• Your eyes leave the camera. You look down at the keyboard, then across at another window, then back. On a webcam that is easy to see.
• You catch only part of the question. While you type the first half, you miss the second, which is often where the real question is.

Where it does fit is checking a forgotten fact in a round where outside material is allowed, after saying “let me check that” out loud.

2. ChatGPT voice mode on a phone

Some people prop a phone beside the laptop with ChatGPT’s voice mode on, so it can hear the interviewer through the speakers. It solves the typing problem, because it hears the question itself. It brings new ones. Voice mode is built for a spoken conversation, so its natural behaviour is to answer out loud, and you have to deal with that sound on a live call. It also hears you, and may treat your own answer as the next thing to respond to. And a phone held near the screen is easy to spot if it comes into the frame, or if the interviewer asks you to show the room.

3. Meeting note-takers that join as a bot

Note-taking services that join a meeting as an extra participant are built for work meetings: they record, transcribe and send a summary afterwards. They are not designed to give you answers during the call, and they are not discreet. When Carleton College’s IT team tested one popular note-taker in 2026, it announced itself, appeared in the participant list and had to be let in by the host.

The built-in options are no different. Teams transcription and Meet’s “Take notes for me” both notify everyone, and both are usually controlled by the host or by people inside the host’s organisation. Microsoft’s documentation says a person from another organisation cannot start Teams transcription at all. In an interview you are usually the outside guest, so these are the interviewer’s tools, not yours.

4. Desktop live helpers

The fourth option is an app on your own computer that listens to both sides of the call (your microphone, and the interviewer’s voice as your computer plays it), transcribes the conversation and drafts a short answer in a small window while the question is still being asked. This is the category PiriPiri AI belongs to, so read this part knowing that, and judge it by the trade-offs rather than our word.

What it does better than the other three:

• No typing, so no key noise, and you hear the whole question.
• No bot in the participant list and nothing extra in the room.
• The draft can be built from your CV and the job description, so it points to your own projects rather than a generic answer.

What it costs you:

• Setup. It needs permission to hear your computer’s audio and your microphone, and it depends on how your audio is routed. Earphones, a new headset or a changed default device can all break it.
• It can fail live. Lag, a missed question or a crash in the middle of the call come up often in reviews. The testing section below is about exactly this.
• Reading risk. A helper makes it easier to read a full paragraph off the screen, and reading is what interviewers notice. How you use it matters more than which one you pick.
• Money. Most charge, either monthly or per interview.

Does it matter whether the interview is on Zoom, Teams or Google Meet?

Less than people expect. For a desktop helper, what matters is where the interviewer’s voice comes out. If Zoom, Teams or Meet is playing the call through your computer, in the desktop app or in a browser tab, a helper that listens to your computer’s audio hears it the same way. What changes from app to app is the screen-share window, the menus and the recording notices, so do your test call on the same app, in the same form (desktop app or browser), that the interview link will open.

On any of them, know how you will share your screen if asked: sharing one window shows less than sharing the whole screen. Practise both in a test call and ask your friend what they actually see.

Can AI help in a phone or telephonic interview?

Only if the call reaches your computer. A desktop helper hears what your computer plays. If the interviewer calls your mobile and you talk on the handset, their voice never passes through the laptop, so the helper hears only you. Putting the phone on speaker beside the laptop is unreliable: the helper then hears both voices through one microphone and can mix up who said what.

If a phone round matters to you, ask the recruiter whether you can join by a video or computer link instead. Many are happy to send one. If not, and notes are allowed, phone rounds are where paper notes work best: a single page with your opening line, five story cue words and three questions to ask is quiet and cannot crash.

How do you use an AI helper well during the call?

Most of the advice here is about restraint. Use the helper as a memory aid: it reminds you of a story, and you tell it in your own words.

Give it your material before the call

A helper with nothing to go on produces generic answers that do not sound like you. Add your CV, the job description and the role before the interview. Then spend twenty minutes writing your five or six best stories as short cue lines (“payments migration, PM wanted to ship early, delayed a week, half the bugs”) so that, whatever the helper shows, you recognise your own material. The tell me about yourself guide covers the opener, which is the one answer you should know without any help.

Read prompts, not scripts

The useful part of a draft answer is usually one line: the project name, the number, the first point. Glance, pick that out, look back at the camera and speak. If you find yourself reading a whole paragraph, stop and summarise it in a sentence instead. The interviewer has already heard how you talk in the small talk, and a sudden change in vocabulary is easy to hear.

Keep answers short and offer more

Give the short version first and then hand control back: “I can go into how we tested it, if that’s useful.” Short answers are easier to say naturally, and follow-ups are where an interviewer checks the story is really yours. Answer those from memory. A helper can remind you of a story; it cannot tell you what you felt or who disagreed with you.

Use the pause lines you would use anyway

A prompt takes a moment to appear and a moment to read. Cover that time the way any candidate would: “Good question, let me think about the best example.” Our mind went blank guide has more of these lines, and they work with or without AI.

Know when to ignore it

Drafts can be wrong. A helper may guess at a detail from your CV, mishear a technical term, or answer the question it thought it heard. If a prompt does not match what you were asked, drop it and answer on your own. Never say a number or a claim from a draft that you cannot back up.

Review afterwards

If your tool saves a transcript, read it the same day. Note which questions caught you and what you wish you had said. That list is your preparation for the next round, which may well be in person.

Test it before the real interview

In about 130 reviews and posts about live interview tools that we read in October 2026, about 8 said the tool was seen or detected. About 30 said it froze mid-interview. A few, in the reviewers’ own words:

“had difficulty recognizing when the interviewer had finished speaking”

Trustpilot review, 13 July 2026

“Buggy, stops working and receiving answers from LLM halfway through.”

Trustpilot review, 7 July 2026

“app closed itself on its own during the interview”

Trustpilot review, 30 January 2026

Another reviewer, in October 2026, described a live mode that jumped ahead to newer answers before they had finished reading the one on screen. These are reviews of different products by different people, and every app in this category, ours included, can hit problems like these on some setups. The lesson is the same for all of them: the first time you use a live helper should never be the real interview. Run at least one full mock call, ideally two days before, so there is time to fix what you find.

The mock call checklist

Ask a friend to call you on the exact app the interview uses and to ask you five or six real interview questions, including a long one and a follow-up. Then check each of these.

1. Does it hear the interviewer? Check this first. On a Mac, apps need permission to capture system audio: open System Settings, then Privacy & Security, then Screen & System Audio Recording, where Apple lets you allow an app “both your screen and audio, or just your audio.” Check your helper is switched on there. On any computer, watch the transcript while your friend talks: their words should appear, labelled as the other person.
2. Does it hear you? Speak an answer and check your words appear too. If not, the helper may be listening to the wrong microphone, for example a monitor or webcam mic instead of your headset.
3. Headphones or speakers? With speakers, your microphone also picks up the interviewer’s voice, which can confuse who said what. Wired earphones are usually the simplest. Bluetooth headsets can switch to a different audio mode when the microphone turns on, so test with the exact pair you will wear, and do not swap devices on the day.
4. How long is the lag? Count the seconds between the end of your friend’s question and a usable prompt. Check it knows when they have finished speaking, and see what happens with a long, rambling question.
5. Can you control it? Make sure you can ask for an answer on demand, scroll back to an earlier one, and stop it moving on before you are done reading. Learn the hotkeys until you do not need to look.
6. What does a screen share show? Share one window, then your whole screen, and ask your friend exactly what they see. Have them record a short part of the call too. Do this on the same app and the same computer as the real interview, because what appears in a capture can differ between apps and operating system versions.
7. Battery and load. Run the test for as long as the real interview, plugged in. Video calls, transcription and a browser full of tabs together can heat up and slow down an older laptop. Close what you do not need.
8. Internet. Most helpers send audio to a server to transcribe it, so they share your connection with the call. Sit close to the router or use a cable, and pause large downloads and cloud backups.
9. Your fallback. Decide now what you will do if it stops: close it, look at the camera and carry on with your own notes on paper. Practise one answer that way during the mock call so it does not feel strange.

The last two days

How PiriPiri AI does this, and where it stops

PiriPiri AI is a desktop app for macOS and Windows built for the fourth option above. During a live call it listens to both sides (your microphone and the interviewer’s audio from your computer), transcribes them, and drafts a short answer on a small floating overlay, built from the CV, job description and role you added. No meeting bot joins, and it works with Zoom, Google Meet, Microsoft Teams, Webex and others because it runs on your own machine.

The hotkeys (Windows uses Ctrl in place of Cmd):

• Answer now (Cmd+Shift+Enter) asks for an answer straight away, for when a question was not picked up or you want one on demand.
• Analyze screen (Cmd+Shift+Space) reads a coding or written question off your screen and suggests an approach.
• Deep dive (Cmd+Shift+6) gives a longer answer when the interviewer pushes for detail.
• Mute mic (Cmd+Shift+M) stops it listening to you.

On screen sharing: the overlay asks the operating system to leave it out of screen capture, so it does not appear when you share your screen or record in the usual way. On Windows 10 (version 2004) and later the window is removed from captures. Developer documentation for this setting notes that some newer Mac capture tools may still pick such windows up, so do a test call with a friend on the exact app you will use before you rely on it. Leaving the overlay out of a capture does nothing about your eyes or your pauses on camera.

After the call, the transcript and notes are saved to your account, so you can review what you were asked and how you answered.

The limits, plainly:

• On a Mac, it needs an Apple Silicon Mac on macOS 14.2 or later. On Windows, it needs Windows 10 or 11 (64-bit).
• There is no phone or browser version, so it cannot help with a call taken on a mobile handset.
• Speech recognition is English only.
• It cannot give you experience you do not have, and reading its answers word for word sounds rehearsed. Treat each one as a prompt.
• It does nothing about a rule that says no AI, a proctored test or a recording consent law. Check the employer’s rules and the platform’s rules before you use it.

The setup steps are in how to use it. Use the free trial for the mock call in the checklist above, on the same app the real interview uses.

Questions people ask

Where to go next

If you are worried about being noticed, read can interviewers tell if you use AI and can the interviewer see my screen. If your interview is soon and you have not prepared yet, start with the interview tomorrow guide. To compare helpers for each meeting app, see AI interview assistants for Zoom, Meet and Teams.

PiriPiri AI costs about $2.50 an interview, paid in one-time packs that never expire, so no subscription is required. Every new account starts with 3 free sessions, so you can run the test call before a real one. See pricing, prices in India and the refund terms.

Keep reading

• Can the interviewer see my screen?
• Live help for one interview, no monthly plan
• Is an AI interview copilot worth it?