That’s just one of five Claude workflows we’re going to cover in this post.
TLDR — here’s what each workflow does:
- Outlier research — proven video topics, each already pulling 10x its channel's average, clustered and ranked by fit for yours.
- Comment mining — the questions your audience keeps asking that nobody has answered, so your video covers what the others missed.
- Idea validation — a make-it or skip-it verdict backed by real demand and competition data, before you spend a day producing the wrong video.
- Transcript analysis — the gaps every top video on your topic leaves open, and the angle that makes yours the one worth watching.
- Performance auditing — a ranked list of what to fix first, highest-impact change at the top.
The thing that makes all of this possible is connecting Claude to VidIQ, giving one of the most capable AI tools direct access to real-time YouTube data. Together they turn hours of work into minutes.
Read (or watch) below to understand how to get this setup, and what to prompt Claude to initiate these workflows.
In our YouTube video, we used simplified versions of these workflows to keep things moving. The workflows in this blog are the full versions we recommend using.
They’re longer for three reasons:
- They give Claude a consistent research method
- They define exactly what the final output should include
- They set clear rules for how it should use evidence, handle uncertainty, and draw conclusions
Looking for an expert YouTube Agency? Get in touch today. 
Setup: build your channel brain
Before the fancy stuff, you need to lay a foundation. Because if you’re just opening Claude and firing off random prompts every time, you’re getting maybe 5% of what it can do.
What you want instead is a Claude Project — a workspace that deeply knows who you and your business are, what your channel is about, and who your audience is. Think of it as your channel’s brain.
Already have a project with VidIQ connected? Skip ahead to Workflow 1.
Step 1: Create the project
Open Claude, go to Projects, and create a new one. In the project instructions, feed it the context it needs to actually be useful:
- Your channel name, niche, and positioning
- Your target audience
- Your content pillars
- Your tone of voice
- Your goals
Copy and fill out the template below to use in your Claude Project. This will give you a great starting point.
MY CHANNEL Channel Name and URL: [NAME — URL] Niche: [BE SPECIFIC] POSITIONING What makes this channel different: [1-2 SENTENCES] What we will never make videos about: [AS IMPORTANT AS WHAT WE DO COVER] AUDIENCE Who they are: [ROLE, EXPERIENCE LEVEL, CONTEXT] What they want: [THE OUTCOME THEY’RE AFTER] What they already know: [SO WE DON’T EXPLAIN THE BASICS] What they doubt: [OBJECTIONS AND EXISTING BELIEFS] CONTENT PILLARS [PILLAR] — [HOW / WHY IT SERVES THE AUDIENCE] [PILLAR] — [HOW / WHY IT SERVES THE AUDIENCE] [PILLAR] — [HOW / WHY IT SERVES THE AUDIENCE] TONE OF VOICE We sound like: [3-5 ADJECTIVES + EXAMPLE PHRASING] We never sound like: [e.g. no hype, no "in today's video"] GOALS AND CONSTRAINTS Primary goals: [e.g. qualified leads / subscriber growth / watch time] How we measure it: [SPECIFIC METRIC AND TARGET] Capacity and limits: [TEAM SIZE, BUDGET, e.g. no on-location shoots] MY BUSINESS (If your channel isn’t connected to a business, remove this section.) Business name and website: [NAME — URL] What the business does: [1–2 SENTENCES] What we sell: [PRODUCTS OR SERVICES] Ideal customer: [WHO THEY ARE AND WHAT THEY NEED] Why customers choose us: [KEY DIFFERENTIATORS] How YouTube supports the business: [e.g. GENERATE LEADS / BUILD AUTHORITY / DRIVE SALES] HOW TO WORK WITH ME - Use vidIQ for all available YouTube data. Never present an estimate as a measured figure, and don’t estimate something you can look up. - If a data source is unavailable, say so rather than guessing. - Be direct. If an idea is weak, say it’s weak and explain why. I’d rather hear “don’t make this” than a hedged maybe. - Cite specifics whenever possible — titles, view counts, timestamps, and quotes — so I can verify anything you claim.
Then upload your best-performing video scripts, your website, your brand guidelines, and any audience research you have. You can always add more later.
Step 2: Connect VidIQ
Now Claude knows you. Next it needs access to YouTube data.
Go to Customize → Connectors → Add Custom Connector. Paste in VidIQ’s MCP server URL and sign in with your VidIQ account.
https://mcp.vidiq.com/mcp
You’ll need a paid VidIQ plan for this, which in our opinion is well worth the money. You can check out their paid plans here if you don’t already have one. (affiliate link)
The moment that’s connected, Claude can pull YouTube data on demand.
Let’s put it to work.
Workflow 1: Researching Video Outliers
This is arguably the most powerful use case in this post.
When a video gets 10 or 20 times more views than its channel normally gets, that’s an outlier. It signals the topic, the format, or the packaging hit on something the audience wants more of.
Doing outlier research can take hours. Going channel by channel, search after search, wrestling with third-party tools. But with VidIQ connected, it becomes a single prompt.
Act as a YouTube research analyst. Find outlier videos in my niche and turn them into a shortlist of video ideas with demonstrated demand that I can consider adapting for my channel. FILL IN: Niche: [BE SPECIFIC] Timeframe: Last [X] months Minimum video age: [e.g. 7 DAYS] Outlier threshold: [10]x Number of results: [25] Channel size band: [e.g. 10K–500K SUBSCRIBERS] Video format: [e.g. LONG-FORM / SHORTS / LIVESTREAMS / ALL FORMATS] METHOD: 1. Search the niche from at least six different angles, including direct keywords, related topics, audience problems, competitor channels, and common video angles. 2. Use vidIQ for every search. Continue adding relevant search angles until you’ve collected approximately three times the requested number of candidates. For example, if I request 25 results, aim to collect around 75 candidates before applying the filters. 3. Filter candidates by the selected timeframe, minimum video age, Outlier Score threshold, channel size, and video format. 4. Use the Outlier Score provided by vidIQ. Exclude a video only if its Outlier Score can’t be verified or there isn’t enough data to confirm that it meets the selected filters. Keep videos with missing secondary data and mark those fields as unavailable. Rank the qualifying videos by Outlier Score. 5. Group the final videos into 4–7 clusters based on the viewer desire they satisfy — not simply their subject matter. 6. Inspect the actual titles and thumbnails before identifying packaging patterns. Do not infer thumbnail elements from titles or metadata. 7. Treat explanations for why a video overperformed as evidence-based hypotheses, not confirmed facts. DELIVER: 1. Search summary: Search angles used, total candidates collected, number remaining after filtering, exclusions broken down by reason, and anything you couldn’t access. 2. Results dashboard: Create a table that is sortable by Outlier Score, views, and publish date, and filterable by cluster. Include: Thumbnail | Clickable Title | Channel | Subscribers | Views | Outlier Score | Engagement Rate | Publish Date (YYYY-MM-DD) | Length | Cluster. Place the likely reason it overperformed in an expandable detail section for each video. 3. Cluster analysis: For each cluster, provide the number of qualifying videos, median Outlier Score, the viewer desire it satisfies, and whether the opportunity is a high, medium, or low fit for my channel based on audience alignment, subject-matter credibility, and production capacity. 4. Packaging patterns: Identify what repeats across titles, thumbnails, promises, and video length. Quantify each pattern using the actual number of qualifying videos — for example, “14 of 22 titles use a number.” Only claim that a pattern is absent from normal-performing videos if you have enough comparison data to verify it. 5. Top five ideas for my channel: Rank the five strongest opportunities. For each, include the source video, transferable insight, working title, opening hook, why it fits my channel specifically, and biggest risk. 6. What I should not make: Identify one pattern that works for other channels but would be a poor fit for mine, and explain why. RULES: - Every number must come from verified data or a calculation using verified data. - Use the Engagement Rate provided by vidIQ. Do not recalculate it from partial engagement data. If vidIQ does not provide it for a video, mark it as unavailable. - Never present an estimate as a measured figure. - If data is unavailable, say so rather than guessing. - If fewer videos clear the chosen Outlier Score threshold than requested, return fewer and state the actual number. Do not pad the list.
What came back was 25 outlier videos, sorted into topic clusters, in a clean interactive dashboard. In minutes.
Get the results as a spreadsheet Instead
If you’d rather work with the raw data, ask Claude to drop it into a spreadsheet instead. Same results, but sheets make it easier to sort, filter, and share with your team. You can even ask it to embed the thumbnails.
Replace Deliverable 2 with a downloadable .xlsx spreadsheet that can be opened in Excel or imported into Google Sheets. Do not also create the interactive dashboard. Embed the thumbnails as actual images where supported. If that isn’t possible, include the thumbnail URLs instead. Complete Deliverables 1 and 3–6 in the response as originally instructed.
Follow-up: steal formats from adjacent niches
To take this one step further, Claude can also hunt for outliers in adjacent niches — formats and angles crushing it somewhere else that nobody in your space has tried yet.
Now find [10] outlier videos from niches outside mine whose format or angle could transfer to my channel. The source audience does not need to overlap with mine. What matters is that the format, packaging, structure, or emotional mechanism can be credibly adapted to my subject and audience. Use vidIQ and follow the same research method, thresholds, filters, and verification rules as above. For each video, identify which niche it comes from and explain why its underlying content mechanism can transfer to mine. If a format only worked because of something I can’t replicate — such as a famous guest, large budget, or platform-specific moment — exclude it and continue searching for a replacement. If fewer than [10] videos qualify, return fewer and explain why. Do not pad the list.
Now, whether one of those outliers sparks your next winning video or you spot a pattern across all of them, you now have a shortlist of proven, data-backed ideas to build from.
This is outlier research made dead simple.
Workflow 2: Mining Comments
YouTube comments are a goldmine, and almost nobody is talking about mining them.
The comments on videos related to your topic tell you exactly what the audience cares about, what confused them, what they felt was missing, and what they’re begging someone to explain properly.
That’s real audience research, sitting there for the taking. Doing it manually is painstaking though.
So here’s what we do instead. Find a few popular videos already covering the topic you’re considering, give Claude the links, and prompt:
Act as a YouTube audience researcher. Analyze the comment sections of these videos and tell me what this audience actually wants, in their own words. FILL IN: Videos: [PASTE 3–6 URLS OR WRITE “FIND THEM”] Topic I’m considering: [TOPIC] My working angle: [ANGLE, OR “NONE YET”] METHOD: - If I haven’t provided URLs, use vidIQ to find 3–6 highly relevant, high-performing videos about the topic with substantial comment activity, ideally from different channels. List the videos selected before presenting the analysis. - Using vidIQ, pull a mix of the most-engaged and most recent comments from each video so the sample isn’t dominated by top-voted jokes or early reactions. Include replies where accessible. Aim for at least 100 comments per video where available. - Report how many comments you analyzed from each video. If comments are disabled, limited, mostly spam, unrelated, or otherwise inaccessible, say so. - Tag each substantive comment with one or more of the following: audience question / pain point / desired outcome / request for follow-up content / debate or objection / praise or criticism of the delivery / correction of the creator. Ignore spam, duplicates, irrelevant comments, and empty praise. - For each theme, count the number of distinct comments, the number of videos it appears across, and the total likes those comments received. Rank primarily by cross-video recurrence, then comment frequency. Use likes as a supporting signal or tiebreaker — not as a substitute for recurrence. - Separate themes that recur across multiple videos from those appearing in only one. Recurring beats loud. DELIVER: 1. Coverage: Comments analyzed per video, how they were sampled, and any limitations. 2. Top 10 audience questions: Rank them by cross-video recurrence and frequency, using likes as a supporting signal. For each, include the number of comments expressing it, the number of videos it appeared across, total likes, and one short verbatim example. 3. Pain points: Group them into themes, each with 2–3 short verbatim phrases in the audience’s own words. Quote exactly — don’t clean up their grammar or present a paraphrase as a quote. Their language is the most valuable part of this analysis. 4. Explicit content requests: Identify where people asked for a specific follow-up, tutorial, example, or deeper dive, with counts. 5. Debates: Show where commenters disagree with each other or the creator, summarize both sides, and explain why that tension could make a strong subtopic. 6. Delivery intel: Identify what commenters said about how the videos were made, including pacing, depth, length, repetition, and ads. This is the format bar my video needs to clear. 7. Must-hit list for my video: Rank the most important points by demand and map each one to the supporting evidence above. Be concrete: “Answer X with a worked example,” not “Cover X.” 8. Three title or hook lines: Create three options inspired by the audience’s language. For each, show the comment phrase or recurring theme it came from. RULES: - Use short verbatim excerpts only. Never present a paraphrase as a quote. - Separate what commenters said from your inference about what it means. - If a theme appears in fewer than approximately five distinct comments, label it a weak signal. - If a theme appears in only one video, label it video-specific. - Treat the findings as signals from the comments analyzed, not proof of what the entire audience believes. - If fewer than 10 meaningful audience questions are supported by the comments, return fewer. Do not pad the list. - Don’t tell me what I want to hear about my working angle. If the comments point toward something else, say so and explain what they support instead.
Because VidIQ is connected, you don’t have to scroll or copy-paste anything. Claude pulls the top comments straight from the videos, the ones people liked and engaged with most, and analyzes them together.
You get back a list of unanswered questions, subtopics the audience is requesting, and creative gaps.
Use it two ways: uncover new angles your audience is already asking for, or identify the must-hit points to build into a video you’re already planning.
This is how you develop content with the audience in mind. And when you keep the audience in mind, you drive better viewer satisfaction.
Workflow 3: Validating Video Ideas
Before you sink time and resources into production, you should validate your idea to know if it’s actually worth making.
Claude can check the two things that predict whether an idea is likely to work or not.
Demand. Claude pulls keyword data to see how many people are searching the topic and how competitive it is. No search interest, or already saturated? Red flag.
Competition. Claude looks at what already exists on the topic: how many videos, how good they are, and whether there are outliers pulling way more views than their channel’s average. This tells you if there’s room to win or if the space is too crowded.
Act as a YouTube strategist and pressure-test this video idea against real data. Your job is to tell me whether to make it — not to validate a decision I’ve already made. FILL IN: Working title or idea: [TITLE OR IDEA] What the video would actually teach or show: [ONE SENTENCE] What success looks like: [e.g. 20K VIEWS, OR 15 QUALIFIED LEADS] METHOD: - Demand: Based on the working title or idea, identify the most likely main topic or search term. Use vidIQ to test and refine it, along with 5–10 relevant variations and question-based searches. Report estimated monthly Search Volume, Competition, and Keyword Score where available, along with the 12-month trend direction if accessible. Determine whether the idea is primarily search-driven, browse-driven, or hybrid. If it’s browse-driven and no meaningful search term exists, say so rather than forcing one. - Competition: Find 10–15 of the most relevant, high-performing videos across the validated main topic, search term, and related variations. For each, collect the title, channel, subscribers, views, publish date, length, Outlier Score provided by vidIQ, and link. - Supply quality: Assess whether the strongest existing videos are current, thorough, and well packaged — or dated, shallow, incomplete, or poorly packaged. Use transcripts or vidIQ’s video-analysis tools to inspect the content of the strongest 3–5 competitors. Don’t judge content quality from titles and thumbnails alone. - Proof and saturation: Assess these separately. Proof means the topic has worked across multiple channels, channel sizes, or periods. Saturation risk means many recent videos are pursuing substantially the same topic, promise, and angle. DELIVER: - Verdict: MAKE IT / SHARPEN THE ANGLE FIRST / SKIP IT. Lead with this. Give your confidence as HIGH / MEDIUM / LOW, the three strongest reasons for the verdict, and the biggest remaining uncertainty. - The data and competitive assessment: Provide a keyword table and competitor table so I can verify the findings. Then assess: b. Search competition: Report the Competition data provided by vidIQ. b. Content competition: Rate it LOW / MODERATE / HIGH based on the number, strength, relevance, and recency of existing videos. Explain who currently dominates the topic, whether channels comparable to mine have succeeded, and where — if anywhere — there is a gap I could exploit. State whether the opportunity appears proven, saturated, both, or neither. - Realistic outcome: Give a directional view range for my channel based on its typical performance and the results achieved by comparable channels. Explain how you derived the range and don’t use the best-performing example as the baseline. If my goal is leads or sales, only estimate that outcome if historical conversion data is available. - Three sharpened angles: If the verdict is MAKE IT or SHARPEN THE ANGLE FIRST, rank three improved angles. For each, include the working title, opening hook, intended viewer, why it improves on the existing supply, and biggest risk. If the verdict is SKIP IT and no viable angle would change that, say so instead of forcing alternatives. - What would change your mind: Identify the one piece of additional evidence most likely to change your verdict. RULES: - Don’t be agreeable. I brought you this idea, which means I already like it. If the data says skip it, say so. A soft maybe is the least useful answer. - Use the Outlier Score provided by vidIQ. Don’t recalculate it using a separate formula. - Clearly label vidIQ Search Volume as an estimate, not an exact measured figure. - Don’t treat low search volume as proof of low demand if the idea is primarily browse-driven. - If keyword, competitor, trend, or content-quality data is unavailable, identify what is missing and lower your confidence accordingly. - Don’t estimate leads, sales, or conversions without historical conversion data. - Judge the idea against my actual channel size, typical performance, audience, and goals — not what a much larger channel could achieve.
This is the quick-and-dirty way to pressure-test any video idea you’re considering.
If you want to go deeper on this workflow, we have a full breakdown of our complete validation process here: “How to Find and Validate YouTube Video Ideas“.
Workflow 4: Analyzing Transcripts
You’ve found outlier ideas, audience gaps, and pressure-tested demand. Now make sure your video actually brings something better or different.
Instead of guessing how to cover a topic, you go in with a complete picture of what the top videos have already said — and what they haven’t.
Grab the links to the top-performing videos on your topic (Claude can find those for you too). Then run this:
Act as a YouTube content strategist. I’m researching the leading videos on this topic so I can create the most valuable version — not another video that repeats what already exists. FILL IN: Topic: [TOPIC] Videos: [PASTE 3–5 URLS OR WRITE “FIND THEM”] My planned angle: [ANGLE OR “NONE YET”] Target length: [e.g. 10–14 MINUTES] METHOD: - If I haven’t provided URLs, use vidIQ to find 3–5 relevant, high-performing videos on the topic, ideally from different channels. List the videos selected before presenting the analysis. - Using vidIQ, retrieve the full transcript for each video. Confirm which transcripts were retrieved and flag any that are unavailable, truncated, or too inaccurate to use. - Don’t analyze a video you couldn’t actually read. If I provided the URL, exclude the video and explain why. If you selected the video, find a replacement so the final analysis still includes 3–5 usable transcripts. - Use timestamps where available when referencing insights so the content-specific findings can be verified. If reliable timestamps aren’t available, say so rather than inventing them. - Base the analysis on what is said in the transcripts. Don’t make claims about visuals, editing, demonstrations, or on-screen elements unless vidIQ’s video-analysis tools were also used. DELIVER: 1. Coverage: List the transcripts retrieved, the quality of each, and any limitations. 2. Per-video analysis: a. Core thesis in one sentence b. The hook: a short verbatim excerpt from the first 30 seconds and the promise it makes c. Structure: the main beats with timestamps where available d. Two or three key points and the evidence or examples provided for each e. Intended viewer, strongest element, and biggest weakness 3. Consensus and disagreement: Identify the points that recur across multiple videos and report how many videos cover each one. Show where the videos disagree or take meaningfully different positions. Flag shared assumptions that could create an opportunity for a counter-angle, but don’t declare them wrong without evidence. 4. Gaps: Looking across the transcripts as a whole, identify points that are mentioned but underdeveloped, unsupported, oversimplified, or potentially outdated. Cite the relevant timestamp where available and label anything requiring further fact-checking. 5. White space: Identify angles, audiences, objections, examples, and counter-narratives not addressed in the transcripts analyzed. Be specific. “Go deeper” isn’t an angle; “Show what this looks like when it fails” is. 6. My differentiated angle: Recommend 2–3 directions and rank them. For each, include the hook, intended viewer, why it improves on these videos specifically, what I would need to prove or show, and biggest risk. 7. Recommended outline: Create a beat-by-beat outline for the top-ranked angle with rough timestamps based on my target length. Show what existing information should be covered quickly and where the video should go somewhere the others didn’t. RULES: - Support content-specific findings with timestamps where available. Use only short verbatim excerpts and don’t reproduce long sections of a transcript. - Separate what the videos explicitly say from your interpretation. - Don’t invent consensus. If the videos diverge, show where and how. - Don’t describe a claim as inaccurate or outdated unless it has been verified. Otherwise, label it as something requiring fact-checking. - If a transcript is unavailable, say so rather than guessing from the title, description, thumbnail, or comments. - Be honest about my planned angle. If the analyzed videos already cover it well, tell me instead of forcing a way to make it sound original.
You get a clear map of what every top video already covers, and just as importantly, what they’re all missing. The repetitive points, the gaps, and the opportunities to make your video feel genuinely different.
Why this matters: if your video says the same thing as every other video on the subject, what reason does a viewer have to stick around? None.
To keep a viewer, they need new information, more efficient delivery, or a different take. The gaps Claude finds here are exactly that. They’re the difference between 80% average view duration and 20%.
Workflow 5: Auditing Performance Data
Your YouTube analytics are telling you exactly what’s working and what isn’t on every video you’ve put out. Most people don’t know what to look for, or don’t have time to look.
Claude does. Run this:
Act as my YouTube content strategist. Audit this channel’s last 12 months of performance data, identify what’s working and what’s holding it back, and give me a ranked action plan to improve results. METHOD: - Use vidIQ for all available channel and video data. If a metric is unavailable, say so and continue the audit using the data you can verify. - Retention: Review the True Audience Retention Curve, 100-point drop-off graph, Average Percentage Viewed, and the relative retention comparison provided by vidIQ, which compares the video with other YouTube videos of a similar length. Identify meaningful drop-offs, spikes, and recurring patterns across the channel’s strongest and weakest videos. - Traffic sources: Identify how viewers are finding the channel and how the mix differs between the strongest and weakest videos. - Topic selection: Group videos by content pillar or recurring topic. Identify which consistently outperform, which underperform, and which earn their place in the strategy. - Video length: Group videos into useful length ranges and look for a relationship between length, retention, and views. Say plainly if there isn’t one. - Packaging: Inspect the actual titles and thumbnails of the strongest and weakest videos. Identify what the top performers share and any potential packaging opportunities. Without CTR data, treat conclusions about packaging performance as hypotheses. - Subscriber growth: Analyze net subscriber growth and subscribers gained or lost by video where available. - Publishing patterns: Check whether cadence, timing, format, or recurring series appear connected to stronger or weaker performance. Don’t claim causation from correlation alone. - Niche context: Using vidIQ, identify 3–5 relevant channels serving a similar audience, including channels of a comparable size where possible. Compare their public performance over the same 12-month period at a pattern level: topics, formats, views, Outlier Scores, video length, publishing cadence, titles, and thumbnails. Identify what they appear to do better, what my channel appears to do better, and any opportunities they’re capturing that my channel is missing. Don’t attempt to compare private metrics such as retention or traffic sources. DELIVER: 1. Headline finding: State the most important finding in two sentences. 2. What’s working: Identify the strongest videos, topics, formats, and packaging patterns with supporting evidence. State the likely influence of topic, retention, timing, traffic source, and packaging where the available data allows. 3. What’s holding the channel back: Identify the most important performance problems and distinguish between: a. A content or retention problem: viewers start watching but leave at consistent points b. A topic problem: the subject consistently underperforms despite adequate retention c. A distribution problem: performance appears limited by traffic sources or reach d. A potential packaging problem: titles or thumbnails appear weaker than those of the channel’s strongest performers, but CTR data isn’t available to confirm it 4. Niche context: Summarize the most important differences between this channel and the relevant channels analyzed. Identify two or three things they appear to do better, one or two advantages this channel has, and any clear missed opportunity. Base this only on verifiable public data. 5. Performance summary table: For each important metric, show the strongest and weakest results, the overall channel pattern, what it suggests, and the data source. 6. Ranked action plan: Rank the recommended changes from highest to lowest expected impact. For each, explain exactly what to change, why the data supports it, the effort required given the constraints in this Project, and how to determine within 30 days or the next three uploads whether it worked. 7. The one thing: If I make only one change, tell me what it should be and why. RULES: - Every number must come from verified data or a calculation using verified data. Identify its source. - Never present an estimate or inference as measured data. Clearly separate verified metrics from your interpretation of what they suggest. - Analyze Shorts and long-form separately. Combining them will distort the data. - If a finding is based on only a small number of videos, state the sample size and label the conclusion as tentative. - Treat relationships in the data as evidence-based hypotheses, not proof of causation. - Only evaluate leads, sales, or other off-platform outcomes if the necessary conversion data is available. Otherwise, keep the audit focused on measurable YouTube performance. - Don’t give vague recommendations. Give specific, actionable advice.
What comes back is a strategic audit. This is the kind of analysis that used to mean hours in YouTube Studio. Now it’s one prompt and about five minutes.
When we ran it on our own channel, it pinpointed our single best-performing topic and told us to prioritize it. It pulled our most recent retention data and flagged specific fixes for writing better hooks. And it handed us a ranked list of what to fix first across the whole channel. Not a vague summary, an actual plan.
Recap & Next Steps
With all of these workflows, you still need to bring your own strategic thinking. Claude will surface patterns, insights, and recommendations far faster than you could manually — but you need to sanity-check the output, apply context, and decide what actually makes sense for your channel.
That’s the point of all five workflows. They’re accelerators, not a substitute for you. The real advantage comes from the collaboration: your taste, judgment, and strategy, paired with Claude’s ability to pull data, spot patterns, and compress hours into minutes.
And the more context you give it, the better that collaboration gets.
Want help turning this research into a YouTube strategy?
Turning all of this data into a focused YouTube strategy, compelling content, and consistent execution is the next step.
Vireo is a YouTube-specialized agency that has helped 300+ leading brands generate billions of organic views through strategy, production, and optimization.
If you want our team to take a closer look at your channel and help create a clear path forward, let’s talk.
Frequently Asked Questions
Yes, you need a plan that supports Projects and custom connectors. The free tier won’t get you there.
Yes. The MCP connection requires a paid VidIQ account. It’s the piece that gives Claude live YouTube data instead of guesswork.
MCP (Model Context Protocol) is the standard that lets Claude connect to outside tools and pull live data. VidIQ’s MCP server is what turns Claude from a writing assistant into a YouTube research tool.
A video that significantly outperforms its own channel’s average — typically 10x or more. The multiple matters more than the raw view count, because it isolates what the audience responded to rather than how big the channel is.
Workflows 1 through 4, yes — they run on public data. Workflow 5 requires access to your own analytics.
Three things, mainly.
1) Depth of context. AI Coach knows your channel data. A Claude Project knows your channel data plus your brand guidelines, your best-performing scripts, your positioning, your tone of voice, and anything else you upload. The gap between “advice for a channel like yours” and “advice for your channel” is almost entirely a context problem.
2) Chained analysis. Workflows 2 and 4 in this post require pulling data from multiple videos and synthesizing across all of them. That’s a fundamentally different operation from answering one question well.
3) Real deliverables. Claude produces spreadsheets, interactive dashboards, and documents you can hand to your team. AI Coach returns answers in a chat pane. For solo work that’s fine. For a team it isn’t.
