Archiving a YouTube Channel as a Knowledge Base: Why Creators Lose 80% of Their Own Insights

Archiving a YouTube channel as a knowledge base means transcribing the entire back catalogue, indexing every transcript passage with its timestamp, and putting a question-answering interface on top — so that anyone, including the creator, can retrieve any specific statement from any upload in seconds instead of never. The problem it solves is brutally simple: the average channel in SageTube’s production index holds about 251 videos, or 67 hours of spoken content, and nobody — not the audience, not the creator — can hold that in memory or scrub their way through it. The catalogue keeps existing, but it stops being retrievable, which in practice means most of its insights are lost to everyone including the person who made them. This post walks through what that loss looks like, and how a searchable, self-updating, shareable archive reverses it.

How big is a typical channel archive, really?#

Numbers from our own production index, queried on 2026-07-22. SageTube currently tracks 46,528 videos across 183 YouTube channels, of which 20,247 are fully transcribed and searchable. The catalogue works out to roughly 12,250 hours of video. Per channel, the averages are the interesting part: 251 videos and about 67 hours of content on the average channel we index. The largest single channel in the index holds 6,065 videos on its own.

Sixty-seven hours is not a number a human works with. It is more than a full-time work week and a half of continuous listening — for one channel, assuming you never sleep and never rewind. A fan who wants “that thing they said about cold exposure, I think it was a couple of years ago” is not going to find it by watching. The creator who said it is not going to find it either, and that is the part that gets too little attention.

Why does a back catalogue stop paying its owner?#

Because retrieval cost, not content quality, decides whether old work gets used. A six-year-old video that perfectly answers today’s question is worth nothing if nobody can surface it — and on a 251-video channel, almost nobody can.

Call it the 80% problem. Be generous and assume a creator can actually recall — well enough to find again — roughly one upload in five: the recent ones, the hits, the handful they’re proud of. Everything else, call it 80% of the catalogue, is functionally gone: the creator can’t cite it, YouTube’s own search won’t surface what was said inside it (we covered why in the time-scarcity pillar), and fans certainly can’t find what the maker themselves cannot. That fraction is an illustration, not a measurement — but whatever the true number is on your channel, it grows with every upload, because your memory does not scale and your catalogue does.

The failure shows up in painfully concrete ways:

• A viewer asks a question in the comments that the creator answered thoroughly in 2022. The creator paraphrases a worse version from memory, or ignores it.
• The same questions arrive over and over — because for a new viewer, the archive may as well not exist.
• The creator wants to reference their own earlier argument in a new script and spends an evening scrubbing timelines looking for it.
• A journalist, sponsor, or collaborator wants to know “has this person covered X?” and there is no way to answer short of watching.

None of this is a content problem. The knowledge was recorded, edited, published, and is sitting in public. It is purely a retrieval problem — and retrieval problems are solvable.

What does “archiving a channel as a knowledge base” actually mean?#

In SageTube, the unit is an Expert: a knowledge base you assemble from YouTube sources — single videos, playlists, or entire channels. Building one from a channel does three things to the catalogue:

Every video gets a transcript. Videos with usable captions take the fast, cheap path — the captions are fetched directly. Videos without captions are not skipped: the audio track is extracted and transcribed with a speech-to-text API. The two paths are priced separately and honestly: caption-based indexing costs $0.01 per minute of video, audio-path transcription $0.05 per minute (config/subscription.php, lines 64–65).

Every transcript passage keeps its timestamp. The transcript is split into passages that remember exactly which seconds of video they came from, so an answer can point at the moment, not just the video. This is the grounding model the whole product rests on — every answer cites the video and timestamp of each claim, and clicking the citation takes you to the moment it was said. Why that citation discipline matters more than the answers themselves is the subject of its own pillar: Citations, Timestamps, and Trust.

The catalogue becomes something you interrogate. Instead of “which video was that in?”, the question becomes the question itself: “What does this channel recommend for beginner deadlift programming?” — answered across every indexed upload at once, with citations telling you which video and which minute each part came from. For a channel with 251 videos, that is the difference between an archive and a landfill.

How does the archive stay current when new videos ship?#

An archive frozen at build time would start rotting immediately — the channels worth archiving are the ones still publishing. This is what Live Knowledge Pulse exists for. Each channel attached to an Expert carries its own sync frequency — daily, weekly, or monthly (app/Services/PulseSyncService.php, lines 242–253) — and a scheduled sweep picks up every channel due for a sync and indexes whatever was uploaded since the last one (app/Services/PulseSyncService.php, lines 40–56). The knowledge base you built in January answers correctly about the video published yesterday, with no manual re-import.

Pulse also closes the loop with the audience: after a sync lands new videos, a digest email goes out to the Expert’s subscribers summarizing what was added — batched with a short delay so a multi-video upload day produces one email, not five (app/Jobs/ProcessPulseDigest.php, lines 39–40). As of this writing, 22 channel–Expert pairs in production sync on a Pulse schedule; it is a young feature, and the digest is the part subscribers actually see.

How do fans use an archived channel?#

This is where archiving stops being a private convenience and becomes an audience feature. A creator can publish an Expert as a public page: SageTube issues a share token, and the Expert becomes reachable at https://sagetube.ai/s/<token> (routes/web.php, line 331). Any visitor can open that page and start asking the catalogue questions directly — the page resolves the token, checks the Expert is active, and serves a chat interface with no visitor account required (app/Http/Controllers/SharedExpertController.php, lines 18–20).

The visitor experience is the fan-mail scenario inverted. Instead of leaving a comment and hoping, a fan asks the archive itself: “What was the supplement protocol from the episode with the sleep researcher?” — and gets the answer with a clickable timestamp into the exact episode. The creator answers thousands of questions per day without being present for any of them, and every answer links the fan back to the original video.

Visitors who want to stay current can subscribe to the Expert by email right on the public page (routes/web.php, line 256; app/Http/Controllers/PulseSubscribeController.php, lines 16–34) — that is the same Pulse digest described above, pointed outward at the audience. You can browse real public Experts on our Explore page to see what this looks like from the visitor’s side.

Can an archived channel earn money?#

Yes — the archive can be a product, not just a service. Creator monetization on public Experts runs through Stripe Connect: a creator connects a Stripe account, sets pricing on their Expert, and paid visitor queries are split at charge time — the creator’s share is routed to their connected account via Stripe’s transfer_data mechanism (app/Http/Controllers/ExpertCheckoutController.php, lines 83–86, 153–154). The platform fee is 30%, meaning creators keep 70% of what their archive earns (config/subscription.php, line 81).

The pricing mechanics are deliberately simple. Visitors get a free allowance to try the Expert — 3 free queries by default (config/subscription.php, line 82) — and can then buy query packs of 5, 20, or 50 (config/subscription.php, line 93). Per-query pricing is set by the creator with a floor of $0.15 (config/subscription.php, line 88), so even the smallest pack stays meaningfully above payment-processing fees. For a creator with a deep, high-trust catalogue — a fitness coach’s 400 workout breakdowns, a mechanic’s decade of repair walkthroughs — this turns the 80% of the archive nobody could reach into inventory.

Which plan does channel archiving need?#

The build-and-ask part works on any tier — what varies is scale and automation:

• Free covers a taste: 1 knowledge base, up to 3 sources, and a $5.00 starting balance to index with (config/subscription.php, lines 52–57). Enough to archive a small channel or a favorite playlist and see whether the answers hold up.
• Builder ($9/month) raises the ceiling to 2 knowledge bases and 5 sources (config/subscription.php, lines 9–10) but, like Free, does not include Pulse auto-sync (config/subscription.php, lines 16 and 60) — archives stay manual.
• Creator ($29/month) is the archiving tier: 30 knowledge bases, up to 200 sources (config/subscription.php, lines 23–24), and no fixed cap on Pulse auto-sync channels — sync cost is governed by your wallet balance and bounded by the 200-source ceiling rather than by an arbitrary channel count (config/subscription.php, lines 30–33).
• Enterprise ($99/month) removes the knowledge-base and source limits and applies a 20% discount to usage rates (config/subscription.php, lines 40–47).

Indexing cost scales with catalogue length at the per-minute rates above, and it is a one-time cost per video — once a video is in the index, every future question against it is a query, not a re-transcription.

Where should a creator start?#

Not with the whole catalogue. The pragmatic path is: build an Expert from the channel’s most question-generating series first — the how-to library, the Q&A episodes, the flagship course — and put the public share link where the questions already arrive: the channel’s About page, video descriptions, the pinned comment. If the archive starts answering real questions, extend it backwards through the catalogue and turn on Pulse so it never goes stale again. The back catalogue took years to make; making it retrievable is a weekend decision.

The broader case — why video knowledge is locked up in the first place, and what retrieval-augmented generation does about it — is laid out in the companion pillar, The Time-Scarcity Problem. Or skip the theory and start with your own channel.

Every product claim in this post is tied to the SageTube codebase as of July 2026: indexing rates (config/subscription.php:64), Pulse sync frequencies and the due-for-sync sweep (app/Services/PulseSyncService.php:242, :40), the batched subscriber digest (app/Jobs/ProcessPulseDigest.php:39), the public Expert page and its token check (routes/web.php:331, app/Http/Controllers/SharedExpertController.php:18), on-page email subscription (routes/web.php:256, app/Http/Controllers/PulseSubscribeController.php:16), the Stripe Connect split and 30% platform fee (app/Http/Controllers/ExpertCheckoutController.php:83, config/subscription.php:81), visitor free queries and query packs (config/subscription.php:82, :93, :88), and per-tier limits (config/subscription.php:9–60). Index statistics — 183 channels, 46,528 tracked videos, 20,247 transcribed, an average of 251 videos (~67 hours) per channel, and 22 Pulse-synced channel pairs — were queried from the production database on 2026-07-22.