Claude Can't Summarize a YouTube Video? Here's Why and What Works

Quick answer: Claude cannot watch a YouTube video, and pasting a URL will not produce a summary of its contents. This is an architectural limit, not a bug or a setting you have missed — the model processes text and images, not video streams. The workflow that does work is to fetch the transcript yourself and paste the text in, at which point Claude becomes genuinely good at the task.

What actually happens when you paste a link

One of three things. Claude tells you plainly that it cannot access the video, which is the most useful outcome. Or it asks you to paste the transcript. Or — the case to watch for — it produces something that reads like a summary but was assembled from the URL slug, your own framing of the question, and general knowledge about the topic. That third outcome is the one that misleads people, because a confident summary of a video nobody read is indistinguishable from a real one unless you already know the content.

The tell is the absence of specifics. Real transcript summaries contain the speaker's actual claims, numbers and examples. Reconstructed ones describe what a video with that title would probably say.

Why this is not going to be fixed by a setting

Video summarization requires either video input or transcript retrieval. The first is a model capability question; the second is a tooling question. A chat assistant with web access can sometimes fetch a page, but YouTube's caption tracks are not simply sitting in the page HTML — they are served separately, which is why dedicated summarizers exist as a category at all. If retrieval were trivial, every assistant would do it well and this article would not need to exist.

The workflow that works

1. Open the video and click the three-dot menu underneath it.
2. Choose Show transcript. If the option is missing, the video has no captions and no tool can summarize it.
3. Use the transcript panel's three-dot menu to toggle timestamps off — this saves you deleting hundreds of lines by hand. Full method in removing timestamps from a YouTube transcript.
4. Select all, copy, and paste into Claude with an explicit instruction.

Claude's large context window is a real advantage here: a three-hour podcast transcript of roughly 30,000 words fits comfortably in a single message, which means you get a summary of the whole thing rather than the first third. That is a genuine edge over tools that chunk aggressively, and it is worth the manual step for a video that really matters.

Prompts that produce better output than "summarize this"

• For an argument-led video: "Give me the central claim, the three strongest pieces of supporting evidence, and any point where the speaker concedes uncertainty."
• For a tutorial: "List the steps in order, with the exact settings or commands mentioned. Flag anything the speaker says is optional."
• For an interview: "Summarize by speaker. What did each person actually assert, and where did they disagree?"
• For triage: "In three sentences, tell me whether this is worth watching in full and who it is for."

Where the manual route stops being worth it

The copy-paste workflow costs four to six minutes per video once you include finding the transcript, stripping timestamps, and pasting. For one important video that is nothing. For a backlog of twenty, it is two hours of clerical work, and the friction means most people simply stop doing it — which is the real cost, since the videos then go unwatched and unsummarized. We put numbers on this in why copying transcripts by hand wastes more time than you think.

YT Summarizer collapses those steps into pasting a URL: it retrieves the transcript, processes all of it, and returns an overview, key points and takeaways in about a minute, saved to a searchable library. Five summaries free, then one-time credit packs from $19 with no subscription and no expiry.

See also: YouTube summarizer vs Claude for a direct comparison of the two workflows.