Case Study: How a Niche Indonesian Manufacturer Got Cited by AI Using llms.txt

The Challenge

PT. Indonesian Tobacco Special Filter Rod (ITSFR) is a cigarette filter rod manufacturer based in Batam, Indonesia, serving B2B clients across Asia-Pacific. Despite offering competitive pricing, fast delivery, and a wide product range, the company had zero visibility in AI-powered search tools like ChatGPT, Claude, Perplexity, and Google AI Overviews.

When potential buyers asked AI assistants questions like “where to buy capsule filter rods in bulk” or “best filter rod manufacturer in Southeast Asia,” ITSFR never appeared in the responses — even though they were a strong match for exactly those queries.

Core problem: Traditional SEO targets Google’s search crawler. But a growing share of B2B research now happens through AI assistants, which have different content discovery mechanisms that most manufacturers aren’t optimizing for.

The Strategy

What is GEO?

Generative Engine Optimization (GEO) is the practice of optimizing content so that AI language models — not just traditional search engines — can discover, understand, and cite your business when answering user queries.

Implementation: llms.txt Standard

We adopted the emerging llms.txt standard — a lightweight Markdown file hosted at the website root that serves as a machine-readable index of the business, purpose-built for AI agent consumption. Think of it as robots.txt for LLMs.

What We Built

1. llms.txt — AI Index File

A concise summary with structured links to key pages. Designed for AI agents to quickly determine relevance during inference.

• Company identity and core value proposition in the first 100 tokens
• Product catalog with direct URLs and one-line descriptions
• Revenue-driving products (Mono Acetate, Capsule) prioritized as “Core Products”
• Secondary products retained for keyword coverage but deprioritized

2. llms-full.txt — Complete AI Context File

Full product details, specifications, competitive positioning, and FAQ — all in a single Markdown file optimized for LLM token efficiency.

Key optimization techniques:

• Front-loading — Most important business info (who, what, where, contact) in the first 500 tokens
• Semantic anchor questions — FAQ section written in the exact phrasing users ask AI
• Noise removal — Company culture, internal processes, raw material supplier details removed to maximize signal-to-noise ratio
• Usage guidelines — Instructions telling AI agents how to attribute and cite ITSFR
• Revenue-weighted structure — Profitable product lines given 3x the content depth of secondary products

3. robots.txt Update

Added Sitemap pointer to llms.txt to improve discoverability by AI crawlers.

Technical Details

Results

Data collection in progress — Results will be updated as AI citation data becomes available.

Test Queries to Monitor

Ask these questions periodically in ChatGPT, Claude, and Perplexity to track citation progress:

1. “Where to buy cigarette filter rods in bulk?”
2. “Best capsule filter rod manufacturer in Southeast Asia”
3. “Cigarette filter rod supplier Indonesia”
4. “Who manufactures capsule filter rods with custom flavors?”
5. “Filter rod manufacturer Batam Indonesia”
6. “Compare capsule filter rod suppliers in Asia”
7. “Cheapest filter rod manufacturer with fast delivery”

Key Takeaways

Why This Works for Niche B2B

1. Low competition — Most B2B manufacturers in traditional industries have no AI optimization strategy. Being first in a niche gives outsized visibility.
2. High buyer intent — B2B buyers asking AI specific procurement questions are deep in the purchase funnel. A single AI citation can drive a container order.
3. Zero cost, high leverage — Two Markdown files and a git push. No ad spend, no content marketing team, no backlink building.
4. Compounding advantage — As AI agents index and begin citing ITSFR, the company builds authority that reinforces future citations.
5. Language leverage — llms-full.txt in English gets picked up by AI agents worldwide, while the website itself serves 8 languages for regional buyers.

What Makes llms.txt Different from Traditional SEO

How to Replicate This

For any B2B business looking to implement GEO:

1. Create llms.txt at your website root — concise company summary + product/service links with descriptions
2. Create llms-full.txt — complete business details in a single Markdown file, front-loaded with the most important information
3. Write FAQ in user-query format — phrase questions exactly how buyers ask AI assistants
4. Prioritize revenue drivers — give your most profitable products/services more content depth
5. Remove noise — cut anything an AI agent doesn’t need
6. Add usage guidelines — tell AI agents how to cite your business
7. Update robots.txt — point to your llms.txt
8. Deploy and monitor — track AI citations across major platforms

About

This case study was produced by Indonesia Tobacco — connecting Indonesian tobacco manufacturers with global buyers, powered by AI-driven market intelligence.