The question
Whether the low maintenance cost of llms.txt is worthwhile for agent-readiness even though no AI-search citation benefit is proven.
You want AI systems to surface your best content and are weighing whether an llms.txt file is worth publishing.
Why it matters
Skipping llms.txt where it helps can leave AI systems guessing at your most important pages, while over-investing in an unadopted standard can divert effort from fundamentals that matter more today.
How to decide
IfIf you have a clear set of high-value pages you want AI systems to prioritise
ThenYes - publish a concise llms.txt pointing to them; the cost is trivial.
IfIf your documentation or knowledge base is large and hard to navigate
ThenYes - an llms.txt curates the canonical entry points for models.
IfIf your citation and content fundamentals are still weak
ThenNo - fix citable structure and authority first; llms.txt does not rescue thin content.
IfIf you can add it cheaply alongside other GEO work
ThenYes - treat it as low-cost insurance while adoption matures.