The playbook that makes YouTube research good.

---
name: tubefield
description: "Use TubeField (tubefield.com) for YouTube-native research: search videos, inspect metadata, read transcripts, sample public comments, and analyze playlists or channels when tools are available. Use when the user asks what YouTube users, commenters, creators, reviewers, fans, customers, or audiences say or think; asks to summarize, quote, compare, fact-check, monitor, or analyze YouTube videos/channels/playlists; asks for product, brand, market, launch, trend, or sentiment research where the discussion is happening on YouTube; or when web search would miss spoken video content and comment-section evidence."
---

# TubeField Research

Use TubeField when YouTube is the evidence source. It complements web search: use web search for official facts, specs, pricing, policy, filings, and news; use TubeField for what creators said and how viewers reacted.

## Capabilities

TubeField can help you:

- Find relevant YouTube videos for a topic or entity.
- Inspect video metadata such as title, channel, publish date, views, likes, comments, duration, and tags.
- Read timestamped video transcripts for creator claims, quotes, summaries, and fact checks.
- Sample public top-level comments for viewer sentiment, objections, praise, comparisons, jokes, and purchase intent.
- Inspect playlists or channels when those tools are exposed.
- Build sourced datasets, briefs, and dashboards from YouTube evidence.

## Tool Use

If TubeField tools are not visible, use tool discovery for `TubeField YouTube search comments transcript`.

Use only the tools exposed in the current session. Common TubeField tools include:

- `search_videos`: discover candidate video ids for a topic.
- `search_result`: search plus detail and transcript for a small enriched result set.
- `video_detail`: get metadata and public statistics for a video.
- `video_transcript`: get timestamped spoken transcript.
- `video_comments`: get public top-level comments with author, text, like count, and timestamp.
- `playlist_videos`: list videos in a playlist.

Typical sequence:

1. Extract video or playlist ids if the user provided URLs; otherwise search.
2. Shortlist sources by relevance, recency, diversity, and reach.
3. Pull details for context.
4. Pull transcripts when creator content matters.
5. Pull comments when audience reaction matters.
6. Synthesize patterns, disagreements, and limits.

## Sampling Defaults

TubeField is inexpensive, so prioritize outcome quality over excessive call minimization.

- For one video: use detail plus transcript; add comments if reception matters.
- For audience sentiment: sample 5-8 relevant videos and 50-100 comments per video.
- For a stronger product, brand, or launch read: sample 8-12 videos across reviews, comparisons, launch reactions, and long-term coverage.
- For competitive research: keep samples balanced across entities and compare the same theme categories.
- For datasets or dashboards: preserve video id, URL, title, channel, publish date, pull date, comment text, comment likes, comment date, transcript timestamps, theme, sentiment, and evidence type.

Scale down when the user asks for a quick answer, the topic is narrow, results are sparse, or more sources would be duplicative.

## Answer Standards

- Lead with the answer, then support it with evidence.
- Distinguish creator claims, viewer reactions, researcher inference, and external facts.
- Link sampled videos as `https://www.youtube.com/watch?v=<video_id>`.
- Include source dates when recency matters.
- State the sample and method briefly, especially for sentiment claims.
- Use "sampled comments suggest" rather than treating comments as a representative poll.
- Quote transcripts or comments only when useful; keep quotes short and paraphrase the rest.
- Mention missing transcripts, disabled comments, sparse data, search ambiguity, or obvious sampling bias when it affects confidence.