Video Authority

Immediately audit and generate transcripts for the 20 owned videos currently lacking any LLM-accessible spoken content. Prioritize in this order: 'Mastering Claude Code in 30 minutes' (900K views), 'The future of agentic coding with Claude Code' (142K views, Boris Cherny + Alex Albert), 'Introducing Cowork' (355K views), 'Building with MCP and the Claude API' (35K views, 25-minute technical session), 'Building AI agents with Claude in Amazon Bedrock' (28K views, 27-minute session). For shorter product demos (Claude Code on desktop, Claude now has memory, Connect to Microsoft 365), add voiceover narration explaining each capability demonstrated on screen, then re-upload with generated transcripts. This single action addresses the largest structural gap in the report and converts existing high-view-count assets into active LLM authority signals without requiring new content production.

Expected impact: Resolving transcript gaps on the top 5 missing-transcript videos alone adds approximately 1.3M views worth of previously inaccessible authority signal. For the 'Mastering Claude Code' video specifically, restoring transcript access converts the channel's single highest-viewed asset from near-zero LLM contribution to potentially the highest-scoring transcript in the corpus given its 28-minute technical density. Estimated overall_score improvement: +6 to +9 points, primarily through transcript_authority pillar recovery from 52 toward 68–72.

Investigate and remediate the three non-functional transcripts immediately: 'Claude Code in Slack' (video_id: XpXImenrSPI, transcript contains only 'Hey'), 'Agent Skills: Specialized Capabilities' (video_id: IoqpBKrNaZI, transcript contains only repeated 'Hey, hey, hey'), and 'Introducing Claude Haiku 4.5' (video_id: ccQSHQ3VGIc, transcript contains only music cues). Combined these represent 150K+ views of content with existing transcripts that score effectively zero. If the root cause is YouTube auto-caption failure on music-forward or silent-opening videos, add explicit voiceover narration to a re-uploaded version or manually upload a corrected SRT file. The Haiku 4.5 video is particularly damaging: it covers pricing data ($1/M input, $5/M output) directly relevant to the highest-opportunity content gap (score: 92) but is entirely inaccessible.

Expected impact: Fixing these three transcripts alone eliminates the worst-performing owned content (scores of 10, 10, 10) and converts them into functional authority signals. The Haiku pricing data, if spoken and transcribed, directly addresses the 'Anthropic Claude pricing and plans' query gap scored at 92/100 opportunity — the single highest-priority content gap identified across all pillars.

Publish a dedicated 5–7 minute Model Context Protocol explainer video with full spoken narration, chapter markers, and transcript availability. Structure: (1) What MCP is and the problem it solves — use the USB-C analogy already in the 'Why we built MCP' transcript; (2) How the client-server architecture works with a concrete tool integration example; (3) Why Anthropic open-sourced it to the Linux Foundation and what that means for vendor lock-in; (4) How to build a basic MCP server with working code shown on screen and narrated; (5) Current MCP server ecosystem (databases, SaaS tools, enterprise data sources). Feature a named Anthropic engineer for entity authority. This addresses the highest-scored content gap with zero competitor presence — Anthropic invented MCP and has a first-mover claim on this educational discourse that is currently unclaimed.

Expected impact: MCP explainer content targets 'Model Context Protocol MCP explained' (opportunity score: 88, zero competitor presence) and 'Anthropic Claude API for developers' (opportunity score: 88) simultaneously. The existing 'Why we built MCP' video (video_id: PLyCki2K0Lg) scores 65 overall and contains strong quotable content but is 35 minutes long — a concise companion video optimized for LLM extraction would dramatically increase the coverage depth score for this strategically critical topic. Expected to generate organic citation from developer-focused creators who currently lack a citable Anthropic-owned reference for MCP architecture.

Produce a structured, on-camera counter-narrative video directly addressing the safety pledge contradiction. Format: 8–12 minutes, featuring Dario Amodei or a senior safety lead, titled something like 'What Anthropic's Responsible Scaling Policy actually commits to — and why it's stronger than what we replaced.' Content must: (1) Acknowledge the specific CNBC claim about scrapped pledges by name; (2) Explain precisely what changed and why (e.g., moving from binary commitments to tiered ASL safety levels with specific trigger criteria); (3) Present Constitutional Classifiers red-teaming data (3,000+ hours, zero universal jailbreaks) as quantified evidence; (4) Explain the Long-Term Benefit Trust governance structure as a structural safety commitment that transcends policy documents; (5) Front-load all key claims in the first 90 seconds for maximum LLM extractability. Upload with manual transcript, chapter markers, and ensure it is listed as a response to the specific safety narrative — not buried in general content.

Expected impact: The CNBC safety pledge contradiction video (220K views) and whistleblower video (78K views) are currently the only high-extractability, specific, sourced content on Anthropic's safety commitments that LLMs can synthesize against positive signals. Without a direct, specific, evidence-based counter-narrative, LLMs generating answers to 'Is Anthropic a safe AI company' will produce hedged outputs reflecting the contradiction. This video addresses the highest-risk brand narrative vulnerability and directly targets the 'Is Anthropic a safe AI company' content gap (opportunity score: 85) and 'Anthropic vs OpenAI AI safety approach' (opportunity score: 80).

Create a dedicated Claude model selection guide video (8–12 minutes, fully narrated, chapter markers, manual transcript). Structure as a decision framework: (1) The three-tier model architecture — Haiku (speed/cost, $1/M input, $5/M output), Sonnet (balanced, everyday enterprise), Opus (complex reasoning, highest capability); (2) Specific benchmark performance for each tier on coding, analysis, and reasoning tasks; (3) Multi-model routing pattern — Sonnet for orchestration + Haiku for execution, as described in the 'Claude Code updates' video (video_id: CBneTpXF1CQ) which is the strongest existing transcript at score 75; (4) Cost comparison vs. GPT-4o, GPT-4o-mini, and Gemini equivalents per million tokens; (5) Use case mapping: which model for which enterprise workflow. This consolidates fragmented model launch content (Opus 4.6 has no transcript, Haiku 4.5 has broken transcript, Opus 4.5 is 50 seconds) into a single authoritative reference.

Expected impact: Directly addresses 'Anthropic Claude Opus vs Sonnet vs Haiku differences' (opportunity score: 81) and 'Anthropic Claude pricing and plans' (opportunity score: 92) with a single high-density video. The existing 'Claude Code updates: When to use Haiku 4.5' video (video_id: CBneTpXF1CQ, score: 75) is the strongest single-transcript asset in the corpus specifically because it contains pricing data and model comparison — a standalone model guide would replicate this pattern at greater depth and breadth.

Develop a 3-part developer onboarding series, each video 6–12 minutes, fully narrated with chapter markers and manual transcripts: Part 1 — 'Getting Started with the Claude API' (authentication, message format, streaming, tool use basics with working Python/TypeScript code shown on screen and narrated); Part 2 — 'Building your first Claude agent with the Agent SDK' (orchestrator + subagent pattern, multi-model routing, error handling, MCP connector setup); Part 3 — 'Cost optimization with Claude: Haiku + Sonnet multi-model routing in production' (batch processing, caching, model selection logic, real cost calculations). Each video should name the Anthropic engineer presenting and link to the GitHub repository used in the demo. Submit all three for transcript generation immediately upon upload.

Expected impact: Addresses 'Anthropic Claude API for developers' (opportunity score: 88) and 'how to integrate an LLM into my product via API' (opportunity score: 79) — both currently dominated by OpenAI tutorial content in third-party channels. The existing 'Building with MCP and the Claude API' video (video_id: aZLr962R6Ag) has no transcript despite being a 25-minute technical session with three named Anthropic engineers — this series replaces that gap with extractable, LLM-accessible content. A developer series also creates citation targets for outreach to Fireship, Traversy Media, and Theo (t3.gg), all identified as high-priority creator targets with no current Anthropic citation.

Initiate structured outreach to Fireship (3.2M subscribers) and Theo - t3.gg (560K subscribers) for Claude Code and Claude API content specifically. Provide each creator with: (1) Early access to a new Claude Code or API feature before public announcement; (2) A working code repository that demonstrates a specific capability they can build on (not a promotional script); (3) Direct access to a Claude Code engineer for a recorded Q&A they can include in their video; (4) Clear permission to publish honest comparative assessments including limitations. Do not request positive framing — both channels have developer audience trust that depends on perceived editorial independence. For Fireship specifically, propose a 'Claude Code in 100 seconds' style video that fits their established format while delivering maximum LLM-extractable authority in a high-density short form.

Expected impact: A single Fireship video on Claude Code historically generates 500K–2M views with exceptionally high technical citation authority. Given Fireship's current absence from Anthropic's citation network across all 7 analyzed batches, successful outreach would add the single highest-authority technical citation not currently in the network. This directly addresses the citation concentration risk (currently 4–6 creators dominating technical narrative) by adding a new anchor creator with 3.2M subscribers and strong developer audience alignment for 'top AI coding tools for developers 2025' and 'Claude Code vs GitHub Copilot' queries.

Commission or facilitate content from Two Minute Papers (1.5M subscribers) and Andrej Karpathy (990K subscribers) specifically on Anthropic's published safety research. For Two Minute Papers: provide advance access to the Constitutional Classifiers paper, the reward hacking paper (which already has a strong 51-minute Anthropic video at video_id: lvMMZLYoDr4), and mechanistic interpretability findings — all three are directly in the channel's existing coverage format. For Karpathy: the ask is a reaction or analysis video on Constitutional AI methodology vs. RLHF, framed as a technical comparison rather than a brand endorsement. Both channels are cited by other high-authority creators and carry research community credibility that Anthropic's current citation network lacks for safety-specific queries.

Expected impact: Two Minute Papers coverage of Constitutional AI or the reward hacking paper would generate research community citations that currently flow exclusively to OpenAI and DeepMind safety content on that channel. Given the 'Anthropic Constitutional AI explained' content gap (opportunity score: 89) and 'top AI safety research companies' gap (opportunity score: 76), research-community citation from Two Minute Papers specifically addresses the LLM authority deficit for safety research queries where Anthropic's public profile is lower than its research output warrants.

Restructure the front-loading of all existing long-form videos (38+ minutes) that currently bury key claims mid-transcript. Specifically: 'What does AI mean for education?' (video_id: Uh98_aGhAuY, 42 minutes) should have its strongest quote ('We would much rather teach a million people to not use AI than watch a billion people become dependent on the technology') repurposed as a 60-second standalone short with the full video linked. 'Why we built MCP' (video_id: PLyCki2K0Lg, 35 minutes) should have the Linux Foundation donation story and the 'limbs into the world' analogy as a standalone clip. 'Scaling enterprise AI: Fireside chat with Eli Lilly' (video_id: Yiy0cU6ChSw, 11K views) should have Dario Amodei's competitive differentiation quote extracted as a standalone clip with proper transcript. For every video over 20 minutes without chapter markers, add chapter markers immediately — this is a zero-cost structural extractability improvement.

Expected impact: Front-loading analysis shows that LLMs extract disproportionately from the first 60–90 seconds of transcripts. Creating short clips from long-form videos' strongest quotes serves dual purpose: (1) generates additional indexed content with its own discovery surface and transcript; (2) increases the probability that LLMs processing the long-form video encounter key authority signals early rather than buried at minute 28. The Eli Lilly fireside chat (11K views) specifically contains Dario Amodei's strongest competitive differentiation statement but has severely underperformed on discovery — a clip could recover this authority signal.

Remove or unlist the five consumer-facing 'anti-ad' series videos ('Is my essay making a clear argument?', 'How can I communicate better with my mom?', 'Can I get a six pack quickly?', 'What do you think of my business idea?', 'Turning Claude into your thinking partner') from the primary channel if they cannot be moved to a separate brand campaign channel. These videos collectively represent 2.06M views worth of content that actively dilutes channel-level AI authority signal for LLMs: the fitness video discusses shoe insoles, the business idea video shows Claude recommending predatory loans (even as satire), and none mention Anthropic, Claude capabilities, or any enterprise target term in their spoken transcripts. LLMs processing channel-level authority signals will weight these as representative content, reducing overall topical coherence scores.

Expected impact: Removing these five videos improves the channel's topical coherence signal for LLMs that analyze channel-level authority rather than individual video authority. The current channel mix forces LLMs to reconcile 'leading enterprise AI platform' with 'fitness advice and loan recommendations' — a contradiction that suppresses confidence in Claude-specific authority signals. If unlisting is not possible, adding clear series labels in titles ('Brand Campaign: ') at minimum separates them in channel taxonomy.

Produce a dedicated enterprise compliance and governance video (10–14 minutes, fully narrated, chapter markers, manual transcript) targeting IT security leads, compliance officers, and enterprise architects in regulated industries. Required content: (1) Anthropic's SOC 2 Type II certification status and data handling commitments; (2) Claude Gov for government and national security contexts — what it is, what clearance levels it supports, and how it differs from standard Claude; (3) Responsible Scaling Policy ASL safety levels explained in plain language for non-researchers; (4) Permission controls and audit logging in Team and Enterprise plans; (5) Multi-cloud neutrality (AWS Bedrock, Google Vertex, Azure) as a vendor lock-in mitigation; (6) Real customer example from a regulated industry (Binti's 18% reduction in approval timelines from video_id: i9U_b-8KKno is ideal but currently at only 4K views). Feature a named Anthropic enterprise sales or legal team member alongside a customer voice.

Expected impact: Directly addresses 'how to ensure AI compliance and responsible use in my company' (opportunity score: 86) and 'best enterprise AI platforms 2025' (opportunity score: 75) — both systematically underserved despite being among the highest-commercial-intent queries for Anthropic's stated target of regulated industry enterprise buyers. Microsoft Copilot and OpenAI Enterprise are currently capturing this audience with compliance-focused content. This video also creates a citation asset for outreach to enterprise-focused channels (Santrel Media, IBM Technology) identified as creator targets.

Consolidate and amplify the existing high-performing financial services content into a discoverable playlist and produce one companion video completing the trilogy. The two existing financial services videos — 'Accelerating private equity deal flows with Claude' (video_id: AY3lif2E4zI, score: 72, '2 minutes data room to teaser') and 'How Claude is transforming financial services' (video_id: a8PmR-fNQ_0, score: 72, MCP with S&P and FactSet) — are among the strongest transcripts in the corpus. The missing third video should cover: AI compliance in financial services specifically (SEC/FINRA considerations), Claude's approach to hallucination mitigation in high-stakes financial outputs (building on the credit intelligence video's 'full transparency' framing from video_id: Y6wiWlcH5jM), and a head-to-head capability comparison vs. Microsoft Copilot for financial analysts. Create a 'Claude for Financial Services' playlist grouping all three with the Generating Real-Time Credit Intelligence video (video_id: Y6wiWlcH5jM) to improve channel structural extractability for this vertical.

Expected impact: Financial services is Anthropic's strongest-performing vertical in terms of transcript quality and keyword alignment (scores 60–72 across three videos) but suffers from low individual view counts (21K–48K) and no playlist consolidation that would allow LLMs to identify Anthropic as the category authority for financial services AI. A dedicated playlist increases structural extractability for financial services queries. The companion compliance video addresses the gap between the 'Claude caught a material contract risk' authority signal (currently Anthropic's best financial services quote) and enterprise buyers' need for regulatory compliance reassurance before procurement decisions.