Solving workflow issues in automated travel content creation

Hi everyone, I’m working on a travel content automation project using GPT-4.5 Deep Research, and I’m facing a workflow issue I’d love help understanding or improving.

What I’m trying to do:

  • I uploaded a CSV file with 25+ tourism-related categories for the destination Dubai.
  • My prompt asks GPT to:
    • Shortlist relevant categories
    • For each selected category, generate 5 high-quality POIs
    • Include real data (Google + TripAdvisor ratings, Viator links, authentic mentions)
    • Output the result as a CSV file with strict formatting
    • All POIs and links must be from authentic, working sources (not dummy data)

The problem:
Even though I upload the full category list and ask for full output, GPT:

  • Only returns 1–2 categories at a time (sometimes just samples)
  • Doesn’t auto-process the entire file or return 5 POIs per category as expected
  • I have to keep asking manually for more categories to be processed
  • It’s unclear whether the model “remembers” the full input file or is chunking it due to size/complexity

My question:

How do I get GPT to fully process all categories from the uploaded CSV — generating 5 authentic, verifiable POIs per category — without having to manually request each one?