Video summary

ChatGPT Work's Most Overlooked (And Useful) Feature

Main summary

Key takeaways

Product Review

Product/Feature Reviewed

ChatGPT Work “Sites” — a feature that lets you create and deploy real websites and web applications directly from ChatGPT.

Key Features Mentioned

  • Create & deploy sites from natural language (via the “@sites” tool in ChatGPT Work)
  • Live, active URL for the generated site
  • Persistent data support (site state/history and stored results remain after conversations end)
  • Authentication and file storage as supported capabilities
  • Analytics tab to track the number of visitors
  • Sharing controls
    • Private option (requires logging into ChatGPT)
    • Option to make it public
  • Customization options
    • Customizable URL beginning (rename slug)
    • Ability to use an own domain
    • Ability to add environment variables
    • Visual/custom appearance can be customized after functionality is ready (per the creator)
  • Availability scope
    • Available in ChatGPT Work or Codex, but the video focuses on Work
  • Integration with tools/skills
    • Can wrap results from a ChatGPT skill into a site
    • Can add history functionality and embed sources/original posts

Pros (User Experience & Benefits)

  • Fast setup: one example claims a site was created in 9 minutes from a simple prompt
  • Easy iteration: asking for changes updates the site; users reload and see updates
  • Turns scattered chat context into a navigable interface
    • Example: converting multi-chat research into a site for a Disneyland trip
  • Useful for decision tools
    • Example: an RV buyer’s guide that adapts recommendations based on parameters like budget, family size, towing, comfort, and storage
  • Repackages complex workflows
    • Example: transforming an agent-based skill (“Last 30 Days”) into a polished research site including:
      • summaries
      • use cases
      • unconventional views
      • embedded original posts
      • embedded sources
  • Persistent storage
    • Example: building a growing research library where results don’t disappear after the conversation ends
  • Automation + scheduled tasks
    • Example: a newsletter aggregation site that:
      • connects multiple Gmail accounts
      • processes a large archive (2,000+ mailings)
      • supports daily scheduled updates via “scheduled tasks”
    • Example outcome: entry count increased from 2081 to 2094 after refresh

Cons / Limitations Mentioned

  • Not all automation is seamless with skills
    • The site can’t directly “launch skills” from the site interface
    • Workaround described: run the skill separately, copy JSON, paste into the site (the creator disliked this)
    • Later troubleshooting improved how automatically updates worked
  • Hardcoded URL issue during updates
    • Changing the site URL caused a previous one to be hardcoded in skill logic
    • The skill had to be updated to match the new site URL
  • Official storage limit unclear
    • Creator couldn’t find an official OpenAI storage limit; assumes not unlimited, especially for large files
    • Text-heavy datasets (thousands of entries) worked fine
  • Access authorization sometimes required
    • Example: content was blocked until explicit authorization during skill/site testing

Comparisons / Positioning vs Other Tools

  • The video positions Sites as the “best solution” for turning outputs into interactive tools, but suggests:
    • For more complex games or deeper capabilities, the creator recommends Codex or Cloud Code
  • Within ChatGPT Work, Sites are framed as a final stage of a larger workflow:
    • research → analyze/connect data → build workflow → publish as an interface
  • Compared to keeping results in chat, a site is presented as:
    • more organized and reusable than static summaries or long chat threads

Unique Points Mentioned (All Distinct Claims)

  1. Sites create and deploy real websites/apps from ChatGPT Work.
  2. Supported capabilities include persistent data, file storage, authentication, analytics, and a URL.
  3. Sites tool usage: switch to Work, then use @sites.
  4. Example “Capybara Protocol” created in ~9 minutes and works via a live URL.
  5. Sites are hosted on ChatGPT hosting and are preconfigured/built-in.
  6. Share settings: public or private (private requires ChatGPT login).
  7. Site settings include customizable URL slug and optional custom domain.
  8. Environment variables supported.
  9. Analytics tab tracks visitor counts.
  10. Example use: convert multi-chat research into a Disneyland trip site to share with others.
  11. Sites help replace long text/chat threads with a structured interface.
  12. Example decision tool: RV buyer’s guide with adjustable parameters and dynamic recommendations.
  13. Example workflow: turn an agent-based GitHub skill (“Last 30 days”) into a persistent research website.
  14. That skill gathers data from YouTube transcripts, Reddit, Polymarket; also mentions scraping from Instagram/TikTok via “scrape creator.”
  15. Site incorporates summaries plus use cases, unconventional views, and links back to original posts.
  16. Ability to embed/show sources and original posts inside the site.
  17. History functionality added after further request.
  18. Limitation: sites may not directly launch skills; requires running the skill separately and inserting JSON (initially).
  19. Troubleshooting: updating the skill helped remove hardcoded URL problems.
  20. Persistent research library grows over time without losing past results.
  21. Example automation: newsletter-to-searchable-database with duplicate removal and workflow extraction.
  22. New Gmail account connection: support for connecting multiple Gmail accounts to ChatGPT.
  23. Large volume ingest: processed an archive of 2,000+ newsletters/mailings.
  24. Creator couldn’t confirm official storage limits; assumes limits exist but text storage worked for thousands of entries.
  25. Automation via scheduled tasks updates the site daily (e.g., noon).
  26. Scheduled processing increased entries from 2081 to 2094.
  27. Sites described as turning research/workflows into “real tools” via persistent storage + scheduled automation.

Speakers / Perspectives

  • Single main speaker (creator/demo author) throughout, providing:
    • feature walkthrough
    • multiple demos (trip planner, RV guide, skill-based research site, newsletter automation)
    • troubleshooting notes (URL hardcoding, skill update flow)

Overall Verdict / Recommendation

Strong recommendation for creators and knowledge/workflow builders. The video portrays ChatGPT Work “Sites” as one of Work’s most valuable overlooked features because it makes AI outputs interactive, shareable, persistent, and automatable—effectively turning research and skills into reusable tools.

Main caveats are some automation/skill-launch friction and occasional setup/troubleshooting needs (e.g., URL hardcoding), but the demos suggest it becomes very powerful once configured.

Original video