Built and run in production for a travel content publisher. One topic line goes into a Google Sheet; a finished article with keyword research, metadata, internal links and schema markup comes out, logged to Airtable for review and pushed to the site and its social channels.
The system runs on Make.com with the Claude API handling research, drafting, fact-checking and caption writing. Airtable tracks every article’s status and QA flags; Google Drive holds the output by run; a GitHub trigger rebuilds the site when an article is approved.
What changed over the project: I consolidated two AI vendors into one, cutting API calls per article from 5 to 2 and token cost by 75 percent. I replaced fragile JSON parsing with strict tool-use outputs, which took parsing failures to zero. I added automated checks for length, links, duplicates and house style that removed a full editing pass per article, and tuned the prompts against the site’s existing writing to strip the patterns that make AI content read as AI.
Every article on the site is now produced this way. The client reviews finished articles and adds topics; nothing else needs a hand.
0 likes · 1 view · Nov 11, 2026
Make.com and Claude API pipeline that turns one topic into a finished, SEO-ready article with research, metadata, links and quality checks. Cut API calls 5 to 2
