conceptinterest-graph-recs
YouTube data-access constraints
Notes
YouTube data-access constraints
One-line summary: Watch history is closed in the Data API; candidate search is capped at 100 search.list calls/day; known-channel uploads are 1-unit playlistItems.list; Developer Policies bar a substitute YouTube browse; google-takeout is a delayed archive.
The insight
The binding constraint on a user-owned rec loop over YouTube is access and policy, not which graph database stores the interests. Public metadata is cheap; history and discovery search are not. From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs.
Evidence
- From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs:
playlistItems.listdocumentsforbidden (403) watchHistoryNotAccessible: “Watch history data cannot be retrieved through the API.” - From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs: Default allocation is “100
search.listcalls, 100videos.insertcalls, and 10,000 units per day combined for all other endpoints.” - From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs:
search.list— “100 calls per day, 1 unit per call; each extra results page is another call.” - From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs: API Clients “must not … use YouTube API Services to create, offer, or act as a substitute for, or substantially similar service to, any YouTube Applications.”
- From 2026-08-21-autoresearch-user-owned-interest-graph-youtube-recs: Takeout archives expire in about 7 days; time-range export is not supported.
- From 2026-08-21-academic-research-user-owned-interest-graphs-vs-platform-recs: Efstratiou (IMC 2025) — YouTube Data API Search “returns highly variable results between queries”; representative historical video samples are “nearly impossible,” especially off-peak.
- From 2026-08-21-academic-research-user-owned-interest-graphs-vs-platform-recs: Academic methods papers discuss APIs vs data-donation vs tracking (Ohme et al. 2023) and higher willingness to donate YouTube data than Facebook/Instagram/Google (Pfiffner et al. 2023). They do not restate
watchHistoryNotAccessibleor measure Takeout delay. - From 2026-08-21-autoresearch-best-architecture-for-a-rec-system-web-app: Official serving path for a known channel is
channels.list→contentDetails.relatedPlaylists.uploads→playlistItems.list(maxResults 50, 1 unit/page) — notsearch.list. - From 2026-08-21-autoresearch-best-architecture-for-a-rec-system-web-app: Policies re-fetched: API Clients “must not recreate the browse experience from any YouTube Application without adding significant independent value to that flow.”
- From 2026-08-21-autoresearch-best-architecture-for-a-rec-system-web-app: Private user/channel data needs OAuth; public metadata can use an API key — a separate auth plane from graph ownership (local-first-software, solid-project).
Contradictions / tensions
- A tool that filters the user’s own Takeout archive against a graph they drew is a different fact pattern than a live YouTube-like homepage. No fetched YouTube ruling on that case.
Related
Referenced by
Mechanisms
Questions
Entities