You are watching a ninety-minute investor call because somewhere in it, someone said something about 2027 guidance. You do not want a summary of the call. You do not want the ten most notable exchanges. You want that one answer, and everything else is in the way.
Summaries are built on an assumption that quietly fails here: that the AI can work out what matters to you. It cannot. It produces what matters on average — which is exactly right for studying a lecture and exactly wrong when you arrived with a question already formed.
Ask inverts the process
With Ask, you type the question and the AI answers it from the video's transcript. Not a general summary you then skim for your answer. The specific thing you asked, in a sentence or two.
"What did she say about the 2027 guidance?" "Which database did they migrate to, and why?" "Did anyone give a date?"
How to use it
- Analyse a video as normal — paste the URL at xyzyoutube.com, or type
xyzin front of it. - On the results page, find the Ask a Question card in the next actions area.
- Click it. A panel opens with a text box.
- Type your question — up to 1000 characters — and click the Ask button.
Worth knowing: pressing Enter in that box adds a line break rather than submitting. Use the button.
Ask requires a Core plan or above, and each question draws credits from your balance the same way every other analysis does.
Where it beats every other format
Verifying a claim before you repeat it. Someone forwards a video with a strong assertion in the description. Ask what the video actually says about it. This takes thirty seconds and has saved people from confidently repeating things that turned out to be the uploader's interpretation rather than the speaker's words.
Research across many videos. When you are working through eight candidate sources with one question in mind, asking each one directly is dramatically faster than summarising all eight and reading for the answer. This is the workflow where Ask earns its plan, and it pairs naturally with triaging a pile of saved videos.
The detail a summary would never include. Which drill bit, which library version, which year. Correctly deprioritised by any summarisation format, and often the only reason you opened the video.
Following up on a summary. Read an executive summary, notice the one thing it raised but did not resolve, ask about it directly. This is the most natural pairing and how most regular users work.
Asking well
The quality of the answer tracks the specificity of the question, more than people expect.
"What is this about?" gets you a worse summary than the summary formats would have produced. That is not the tool failing — you asked a summarisation question of a retrieval tool.
"What reason did he give for missing the deadline?" gets you a precise answer, because it tells the model exactly what to find. Name the thing you want. Quote the term the speaker would have used if you know it.
The honest limits
Ask reads the transcript. If something was never said aloud, no phrasing of the question will retrieve it — a chart on screen that the presenter waved at without describing is not in the transcript and cannot be in the answer.
It also does not carry a conversation. Each question is its own analysis, not a chat thread with memory of what you asked before. If your follow-up depends on the previous answer, include the context in the new question rather than assuming it carried over.
And the usual caution: answers inherit auto-caption errors. When the answer is a number or a name and you plan to use it publicly, check that moment in the video.
Ask or Q&A pairs?
Use Q&A pairs to study a video properly — the AI generates the questions, covering the material broadly, and the output works as flashcards. Free on any plan.
Use Ask when you already know what you want and everything else is noise. Core plan and above.
Next time you open a long video hunting for one fact, skip the scrubbing. Ask it directly.