Preparation tools and live help are different products

Two products, one search term

The preparation category is older, larger and much more crowded. It covers mock interview generators, question banks sorted by company, CV rewriters, answer scoring, filler-word counters and video review. The live category is newer and small: tools that follow the conversation while it happens and put something in front of you inside it.

They share vocabulary because they share a buyer, and the vocabulary hides the only difference that matters: one of them has time and the other does not. Everything else — what they can know about you, how long their output is allowed to be, what happens when they are wrong — follows from that.

What preparation tools genuinely do well

Three things, and they are not small. The first is inventory. Most people arrive at an interview having never written down the six stories they will be telling all month, and a mock round forces that list into existence. Once the stories exist, they get reused; the tool’s value was making you produce them, not producing them for you.

The second is calibration on length. Almost everybody’s first answer to “tell me about a time you disagreed with a colleague” runs four minutes and should run ninety seconds. Hearing yourself back is unpleasant enough that it works in one sitting.

The third is exposure to the shape of a format you have not done. A case interview, a system design round and a behavioural loop have different unwritten rules, and finding out what they are while it counts is expensive. Two rehearsals is usually enough to stop the format itself from being the obstacle.

Where mock interviews stop working

They stop at the first follow-up. A rehearsed answer survives being asked; it does not survive “and what would you have done if the deadline hadn’t moved?”. Real interviews are mostly second and third questions, and those are not in any question bank because they are generated from what you just said.

They also stop at the scoring. A model marking your answer out of ten is marking fluency, structure and keyword coverage, because that is what is legible in a transcript. It cannot mark the thing being marked in the room — whether the interviewer believed you had done it. A high score on a rehearsed answer and a rejection are entirely compatible, which is confusing if you thought the score was the signal.

And they stop at the surprise. Preparation covers the questions you imagined. The round is usually decided by the one you did not.

What live help does instead, and the three constraints it lives under

A live copilot listens to the call, works out which turn was addressed to you, and puts two or three sentences on screen while the topic is still open. That is a narrower job than it sounds, because it is done under conditions preparation tools never face.

• Seconds, not minutes. An answer that arrives after the conversation has moved on is not a late answer, it is no answer. This is why live tools work from a short window of recent lines rather than anything thorough.
• A glance, not a read. Somebody is watching your face. Anything you have to read carefully is worse than nothing, because reading it is visible. Output that would be a fine chat response is unusable here.
• No second draft. Preparation output gets edited before it ever meets an interviewer. Live output might come out of your mouth. A confidently wrong sentence costs the round.

The test that separates them in one question

If the output would still be useful arriving an hour later, it is a preparation tool. If lateness makes it worthless, it is a copilot. That is the whole boundary, and it survives whatever the marketing on either site says.

It also explains why the two do not merge into one product as often as you would expect. A tool built to be thorough is built around having time, and a tool built to be fast has given that up on purpose. Feature lists can overlap; the constraint cannot.

The week before, and the day of

In practice the sequence is not a choice between them. In the week before, the preparation work is inventory: the six stories, the numbers you can defend, the twenty words your field uses constantly, the format of the round you have not done. None of that is live work and none of it can be done live.

On the day, one line of context is worth more than all the rehearsal. Whatever tool is in front of you during the call, telling it what this call is — the role, the round, the stack, one thing about you that ought to come up — is the difference between a well-formed paragraph about somebody who does not exist and a first sentence you can actually finish out loud. In Trippi Cue that is the “About this meeting” field; in anything else it is whatever you set before you join.

The rehearsal is what makes the live suggestion usable, incidentally. A first clause only helps somebody who already has the story it opens.

What neither of them does

• Neither knows your history. Both will invent one if you let them, and the invented version fails on the follow-up rather than on the answer.
• Neither sees the coding pad. A browser-based copilot hears the tab; a preparation tool was never in the room. Screen content is outside both.
• Neither settles whether you are allowed to use it. Preparation is almost never disputed; live assistance frequently is. That is a question about your interview, not about the tool — we wrote it up in is using AI in an interview allowed.
• Neither makes you good at a job you cannot do. Both shorten the distance between what you can do and what comes across. That is the honest size of the claim.

If you have to pick one

Pick by which failure is yours. If your answers wander, run long, or fall apart because you have never said them out loud, the problem happens before the call and a preparation tool addresses it. If your answers are fine when you have five seconds to start and disappear when you have none — the unexpected question, the round in a second language, the panel where you missed what was asked — the problem happens inside the call and rehearsal will not reach it.

Most people who buy both end up using preparation once per job search and live help once per interview. That ratio is about right, and it is a reasonable way to decide what to pay for.

Read next

• Why an Interview Copilot Should Be a Browser Extension
• What Is an Interview Copilot and How Does It Work?
• How to Use AI to Answer Interview Questions Live