Ask your evidence
Every answer is composed from the sources currently available and cites what it retrieved. Transcript Q&A works first; active visual evidence adds screen and frame citations without changing earlier answers.
- Updated
- 30 August 2026
- Read
- 6 min
- For
- Everyone; works on every plan
Four scopes, from one recording to the whole workspace.
Pick the smallest scope that could contain the answer; tighter scope, sharper retrieval.
Ask from the recording page. Best for "what happened in this session" questions: what failed, what was agreed, what did the customer object to.
Ask from the document page. Answers cite the page and section they were read from.
Ask across everything filed in a project - recordings and documents together. One question about a client returns spoken evidence by timestamp and written evidence by page.
Ask across sources you select from the whole workspace, when the answer could be anywhere.
What a citation carries.
This is what makes an answer verifiable instead of plausible.
The verbatim words the answer rests on, not a paraphrase of them.
Who said it, by name when the speaker is known.
The second it was said. One click seeks the player there.
After optional visual evidence is ready, a visual citation can carry the exact frame - an error toast, a form state, or a dashboard value.
Document citations name the page and heading they were read from, scanned pages included.
Each answer shows how well supported it is, so a thin answer looks thin.
One question, one checkable answer.
"What did the customer say about SSO?"
Citation: Priya N. at 22:41 - "we can pilot without SSO, but procurement will block the wider rollout until it is there". If this recording has ready visual evidence, the answer can also cite the requirements slide shown on screen.
Click the citation, hear the sentence, see the slide. If someone later remembers the commitment differently, the dispute takes thirty seconds to settle.
Four habits that get better answers.
- Ask about facts, not opinions
"What did the tester say when checkout failed?" retrieves evidence. "Was the meeting productive?" asks for a judgement the record cannot support.
- Name the thing
Use the words people actually said where you can - feature names, error text, clause numbers. Retrieval searches meaning and exact wording together, and exact wording is precise.
- One question at a time
Compound questions dilute retrieval. Ask the follow-up as a follow-up; the conversation keeps its context.
- Then open the citation
The answer is a claim; the citation is the proof. For anything that matters, click through and read or watch the source. That is the workflow the product is built around.
If an answer comes back with no supporting evidence, the page says so and suggests rephrasing or waiting for indexing to finish - see troubleshooting for the full list of states.
Evidence Q&A, answered.
- It means retrieval found nothing that supports an answer, and the system declines rather than guessing. That is information: the record is silent on the point. If you believe the evidence exists, rephrase using the words that were actually used, or check that processing has finished.
- Yes. Questions run as conversations: each follow-up keeps the context of the ones before it, and past conversations are saved so you can return to them.
- Each plan includes Evidence Q&A questions per seat per month - 50 on the free plan, more on paid tiers - pooled across the team and shared between recording and document questions. Reading transcripts and recaps never costs questions.
- Yes, after that recording has active visual evidence. Before then, recording answers use transcript and transcript artifacts; a screen-dependent question explains that the screen has not been analyzed and offers the visual-evidence action.
- Send it, citations included, to Slack or Notion, or to your own systems through a webhook. The answer arrives with its quotes and timestamped links, not as a bare paragraph.
- Yes - search covers transcripts, findings, documents, and on-screen text for recordings whose visual evidence is active. It is often faster when you know roughly what you are looking for and just need the moment.