How Volkswagen Built the Best llms.txt in the Automotive Industry
Published 2026-04-20 · PROGEOLAB Research
The Volkswagen llms.txt at vw.com is 215 lines long, lists 198 URLs, and is organized into five Markdown sections. It is the largest accessible llms.txt file in the Fortune Global 500 and the clearest example of what a deliberate AI content strategy looks like at enterprise scale. Every element — the structure, the link selection, the WAF configuration that keeps the file reachable — reflects choices made across multiple teams.
VW's AI Readiness Score (9 / 12) is the highest of any Fortune 500 automaker and higher than most technology companies. Salesforce, Oracle, and IBM — all larger than VW in market cap — score 2, 0, and 2 respectively. The top-performing llms.txt in the world's largest enterprises was built by a company founded in 1937, not by a tech giant.
The anatomy of 198 links
VW's llms.txt opens with the header # Volkswagen U.S. and declares five sections as H2 headings:
| Section | Links | Purpose |
|---|---|---|
| Models | 10 | Product pages for each nameplate (Atlas, ID.4, Jetta, Tiguan, etc.) |
| Shopping Tools | 9 | Configurator, dealer locator, finance calculator, trade-in value |
| Owners | 130 | Manuals, service schedules, Car-Net pairing, Tech Help |
| Financial Services | 12 | Leasing, financing, lease-end |
| Newsroom | 25 | Press releases and corporate news |
The 65% owner-first strategy
The most revealing design choice is the Owners section consuming 65% of the total links. VW didn't optimize for prospect searches ("compare Tiguan with CR-V") — those queries route through Shopping Tools (9 links). Instead, VW optimized for the existing-customer intent ("how do I pair my phone with my Tiguan", "when is the Atlas maintenance schedule", "how to use Car-Net").
This is a sophisticated content-strategy bet. Existing customers asking AI for help have already committed. The question is whether the AI's answer directs them back to VW's official support content (reducing support-call volume, maintaining brand trust) or to third-party forums whose accuracy is unreliable. By mapping 130 owner-resource URLs into the llms.txt, VW ensures AI systems have a curated path to authoritative VW content when owners ask.
VW vs Salesforce: accessible vs blocked
Salesforce's llms.txt has 205 links — seven more than VW. Yet Salesforce's AI Readiness Score is 2 while VW's is 9. The difference is simple: VW's llms.txt is readable by the AI crawlers it was designed for. Salesforce's is not. Salesforce's WAF blocks ChatGPT-User entirely — the 205-link content directory might as well not exist from the perspective of an AI asked about Salesforce products.
This is the Content-Access Contradiction in concrete form: editorial investment without infrastructure investment doesn't produce visibility. Salesforce probably spent more engineer-hours assembling those 205 links than VW spent on 198. The investment is wasted because the security team's WAF rules block the audience.
VW vs Dell vs Subaru: three strategies
The top three Fortune 500 llms.txt implementations reflect different content strategies:
- VW (198 links, 65% owners) — the existing-customer playbook. Optimize for support and retention.
- Dell (131 links, multi-audience) — enterprise buyer + consumer + developer. Each audience gets a section.
- Subaru (100 links, lifestyle) — heavy on model pages, dealer resources, and brand storytelling. Optimize for consideration.
All three work. There is no single "correct" llms.txt structure; the file's value comes from matching the content architecture to the AI use case the brand wants to inform. What VW demonstrates is that the work of curation produces measurable accessibility improvement — and that the work is cheap relative to the infrastructure spend most enterprises already make on WAF, CDN, and CMS.
The replicable pattern
What makes VW's llms.txt replicable:
- Pick the audience. Owners, buyers, developers, analysts — which group's AI queries matter most?
- List the URLs. Not every URL. The ones that answer recurring questions from that audience.
- Organize into 3-5 sections with H2 headings. Fewer sections = clearer extraction.
- Verify the WAF permits AI access to the llms.txt file itself. Many companies' WAFs block
/llms.txtas an unusual path even when the homepage is open. - Put the file at the domain root, not a subdirectory.
yourdomain.com/llms.txt, nothing else.
The entire VW file took a content-team-sized effort, not an engineering-team-sized effort. The 53-point checklist item-level walkthrough sits around the llms.txt creation; the 5-hour transformation places VW's level of investment at Hour 2 of the roadmap. Within one quarter, any enterprise with a content team could ship VW-grade AI visibility. Most haven't.