Key Takeaways

  • TikTok blocks direct programmatic scraping even through its official API, stopping AI agents from pulling audience feedback directly from the backend.
  • Lacey bypasses this block by exporting videos from her phone into Google Drive, giving Claude Code direct access to the source media files.
  • By leaving an authenticated TikTok browser window active, Lacey directs Claude to inspect live comment sections alongside the archived Drive videos.
  • Claude evaluates user questions and confusion against the video content to draft targeted follow-up scripts, such as answering whether hospitals will adopt AI for billing and coding.
  • This workflow fuels Lacey's "WTF even is" explainer series on complex blockchain jargon while keeping her production cadence to one video per week.

The Method

Short-form platforms treat their comment sections like fortresses. As Lacey pointed out, “The thing with Tik Tok first of all on a technical level is that their algorithm their back end software their back end whatever is so um precise that you can't scrape it. You can't have an AI even with their API.”

Standard scrapers fail immediately. Lacey solved this by splitting storage and analysis into two simple pipes.

First, she exports her raw TikTok video files directly from her phone into Google Drive. This gives Claude Code a stable local repository to inspect the exact media she published.

Second, she keeps an authenticated browser window logged into TikTok on her desktop. Instead of trying to run headless API calls that trigger bot defenses, she points Claude directly at the active browser session. As Vo summarized during their conversation: “You said 'Tik Tok's open. Go look at Tik Tok and read the comments for this particular video. Pair it with the download of the video that's up in Google Drive. And then pull it together and say based on how this did, the content in this video and the comments give me new stuff.'”

Claude compares what Lacey said in the video against the specific questions viewers dropped underneath it. In one case, a viewer responded to an explainer by asking: “I wonder if hospitals are going to start using AI for the billing and coding.”

Because TikTok allows creators to post video replies directly to specific comments, Lacey uses Claude to write short, punchy scripts that answer that precise query. She applied the same technique to her "WTF even is" series, where she unpacks dense blockchain terminology. The comments show where the audience got lost, and Claude turns those friction points into the next script.

Where This Breaks Down

This setup works because Lacey posts once a week. If you publish three videos a day, manually exporting files from a phone into Google Drive becomes a bottleneck within 48 hours.

Authenticated browser scraping is also brittle. TikTok frequently throws interactive verification puzzles or cycles session cookies when it detects automated tab inspection. If Claude loses the active session, your automated analysis halts until you complete a manual captcha.

Finally, comment sections contain noise. Unmoderated short-form comments skew heavily toward contrarian arguments and cheap jokes. If you instruct an AI model to build scripts from raw engagement alone, it will optimize for outrage rather than useful technical clarity. You still have to filter the feed for genuine questions like the hospital billing inquiry before telling Claude to draft the response.

What to Do With This

Open your highest-performing video from the past month on your phone and save the original file to a designated Google Drive folder. Log into TikTok or YouTube Shorts on your desktop browser, open that video's comment section, and launch Claude Code in your terminal. Prompt Claude to review the comments on screen, flag three distinct points where viewers expressed confusion, and write a sixty-second script that answers the strongest question.