Turn Your YouTube Analytics Into Your Next Video (With AI)

There are two patterns hiding in your YouTube analytics: videos people don't click on, and videos people click on but don't stay to watch. Fix one or both, and you can fix your whole channel.

Most of us open the analytics tab in YouTube Studio, have a quick look and aren't sure what we're seeing. AI has changed that. I walk you through the quarterly process I use to turn my YouTube analytics into my next set of videos, using AI as a YouTube strategist. I use Claude, but any AI model will do.

Key Takeaways

• Ask your AI which YouTube analytics fields it needs, export them from YouTube Studio as a CSV, and let it analyse what's working and what isn't.
• The analysis showed a handful of videos carrying my whole channel, so the plan is to do more of what works and less of what doesn't.
• AI can then suggest a quarter's worth of video titles, but treat them as a starting point and keep a human in the loop.

Set Up a Claude Project for Your YouTube Content

You can do this in a normal chat, but I use a project in Claude called my YouTube content pipeline. A project has its own description, instructions and files, so every chat inside it has the same context.

I've uploaded my brand bible, my YouTube script template and my content strategy. I explain how I set up projects in my post on how I use Claude to run my business.

Step 1: Ask AI What YouTube Analytics It Needs

I do this at the start of every quarter. In a fresh chat, I tell Claude what I'm trying to do, roughly like this:

"I'm planning the next 13 videos for Q3. Help me analyse my YouTube Studio analytics. Tell me exactly which fields to export into a CSV, so you can see which videos aren't getting clicks, which aren't keeping viewers, and what I should do more and less of. Ask me any questions before we start."

I speak this using Wispr Flow rather than typing it. Within about 20 seconds, Claude explained how to export from YouTube Studio, offered the last 365 days or lifetime data, and listed the exact fields it wanted. If you don't have many videos yet, just export what you have.

Step 2: Export Your YouTube Studio Analytics as a CSV

1. In YouTube Studio, click Analytics, then Advanced mode.
2. Make sure your channel is selected and set the date range. I used the last 365 days, as I have over 300 videos.
3. Choose Content, then tick the metrics Claude asked for. Use the search box in the metrics list to find each one, such as watch time.
4. Click Apply, then Export current view in the top right.
5. Choose Comma-separated values (.csv) rather than Google Sheets, so you can upload it straight to your AI.

The export downloads as a zip file containing totals, chart data and table data. If you're not sure which file to share, ask. Claude asked for the table data.

Step 3: Let AI Analyse What's Working

Upload the CSV and ask Claude to analyse it. It took a couple of minutes, so it's a good time to grab a coffee.

The findings were eye-opening. My channel isn't really carried by 300-plus videos, it's carried by about 10. One video, on Google Gemini Gems, brought in 22% of my views and 41 net subscribers over the year. I wrote it up as a blog post on Gemini Gems.

What the analysis told me:

• Click-through rate: not my bottleneck. Subscribers gained per 1,000 views was the more useful measure.
• Do more of: AI for solo businesses, Kajabi tutorials and YouTube channel growth. Specific AI content and my Kajabi videos were the two clear winners.
• Do less of: generic tool tutorials, like my Loom tutorial, that drift away from my mission.

For more on spotting patterns in your own videos, see my post on the 3 patterns that drive views, and my guide to YouTube traffic sources.

Step 4: Turn the Analysis Into Your Next Videos

Claude offered to pull in lifetime data, which you can export the same way by switching the date range to lifetime. For this exercise, I stuck with the last year and asked for 13 working titles for the next three months.

That took about 45 seconds, and some of the titles were genuinely good. Because Claude knows my story from previous chats, it suggested angles like leaving a 16-year RAF career to build a one-person business, whether a one-person business can replace a full-time salary, and a one-person business versus an agency comparison.

Keep a Human in the Loop

Treat the output as a starting point. AI can do 80% of the grunt work, but the other 20% needs you. Some titles might inspire a video you'd never have thought of.

What I do next with a title:

1. Search it on YouTube to see what already exists and how I can put my own spin on it.
2. Ask the AI for variations of the title, then pick my favourite.
3. Give it the final title and target keyword, and ask for a hook and intro, or a full script if I'm using a teleprompter.

I cover the rest of the process in my YouTube AI workflow.

Conclusion

Your YouTube analytics already tell you what to make next. Ask AI which fields it needs, export them as a CSV, and let it show you which videos are carrying your channel and which aren't.

Then use it to plan a quarter of working titles, and add your own story and judgement to make each video truly yours. If you're new to YouTube, run the same process on whatever videos you have.