LLM-Powered Blog Production: 30 thMenu Affiliate-Focused Topics

Berke, a food-tech creator from Eskişehir, opened his laptop Saturday morning at 09:00 and by Sunday night had produced 30 fully outlined blog posts. His secret was a sequenced prompt chain between ChatGPT-5 and Perplexity. This guide walks through the exact workflow so any affiliate marketer can replicate it.

Master Prompt: Define the Role

The first step is telling the model who it is. Berke used: "You are a 10-year veteran Turkish food-tech writer. You speak directly to restaurant owners, warm but professional. You are an affiliate of thMenu, a QR menu platform, and will mention it organically 1-2 times per post — never spam, always positioned as solution." This role anchor keeps tone consistent across all 30 posts.

Layer in context: audience (SMB restaurant owners, ages 35-55), Turkish market dynamics, and forbidden zones (no naming competitors, no pricing promises, no fabricated statistics). The richer the context, the more every sentence in the next 30 posts adheres to it.

Generate 30 Topics in One Shot

Berke's second prompt: "Produce 30 long-tail blog topics. Each: (1) Turkish title 60-80 chars, (2) target keyword, (3) two-sentence angle, (4) how thMenu feature enters organically. Distribution: 40% industry trends, 30% operational tips, 20% case studies, 10% tool comparisons." The percentage split is critical — without it, the model collapses all 30 into "what is a QR menu" variants.

Then pipe the output into Perplexity for validation: "Compare these 30 titles against Google Trends Turkey 12-month data. Flag low-volume ones and suggest 2026-trending alternatives." Because Perplexity has live browsing, this eliminates ChatGPT's hallucination risk. Berke swapped 7 of 30 titles at this step.

Convert Each Topic to Outline

The third step is bulk outline generation. Prompt: "For each of the 30 titles, generate 4 H2 sections and a 3-question FAQ. Each H2 has 2-3 sub-bullets. Output JSON with slug and readTime included." JSON output feeds a Next.js blog seed script directly. Berke ingested the output into his repo with one line of code.

Final QA step: feed each outline back to ChatGPT with "What question would a restaurant owner reading this outline NOT find answered?" The model identifies its own gaps and patches them. Result: 30 titles × 4 H2s × ~3 sub-bullets = 360 distinct content blocks, all woven into the affiliate narrative.

FAQ

Can 30 posts actually be finished in one weekend? Outlines yes, polished drafts still need 30-45 minutes each. Berke spread publication across 6 weeks.

Why not just use Perplexity alone? Perplexity is weak at creativity. ChatGPT creates, Perplexity validates. The combo outperforms either solo by ~40%.

How many affiliate mentions per post are safe? 1-2 contextual mentions stay within Google E-E-A-T. Three or more reads as spam. Berke averages 1.3 mentions per post.

Found this helpful? Share it.