A Watch Later list rarely gets watched. It grows one 90-minute podcast at a time until opening it feels like homework, and the videos that actually mattered are buried under the ones saved out of guilt. The workaround that has caught on is not watching faster, it is reading first: pull the transcript, skim a summary, then decide whether the video earns the full 90 minutes. We tested the best apps for AI YouTube video summaries on Android and judged them on the parts that break in practice. Whether the app takes a YouTube link straight from the Android share sheet, whether it reads the real transcript or improvises from the title and description, whether it survives an hour-long video, and whether the summary carries timestamps you can jump back to. Every pick has a free tier. None of them are accurate enough to quote without checking.
What to look for in a YouTube summary app
- Does it accept a link, or do you have to paste text? Share sheet support is the difference between a five-second triage habit and a chore you abandon in a week.
- Does it read the actual transcript? Some tools reconstruct a plausible summary from the title, description, and top comments. Those summaries read well and are frequently wrong.
- What happens to a two-hour video? Long transcripts run into context limits. The better apps chunk the text; the weaker ones quietly summarise the first stretch and stop.
- Are there timestamps or citations? A summary you cannot verify is a rumour. Timestamped bullets let you jump to the 40-minute mark and check the claim yourself.
- Does it work when captions are missing? Almost all of these tools depend on YouTube’s caption track. No captions usually means no summary, unless the app processes the video itself or you transcribe the audio another way.
- What does the free tier actually allow? Summary counts, notebook caps, and transcription minutes vary wildly, and the limits are where most of these apps hurt.
Quick comparison
| App | Best for | Takes a YouTube link | Timestamps or citations | Free tier |
|---|---|---|---|---|
| NotebookLM | Building a queue you can question later | Yes, via share sheet | Citations back to the transcript | Free, with caps on sources and notebooks |
| Google Gemini | Fastest one-off summary | Yes, pasted into chat | Sometimes, on request | Free, with daily limits |
| Glasp | A dedicated summarise-and-save app | Yes, via share sheet | Yes, timestamped sections | Free, limited summaries |
| Perplexity | Summaries you can trace to sources | Yes, pasted into chat | Yes, source citations | Free, limited deeper searches |
| ChatGPT | Reshaping a transcript into notes | No, paste the transcript | Only if the transcript has them | Free, with message limits |
| Microsoft Copilot | Microsoft 365 users | Best inside Edge | Sometimes, on request | Free |
| Otter | Videos with no captions at all | No, import the audio | Yes, timestamped transcript | 300 minutes a month |
| PocketPal AI | Keeping the transcript off the cloud | No, paste the transcript | Only if the transcript has them | Free, open source |
The apps
1. NotebookLM, best for turning Watch Later into a queue you can question
NotebookLM (renamed Gemini Notebook in July 2026, and still listed under both names in stores) is the closest thing to a real replacement for a video backlog. Share a YouTube link to it from the Android share sheet, pick a notebook, and it imports the transcript as a source alongside whatever PDFs, web pages, and notes are already in there. From that point the video is text you can interrogate: ask what the guest said about pricing, ask which of your six saved videos disagree with each other, and every answer comes back with a citation chip that opens the exact passage it came from. That citation behaviour is what separates it from a chat app. You can see whether the model actually read the line it is quoting.
The Audio Overview feature flips it back the other way, generating a two-host discussion of your sources for a commute. It is a gimmick with one genuine use: reviewing material you already skimmed.
Where it falls short: YouTube import only works for public videos that have captions, user-uploaded or auto-generated, and only the text is imported, so anything conveyed on screen is lost. Private, unlisted, and caption-free videos fail outright. The notebook structure also adds friction when all you want is one quick summary, and the mobile app trails the web version on newer features.
Pricing:
- Free: full notebook features, with caps on sources per notebook and notebooks per account
- Paid: a Google AI subscription raises those caps substantially
Platforms: Android, iOS, web
Download: Aptoide · Google Play
Bottom line: The pick if your problem is a backlog rather than a single video, and the only one here that makes a saved queue genuinely searchable months later.
2. Google Gemini, best free one-off summary
Google Gemini is the fastest route from link to summary on Android. Copy the URL, paste it into the prompt box, ask for the key points, and the answer arrives in seconds. The YouTube connection is on by default in recent versions, and if it has been switched off you can re-enable it under the app’s extensions or connected apps settings, or invoke it inline with the @ mention.
Gemini for YouTube summaries is also the most forgiving pick when captions are patchy, because it works from the video itself rather than depending purely on a caption track. Ask it to return the summary as timestamped bullets and it usually will, though the timestamps drift on longer videos and need spot checking.
Where it falls short: Nothing is saved. Close the chat and the summary is a thread you have to hunt for, which is the opposite of a triage system. The strongest model is rate limited on the free tier, and Gemini is the most confident hallucinator in this list. It will happily attribute a claim to a video that never made it.
Pricing:
- Free: daily limits on the strongest model, generous limits on the fast one
- Paid: a Google AI subscription raises limits and unlocks longer context
Platforms: Android, iOS, web
Download: Aptoide · Google Play
Bottom line: Use it for the video you are deciding about right now, not for the forty you saved last month.
3. Glasp, best dedicated YouTube summariser
Glasp is the only app here built specifically around the habit this article describes. Share a video to it and you get a summary broken into timestamped sections plus the full transcript underneath, so you can read the outline, tap a timestamp, and land at the right moment in the video. Saved videos go into a library rather than a chat log, and you can follow channels in a feed and clear a week of uploads by reading instead of watching. Highlighting works on articles and PDFs too, which makes it a reasonable read-later app in general.
Where it falls short: The free tier caps how many summaries you get, and heavy users hit that wall fast. Summary quality on rambling, unscripted video is noticeably weaker than Gemini’s, because it is working from the caption track alone and caption tracks handle crosstalk badly. It is also a smaller product than everything else here, with the support and update cadence that implies.
Pricing:
- Free: a limited number of summaries, with the transcript view included
- Paid: a Pro subscription removes the summary cap
Platforms: Android, iOS, web, browser extension
Download: Google Play
Bottom line: The best fit if you want one app that eats YouTube links and files the results, and the timestamp handling is the most reliable in this list.
4. Perplexity, best for summaries you can trace
Perplexity pulls the transcript from a pasted YouTube URL and answers with numbered citations, which makes it the easiest tool here to fact-check. Ask a follow-up question about a specific claim and it will point at the passage rather than paraphrasing again. For technical talks and interviews where you care about the exact wording of a claim, that citation discipline matters more than prose quality.
Perplexity for YouTube summaries also handles the wider question well: ask it what else has been said about the topic and it pulls in articles alongside the video, which is useful when a video is the only source you have and you do not trust it.
Where it falls short: YouTube retrieval is inconsistent. Some videos come back with a full summary, others return an apology that the content could not be retrieved, and the difference is not predictable from the outside. The free tier limits how many deeper searches you get per day. Like Gemini, it saves nothing you would call a library.
Pricing:
- Free: standard searches with a daily cap on the deeper research modes
- Paid: a Pro subscription raises the caps and adds model choice
Platforms: Android, iOS, web, browser extension
Download: Aptoide · Google Play
Bottom line: Worth having for the videos where being wrong would cost you something, provided you accept that retrieval fails often enough to be annoying.
5. ChatGPT, best for turning a transcript into usable notes
ChatGPT cannot open a YouTube link in the consumer app, and pasting one in produces a summary assembled from whatever the model already knows plus the words in the URL. That failure mode is worth knowing about, because the output looks completely plausible. The correct workflow is manual: open the video in the YouTube app, expand the description, tap the transcript panel, copy the text, and paste it in.
Once the transcript is in the chat, ChatGPT is the strongest writer here. It will restructure a two-hour conversation into an argument outline, extract just the action items, or rewrite the whole thing as a briefing for someone who never watched it. That second step, doing something with the transcript rather than shortening it, is where it beats the link-based tools.
Where it falls short: The copy-paste step kills the habit for most people. Very long transcripts exceed what fits in one message, so an hour-plus video has to be split and summarised in parts, then summarised again. Timestamps survive only if you copied them, and free-tier message limits arrive quickly on long inputs.
Pricing:
- Free: full chat access with usage limits on the stronger models
- Paid: subscription tiers raise limits and add longer context
Platforms: Android, iOS, web, Windows, macOS
Download: Google Play
Bottom line: Skip it for triage, reach for it when you need the video turned into something you will actually reuse.
6. Microsoft Copilot, best for people already in Microsoft 365
Microsoft Copilot handles video summaries competently, and its real advantage is where the summary ends up. Notes flow into the Microsoft tools you already use, so a video summary can become a OneNote page or a document draft without a detour. On Android, pasting a YouTube URL into the chat works for many videos, and the answer style is closer to a briefing than a bullet dump.
The stronger path is the desktop one. Copilot inside Microsoft Edge has a summarise-this-video action that reads the open video’s caption track directly, and it is more reliable than the mobile chat route. If you triage on a phone but write on a laptop, that split is workable.
Where it falls short: Captions are effectively mandatory, and videos without them return nothing useful. Mobile link handling is the least consistent of the three big assistants here, and the phone app’s advantage evaporates entirely if you do not use Microsoft’s ecosystem. Summaries also tend towards the corporate register, which is grating on casual content.
Pricing:
- Free: chat, image generation, and video summaries with usage limits
- Paid: a Microsoft 365 subscription adds Copilot inside the Office apps
Platforms: Android, iOS, web, Windows
Download: Aptoide · Google Play
Bottom line: Good if your notes live in Microsoft tools, unremarkable if they do not.
7. Otter, best when the video has no captions
Otter solves the problem every other app on this list gives up on. When a video has no caption track, auto-generated or otherwise, there is no transcript to summarise, and transcript-based tools return nothing. Otter transcribes the audio itself, produces a timestamped transcript with speaker labels, and generates a summary with an outline you can tap through. For recorded lectures, conference talks, and community uploads with captions disabled, it is the only route that works.
Otter for YouTube summaries means importing an audio or video file rather than sharing a link, so you need the file on your device first. It is a deliberate, slower workflow for the handful of videos that justify it.
Where it falls short: The free tier is restrictive in a way that matters here. It allows 300 transcription minutes a month, caps any single conversation at 30 minutes, and permits only three lifetime file imports, which means most people get three shots at this before the paywall. Speaker labelling degrades on crosstalk, and heavy accents or poor audio produce transcripts that need cleaning before a summary is worth anything.
Pricing:
- Free: 300 minutes a month, 30 minutes per conversation, three lifetime file imports
- Paid: a Pro subscription raises the minute cap and lifts the import limit
Platforms: Android, iOS, web, browser extension
Download: Aptoide · Google Play
Bottom line: Keep it installed for the caption-free videos nothing else can touch, and do not plan on the free tier lasting.
8. PocketPal AI, best for keeping transcripts off the cloud
PocketPal AI runs an open-source language model entirely on the phone, with no account and no network call. You download a GGUF model from Hugging Face inside the app, paste in a transcript, and get a summary produced by hardware you own. For internal recordings, private uploads, and anything under an NDA, that is the difference between using an AI summariser and not being allowed to.
It is also the only pick here that keeps working on a plane, in a basement, or on a phone with no data left. The code is open source, so the privacy claim is inspectable rather than a marketing line.
Where it falls short: It does none of the fetching for you. There is no share sheet integration, no YouTube awareness, and no transcript retrieval, so you copy the text from the YouTube app yourself. Phone-sized models have small context windows, so an hour-long transcript has to be pasted in chunks and the summaries stitched together by hand. You want 6GB of RAM at minimum and 8GB for anything comfortable, the model files run to gigabytes, and summary quality is a clear step below the cloud models. Long generations warm the phone and drain the battery.
Pricing:
- Free: the app and the models, with no subscription or account
- Paid: nothing
Platforms: Android, iOS
Download: Aptoide · Google Play
Bottom line: The right answer when the transcript must not leave the device, and a poor one for anything else.
How to pick the right one
If your Watch Later list is the actual problem, install NotebookLM and nothing else. It is the only app here that treats a backlog as a library instead of a series of one-off questions, and being able to ask a question across eight saved videos at once is worth the extra tap of choosing a notebook.
If you just want to know whether to watch this video, use Gemini. Paste, read, decide, move on.
If you want a summarise-and-save app that does one job properly, Glasp is the pick, and its timestamped sections are the most reliable way to jump back into a video mid-way.
If accuracy matters more than speed, use Perplexity and click the citations. If the video has no captions, Otter is the only option that will not simply fail. If the recording is confidential, PocketPal AI, and accept the manual work.
If you tried ChatGPT for this and got a summary that sounded right but described a video that does not exist, that is the expected behaviour, not a bug you can configure away. It cannot open the link. Feed it the transcript instead and it becomes the best writer in the group.
One rule regardless of which you pick: treat every summary as a pointer, not a quote. All of these tools invent details, most often names, numbers, and who said what. When being wrong would cost you something, open the video at the timestamp and check.
FAQ
Can an AI app summarise a YouTube video without captions?
Most cannot. NotebookLM, Glasp, Perplexity, and Copilot all depend on the video’s caption track, and they fail on videos where captions are disabled. Gemini has the best chance because it works from the video rather than a caption file alone. The reliable fallback is transcribing the audio yourself with Otter and summarising that transcript.
What is the best free app for YouTube summaries on Android?
Google Gemini, if you want one summary right now with no setup. NotebookLM, if you want summaries you can come back to, since its free tier gives you the full notebook feature set with caps on how many sources and notebooks you can keep. Both accept a YouTube link directly.
Can these apps handle a two-hour video?
Link-based tools like NotebookLM, Gemini, and Perplexity handle long videos better than paste-based ones, because they process the transcript server-side rather than fitting it into a chat message. With ChatGPT or PocketPal AI you will need to split a long transcript into chunks and summarise each part separately. Quality drops on very long videos either way, since the middle of a two-hour conversation tends to get compressed hardest.
Do AI YouTube summaries include timestamps?
Glasp and Otter produce timestamped output by default. Gemini, Perplexity, and Copilot will produce timestamps if you ask for them, though they drift on long videos and need spot checking. NotebookLM gives you citations into the imported transcript instead, which serves the same purpose of letting you verify a claim against the source.
How accurate are AI video summaries?
Good enough to decide whether to watch, not good enough to quote. All of these models hallucinate, and the common failure is attributing a claim to the wrong speaker or inventing a specific number. Citation-first tools like Perplexity and NotebookLM make errors easier to catch because you can open the source passage. Nothing here removes the need to check anything you plan to repeat.
Can you summarise a YouTube video without sending it to a cloud service?
Yes, with PocketPal AI, which runs an open-source model on the phone with no network connection. You copy the transcript from the YouTube app yourself and paste it in. Expect weaker summaries than the cloud tools, expect to work in chunks because of the small context window, and expect to need a phone with at least 6GB of RAM.
