AI Use and Limitations

Most products in this category will not tell you this. Here is exactly where the software is generating text, and exactly where it can be wrong.

Where AI is used

Step	What the model does	How wrong can it be?
Reading the job description	Pulls out the requirements and which ones repeat	It can miss an implied requirement, or over-weight a boilerplate line
Writing your six questions	Builds questions from those requirements	A question may be slightly off-target for the real role
Deciding on a follow-up	Judges whether your answer lacked an action or a result	It sometimes asks for a result you already gave in different words
Writing the feedback	Quotes your words and explains what worked or did not	This is where errors matter most. It can misread tone, or mark something as missing that was implied
Suggesting a stronger shape	Rearranges only facts you gave	Constrained to your own words — see below
The practice plan	Summarises the session into three actions	Can over-weight one weak answer

Where AI is deliberately not used

• Transcription in the free version happens in your browser, not on a server. The audio does not leave your device.
• No scoring model. There is no number, so there is no false precision to defend.
• No analysis of your voice, face, tone or confidence. Claims made in this area by other products are not well supported, and we are not making them.

The invention problem, and what we do about it

Language models are eager to help. Asked to improve your answer, an unconstrained model will add a plausible metric you never mentioned — and if you repeat it in a real interview, you are now describing work you did not do.

Three defences:

1. The model is instructed to use only facts present in your transcript or your resume.
2. Its output is validated against a schema with a flag that fires when it introduces an unsupported fact; flagged output is not shown to you.
3. Where a fact is genuinely missing you see a visible bracket — [ADD YOUR REAL NUMBER] — instead of an invention. You can see the gap, which means you can fill it truthfully.

This is not perfect. If you ever see the product state a fact about your career that you did not give it, that is a serious bug and we want to hear about it immediately.

What it is bad at

• Industry-specific judgement. It does not know your sector the way a specialist would.
• Cultural and regional interview norms. Its defaults lean North American.
• Anything about the specific company, its team, or the person interviewing you.
• Knowing when you are right and it is wrong. Trust your own knowledge of your work over its feedback, always.

Who processes your data

Processor	What they handle
[HOSTING PROVIDER]	The application and its database
[MODEL PROVIDER]	The text of your answers, job description, and resume facts, at the moment feedback is generated
Stripe	Payment. We never see your card.
[EMAIL PROVIDER]	Sign-in links and receipts
[ANALYTICS PROVIDER]	Anonymous event counts. No text you wrote.