Notes
Notes
One finding per note, each answered in a few sentences first. The numbers come from The Dark Archive, our study of long-form YouTube channels, or from a named outside source. New terms are in the glossary.
What The Dark Archive found
- How machine-readable are creator websites?
We scored 528 creator websites on machine readability out of 100. The median score was 20, and none scored above 70. Why, and how to raise your own score.
- How we chose 1,284 channels for The Dark Archive
We filtered 14,713 candidate channels down to 1,284 using six strict criteria. How we built the sample, and which channels its findings speak for.
- Is a link-in-bio page enough for a YouTube channel?
A link-in-bio page routes fans, but gives readers and AI engines nothing of yours to quote. What an owned written archive adds, and how to get one built.
- The Dark to Sovereign ladder: what separates each rung
We mapped 1,206 long-form YouTube channels across five rungs from Dark to Sovereign. What separates each rung, and how your channel climbs to the top.
- 121,294 hours: the unwritten archive of long-form YouTube
In 24 months, the 1,284 channels in our study published 121,294 hours of long-form video. Almost none of it exists as text a reader or an AI engine can quote.
- What a website does for a YouTube channel
Creators who run a site report three returns: direct contact, off-platform sales and a second way to be found. What separates the sites that pay off.
- Does your channel need a website? A quick qualifying test
Before you build a site, run this test: the signs a website would waste your time, and the signs it pays off. Channels with evergreen videos pass it.
- Why podcasts go Dormant and video essays stay Dark
In our study of 1,284 channels, podcasts were the niche most often Dormant and video essays the one most often Dark. Why the two sit at opposite ends.
- 0.83%: how many YouTube channels keep a written archive
Of 1,206 long-form YouTube channels we could measure, 10 republish their videos as text on a site they own. Why that gap is your opening in AI search.
- What blocking AI crawlers costs a creator's site
Most creator sites block no AI crawler: 476 of the 528 we scored. What a block changes in ChatGPT, Google and Claude search, from the vendors' own docs.
How AI engines read creator sites
- AI citations name the source. Make sure the source is you.
AI answers send few clicks, one traffic report shows. Their value is credit: being named as the source. How to make your own site that named source.
- Is a transcript enough to turn video into text?
A raw transcript is spoken text, not an article. In one publisher's test, articles built from transcripts were cited on every platform tested. What they add.
- llms.txt is hygiene. The written archive is the lever.
Almost no llms.txt files get traffic, Ahrefs found. Ship one as hygiene, then build what earns AI citations: a written archive of your videos on your own site.
- Schema labels the page. Written answers earn the citation.
A matched Ahrefs study found no clear citation lift from adding schema. Schema labels a page; written answers earn the citation. Here is what to build first.
- AI engines cite YouTube. Make your own domain citable too.
YouTube ranks among the most-cited domains in AI answers, and small channels get cited too. The catch: the clicks land on YouTube. How to own the source.
- Which AI engines cite video, and how to reach the rest
Perplexity and Google cite YouTube often. In one controlled test, ChatGPT, Claude and Copilot cited only the written versions. How to get quoted on both.
Follow new notes by RSS.