Using Interview Signal

How follow-up suggestions work in Interview Signal

On this page
  1. When do follow-ups appear?
  2. What do probe, clarify and evidence mean?
  3. What are the suggestions based on?
  4. What makes a suggestion worth asking?
  5. How do I ask one of the suggestions?
  6. Which AI writes them, and can I see two sets at once?
  7. What is recorded about follow-ups?
  8. Limits
  9. Questions people ask

During a live interview, Interview Signal suggests three follow-up questions after each substantial answer from the candidate. They appear under the transcript, best first, each tagged probe, clarify or evidence with a short reason, and each is meant to dig into something the candidate actually just said. Press Ask this and the question moves onto your guide, pinned, so it stays readable while you ask it. This describes version 0.9.10, current as of October 2026.

When do follow-ups appear?

A new round is requested whenever the candidate finishes a stretch of speech of about 25 words or more. Short replies such as "yes, about two years" do not trigger one, which keeps the suggestions about real answers rather than half sentences. Until the first set arrives, the panel reads "Follow-ups appear here after the candidate has been talking for a moment."

Each round replaces the last. The ↻ button beside the Follow-ups heading (tooltip "Suggest again") asks for a fresh set at any moment, and the chip beside the heading names the AI model that wrote the current set.

What do probe, clarify and evidence mean?

TagWhat the suggestion is forIllustration of the shape
probeGo deeper into something they said"You mentioned the migration ran three weeks. What went wrong in it?"
clarifyThey were vague"When you say 'we' rebuilt it, which part was yours?"
evidenceGet a concrete example"Can you walk me through one account you turned around?"

Under each question is the reason it was suggested, in under ten words. A suggestion can also be aimed at one of your scorecard areas, so the follow-ups lean toward what the role is scored on.

What are the suggestions based on?

  • The question on the table: the one you pinned in the guide, or, with nothing pinned, your most recent turn that sounds like a question.
  • The last minute and a half of the conversation, not the whole transcript. For a follow-up, what matters is what was just said, and a short excerpt keeps the suggestions quick.
  • The guide questions still to cover, so a suggestion does not duplicate a question you are going to ask anyway. Its job is the ground between them.
  • Your scorecard areas and the interview language.
  • The candidate's CV, if you attached one, so a suggestion can go after a gap between the CV and what they said. It still quotes what was said, never what the CV claims. See attaching a CV.

What makes a suggestion worth asking?

A suggestion may be read aloud to a real person seconds after it appears, so the model is told what a good one looks like. It names or quotes something the candidate actually said. It goes after the places candidates tend to be vague: scope ("we"), outcome ("it went well") and their own part. It is one sentence you can ask without rehearsing.

The model is told to avoid compound questions, leading questions that hand over the answer, rephrasings of questions already covered or still on the guide, and anything illegal or inadvisable to ask, such as age, family or marital status, health, religion, ethnicity, pregnancy, or salary history where it is restricted. These are instructions to an AI model, not guarantees, so read a question before you ask it.

How do I ask one of the suggestions?

  1. Read the three suggestions while the candidate finishes.
  2. Press Ask this on the one you want. It leaves the panel and appears on the guide, pinned, under the pinned question's category or under "Follow-ups you took". A note confirms: "On the guide and pinned. It ticks off when you ask it."
  3. Read it aloud. It ticks off like any guide question, carries its own clock, and the next scoring pass scores the answer against it.
  4. Ignore the others. A new set arrives after the next substantial answer.

Our guides to interviewing for a role you have never done and probing resume gaps cover follow-up technique in more depth.

Which AI writes them, and can I see two sets at once?

Follow-ups use the model chosen under AI in the live screen's top bar, or in Settings under "AI during interviews"; Auto picks the fastest model for each job, with fallbacks. A change mid-call applies from the next round. Turn on Dual AI to see suggestions from two models side by side, in columns A and B, with B's repeats of A removed.

On technical questions the same request also carries the accuracy check, which marks the candidate's claims in the transcript. For a question of your own, use Ask the AI, the box directly under the follow-ups.

What is recorded about follow-ups?

Two numbers: how many suggestions were shown and how many you pressed Ask this on. They are among the interview counts listed on the security page and feed your coaching report. The suggestions themselves are not stored; a follow-up you asked is saved with the interview on your computer, as a guide question.

Limits

  • Follow-ups need a connection. Without one the panel stays as it was, and the interview, transcript and guide carry on.
  • When your free minutes or bought hours run out, follow-ups stop with the other AI features (see pricing).
  • They depend on the candidate being transcribed. If the meeting's sound is not being captured, there is nothing to follow up on.
  • They come from a fast model reading ninety seconds of conversation, so some will be generic or miss the point of your field. You decide what to ask.

Questions people ask

Does Interview Signal suggest follow-up questions during a live interview?

Yes. After each substantial answer from the candidate, about 25 words or more, Interview Signal shows three follow-up questions under the transcript, best first. Each is tagged probe, clarify or evidence, carries a short reason, and is meant to dig into something the candidate just said. Ask this moves one onto the guide, pinned.

Why are no follow-up suggestions showing in Interview Signal?

Follow-ups only start after the candidate has spoken a stretch of about 25 words, and they need the candidate's side to be transcribed, which requires your computer's sound to be captured. They also need a connection and minutes left on a metered plan. Check the transcript first: if the candidate's lines are missing, fix the audio.

Can Interview Signal's follow-ups use the candidate's CV?

Yes. If you attach a CV on the pre-interview screen, its opening part goes with each follow-up request, so a suggestion can go after a gap between the CV and what the candidate said. The suggestion still quotes what was said in the interview, never what the CV claims.

Do follow-up suggestions change the candidate's score in Interview Signal?

No. Follow-ups are questions for you to ask or ignore. The scorecard scores what the candidate says, so a follow-up only affects scores through the answer it draws out, and every quote on the scorecard is still checked against the transcript.