An open-source dev tool gets cited in 6 of 10 LLM answers in 60 days
GitHub repo with weak landing page. Fixed docs site, llms.txt, BLUF answers across docs. Strong GitHub-stars → strong brand-entity.
Open-source dev tool, ~4K GitHub stars, used by a few thousand devs, but invisible to ChatGPT + Claude for the canonical 'how do I X with [tool]' queries. Maintained by a 2-person team plus contributors.
Narrative
Week 1–2 was the diagnostic: scored 51. The shocking finding was that the existing docs site at /docs was a clean SSG built with VitePress but had zero LLM-discoverability features. Week 3 deployed /llms.txt + restructured top 30 pages with BLUF. Week 4–5 was the FAQPage + structural-depth pass. Week 6 was Wikidata submission (approved in 6 days). Week 7–8 was the case-study showcases. The key insight: an open-source tool with 4K stars has more brand-entity authority than most SaaS at 100K MRR — the OSS metric (stars + contributors + downloads) is actually a stronger Wikidata-citation signal than press mentions. By day 60, the cite rate on 'how do I X with [tool]' queries was 6 of 10 (vs <1 of 10 baseline) across ChatGPT, Claude, Perplexity, and AI Overviews.
Signals moved
Published /llms.txt + /llms-full.txt per the llmstxt.org spec. Linked to docs sitemap. Validated via VectorCite tool.
Rewrote the top 30 docs pages to lead with a 2-sentence direct answer before the conceptual context. Moved 'why we built this' below the fold.
Restructured docs from 'Getting started → API reference' to H2-segmented sub-question landing pages: 'How to install', 'How to configure auth', 'How to handle errors', etc.
Added FAQPage schema to every docs page with the top 5 most-asked questions on Discord / GitHub Issues.
4K GitHub stars + Wikipedia + Crunchbase already qualified the brand for Wikidata. Submitted Wikidata entry referencing the GitHub repo as primary external ID. Approved in 6 days.
Added citations to relevant RFCs, official docs (Node.js, Python, MDN), and academic papers in the conceptual sections.
Published 6 'How [company] uses [tool]' showcase pages, each with concrete code snippets + scale numbers from real users.
Takeaways
- 1.Open-source tools have an unfair advantage on brand-entity signals — GitHub stars + npm/PyPI downloads + contributor lists are all Wikidata-citation-grade external IDs. Use them.
- 2.Documentation-first sites need to invert their structure for AEO: top pages should answer 'how do I X' not 'what is X'. The conceptual content goes below the fold.
- 3.llms.txt + llms-full.txt deserves to be treated as a publishing standard for open-source projects — it's how you tell AI engines what your canonical docs are vs the noisy GitHub README.
Run the audit on your site.
The full 47-signal audit tells you which of these moves applies to your URLs.
Run free audit