Interview Signal interview kit: questions from a job description

Interview Signal builds a role's interview kit from the job description you paste, or from a posting link it fetches for you: 8 to 14 questions in the order a real interview runs, each with a note on what a strong answer contains, 3 to 5 of them marked must-ask, and a scorecard of 4 to 7 weighted areas with a rubric for each. Everything is editable before you save, and the saved kit becomes the question guide for every interview on that role. This page describes version 0.9.10, the current release as of October 2026.

How do I build a kit from a job description?

1. On the home screen, press New role (or the N key).
2. Type the Role title, then paste the posting into Job description. If the posting is online, paste its address into Paste the posting link and press Fetch instead.
3. Optionally, fill in What the client actually wants. This is the brief: anything the job description does not say. The box itself warns that the brief "outranks the JD wherever they disagree."
4. Choose the Interview language, the Submittal template, the Interview type and the Difficulty. A hint under the last two says what each choice does.
5. Press Generate the kit. The questions and scorecard appear on the right.
6. Edit anything you disagree with, then press Save role. The app confirms: "Role saved. Every recruiter on the team sees it."

The job description has to be more than a few lines; with too little text the app asks you to paste the full description first. The walkthrough video on the Try it page shows the whole flow in its "Build the interview kit" chapter, which starts at 0:53.

What does a generated kit contain?

• 8 to 14 questions, ordered the way interviews run: context and background first, then depth, then logistics last. Each has a category (Experience, Technical, Coding, System design, Behavioural, HR, Logistics or Motivation) and a planned answer time.
• What a strong answer contains, under every question. It is written in concrete terms for an interviewer who is not a specialist, so you can tell a real answer from a fluent one: "Mentions a real migration they ran, with a rollback plan", not "demonstrates deep knowledge".
• Must-ask marks on 3 to 5 questions, explained below.
• A scorecard of 4 to 7 areas, each with a weight from 1 to 3 and a rubric saying what a 1-3, a 5-6 and a 9-10 answer looks like. The model is asked only for areas the interview can actually evidence from what the candidate says.
• A consent notice: one first-person sentence to read at the start of the call, written for the jurisdiction in your settings. The editor's note reads "Read at the start of the call. Edit it to sound like you."

Chips above the questions show the totals, the planned minutes of answers, and which AI model wrote the kit.

What do Interview type and Difficulty change?

Interview type decides the shape of the questions. The labels and hints below are the app's own:

For Balanced and HR screen kits, the model is told to include notice period, compensation expectation and right-to-work questions, because a submittal without them is unusable to a client. The four single-type kits contain only that type; off-type questions the model returns anyway are removed, and the app tells you how many.

Difficulty sets the bar an answer has to clear, not how obscure the questions are: Easy (junior), Medium (mid-level), Hard (senior), Expert (staff and above) and Mixed, which starts easy and gets harder so a weak candidate shows early. The role remembers both choices and its language.

How are must-ask questions chosen?

A must-ask is a question that, left unasked, makes the submittal unusable: notice period, compensation expectation, work authorization, and location when the brief implies it. On top of those, the model may mark at most two capability questions, and the brief decides which. It is told not to mark a skill must-ask when the job description lists it but the brief never raises it, however important it looks.

The app enforces a cap of five. If the model marks more, the extras stay in the kit as ordinary questions and a message says so, so you can tick Must ask on any you disagree about. The cap matters because every must-ask stays prominent in the live guide until it has been asked; see how the question guide works during the call.

Can I fetch the job description from a posting link?

Yes, for many postings. Fetch reads the page, strips navigation, cookie banners and "similar jobs" panels, and puts the text into the Job description box, filling an empty role title from the page title. Read the result before generating; the app's own note says "job boards put a lot of things on a page." A very long posting is trimmed. Some job boards build the description with JavaScript, which the fetch cannot run; then the app asks you to copy and paste the description instead.

How do I edit the kit and regenerate without losing questions?

• Drag a question by its handle to change the running order.
• Edit the question text and its "What a strong answer contains" line directly. Rewording a question makes it yours, and its Keep box ticks itself.
• Tick Keep on any question you want to survive a regenerate. The button then reads, for example, Regenerate the rest (3 kept): new questions are written around the kept ones, which stay where they were, and a new question that repeats a kept one is left out with a note.
• Add a question adds a blank one, kept automatically. The × removes a question.
• Rename scorecard areas, change a Weight, rewrite the rubric lines, or press Add a scorecard area. A kit needs at least one area, so the last one cannot be removed.

If a regenerate fails, the kit stays exactly as it was, kept questions and all.

Which questions does it avoid?

The model is instructed not to write questions that are illegal or inadvisable in a hiring context: age, marital or family status, health or disability, religion, ethnicity, pregnancy, or salary history where it is restricted. It asks about salary expectation, never history. Read the kit before you use it anyway. This is an instruction to an AI model, not a legal review, and Interview Signal is not legal advice.

What is sent, and where is the role kept?

The kit screen says it plainly: "The job description and brief are sent to generate the kit. Nothing from a live interview is ever sent." Save role keeps the role in your account so that, on a Team plan, every recruiter runs the same questions against the same rubric. A role holds the title, job description, brief and kit, and nothing about a candidate. For what stays on your computer, see where your data is stored and the security page.

What the kit builder does not do

• It does not pull job descriptions from an ATS. Interview Signal has no ATS integration, so you paste the text or fetch the posting.
• It needs a connection. Generating a kit is AI work, so once your free minutes or bought hours are used up it closes along with scoring, showing the same message and a button to buy hours or upgrade (see pricing).
• It drafts; it does not know your client. The model can misjudge a specialist field, so treat the kit as a first version and edit it. A job description that screens makes a better input, and weighted scorecards explains the weights.

To try it without a live role, practice roles come with sample kits. Download the app for Windows 10 or 11, or macOS 13 or newer; the first 20 minutes of transcription are free, with no card.