Full YouTube Channel Audit (Advanced)

Process

1. Resolve channel: accept @handle, /channel/UC..., /c/name, or a video URL. Normalize to channel ID via API or WebFetch.
2. Ask for Studio exports before starting any retention or CTR analysis:
   • Analytics → Advanced Mode → Last 28/90 days CSV for:
     • Content (video-level stats)
     • Traffic source
     • Audience → geography + top subtitles
     • Key moments (retention curves for top 10 videos) Without these, score only metadata and structure; label retention metrics as "unknown".
3. Fetch channel surface: /about, /videos, /shorts, /playlists, /community, /featured. Extract:
   • Channel name, handle, description, country, links, join date
   • Banner and profile image presence
   • Total videos, views, subs (if public)
   • Topic categories (from API channels.list?part=topicDetails)
   • Upload cadence from last 25 long-form + last 25 Shorts
4. Sample videos: pull the last 25 uploads and the top 25 all-time by views. For each: title, description, visible tags, view count, like count, comment count, duration, upload date, thumbnail URL, chapters.
5. Delegate (parallel via Agent tool when available):
   • youtube-seo-channel — identity, about, playlists, trailer, session chains
   • youtube-seo-video — run on top 3 + bottom 3 performers (contrast analysis)
   • youtube-seo-thumbnail — review last 12 thumbnails as a grid + SERP differentiation
   • youtube-seo-keywords — topical authority score + gap analysis
   • youtube-seo-competitor — detect 3-5 direct competitors + pattern extraction
6. Compute topical authority: cluster the channel's videos by topic and measure concentration (entropy). High concentration = high authority on that cluster. Diffuse = weak algorithmic classification.
7. Compute velocity consistency: stddev of days-between-uploads. Low stddev + regular cadence = Browse feed promotion friendly.
8. Compute session chain strength: % of videos that appear in at least one playlist; % of playlists with ≥5 videos; avg playlist view count.
9. Score: aggregate into YouTube Health Score (0-100)
10. Report: write YT-AUDIT-REPORT.md + YT-ACTION-PLAN.md to the current directory.

Scoring Weights

If Studio CSV is missing, Retention & Watch Time drops to 0 weight and other weights rescale proportionally — the report flags this explicitly.

Report Structure

Executive Summary

• Overall YouTube Health Score (0-100) with CSV-available flag
• Channel stage: Starter (0-1k) / Growing (1k-100k) / Established (100k+)
• Niche detected + topical authority concentration score
• Top 5 critical issues with estimated impact
• Top 5 quick wins with expected lift

Retention & Watch Time (only if Studio CSV provided)

• APV distribution across sample (median, p25, p75)
• Retention-cliff inventory: which videos have cliffs, where, why
• Intro-retention (0:30) median vs niche benchmark
• Session contribution: which videos drive next views
• End-of-video drop analysis

Traffic Source Mix (from Studio CSV)

• % Browse / Suggested / Search / External / Notifications / Shorts feed
• Which surface is over/under-indexed vs niche norm
• Strategy implication: optimize for the dominant surface, then diversify

Video Metadata (aggregate)

• Title patterns: avg length, keyword-front usage, hook types, emotional intensity
• Description: above-fold usage, avg length, entity coverage, link block presence, chapter presence
• Tags: avg count, first-tag alignment with title
• Chapters: % of videos with ≥3 starting at 0:00
• Captions: manual vs auto, languages translated
• Key Moments schema eligibility

Thumbnails

• Style consistency across the last 12
• Text legibility at 120px
• Face / emotion usage vs niche
• Color palette analysis (dominant colors across the grid)
• SERP differentiation sample for top 3 target keywords
• Specific thumbnails to redesign

Topical Authority

• Topic cluster map (from video titles, tags, transcripts)
• Concentration score (0-100, higher = stronger classification)
• Cluster drift: is the channel becoming more or less focused over time?
• Recommended consolidation: playlists to build, videos to unlist or reclassify, topics to stop covering

Content Strategy

• Upload cadence: median days-between, stddev, regularity score
• Format mix: long-form / Shorts / live %
• Length distribution: median, with mid-roll ad eligibility %
• Series / playlist alignment
• Premiere usage

Channel Setup

• Delegated to youtube-seo-channel

Engagement Snapshot

• Like-to-view ratio trend
• Comment velocity and creator reply rate
• Pinned-comment usage and reply rate
• Community tab activity (last 30 days)

Discoverability

• Target keywords present in titles/descriptions
• Translated metadata for top 3 audience countries
• Caption languages available
• Hashtag strategy

Velocity & Freshness

• First-24h view velocity distribution for last 25 videos
• Videos decaying fast vs holding evergreen
• Evergreen:topical mix recommendation

Safety & Compliance

• Made-for-Kids flag consistency
• Altered/synthetic content disclosure usage (for AI content)
• Copyright claims on top videos
• Community guideline strikes
• Monetization status

Priority Definitions

• Critical: Blocks discovery, damages CTR, or breaks algorithmic classification (fix this week)
• High: Significantly limits ranking or watch time (fix this month)
• Medium: Optimization opportunity (fix this quarter)
• Low: Nice to have (backlog)

Each recommendation includes estimated impact and the Studio metric to watch after implementation.

Data Sources Policy

First-party only. YouTube Data API v3, YouTube Studio CSV exports (user-provided), yt-dlp, YouTube suggest API, WebFetch on youtube.com, and the local helper scripts in scripts/. Do NOT pull data from SocialBlade, VidIQ, TubeBuddy, NoxInfluencer, HypeAuditor, or similar third parties — they rate-limit and error, which blocks the audit.

The seo-dataforseo MCP is OPTIONAL. If it is installed AND returns data cleanly, it may supplement SERP position checks. On any error, skip it silently and continue with native sources — do not retry.

Error Handling

Core principle: degrade gracefully, score what you can, flag what you can't, never fabricate Tier 1 numbers.