Add llms.txt to the root [https://next-intl.dev/docs/llms.txt]

[ii-vo]

Is your feature request related to a problem? Please describe.

This would help AI to understand the documents easier

Describe the solution you'd like

Add the llms.txt

Describe alternatives you've considered

Building my own scraper to get the documentation to a LLM easy to read format.

[amannn]

I‘m currently sharing the stance here from the Tailwind team: tailwindlabs/tailwindcss.com#2388

Apart from that, i18n setups vary quite a lot, I think you’ll have a better time if you set up a skill or similar based on your preferences.

[stevenceuppens]

Hi @amannn,

I have a slightly different perspective on this.

I’ve seen the discussions regarding Tailwind and understand the tension there regarding their business model and site traffic. However, I believe next-intl is in a different position where adoption is the primary currency.

I view llms.txt as a key component of modern Developer Experience. Not having one doesn't prevent LLMs from learning about the library—they just have to process significantly more requests and tokens to eventually reach the same understanding. It’s analogous to a shop that only accepts cash: customers can still buy things, but it adds unnecessary friction. Adding card payments (or in this case, llms.txt) just smooths out the transaction.

While many i18n setups differ, setting them up correctly takes time. llms.txt can provide strong guidance on the "best approach" right at the start of a project. If an AI agent can flawlessly implement next-intl because the context is readily available, it becomes an easier choice for new projects.

If a competitor is easier for an LLM to "read" and implement, developers (and their AI assistants) will naturally gravitate toward that path of least resistance. Providing this file is a good way to ensure next-intl captures that market share early on.

[amannn]

Don't get me wrong, I'm all for making sure next-intl works well with LLMs and this was a core driver behind the introduction of useExtracted to make sure next-intl stays relevant in a world where code might be primarily written by agents.

I'd also like to make best practices around i18n more accessible to agents, but there's more work to be done to see in which way this makes sense for next-intl as a project.

Adding card payments (or in this case, llms.txt) just smooths out the transaction.

If agents automatically pay per crawl, then I'm all here for it. But "adoption is the primary currency" is unfortunately not really the reality I'm in, and if I fail to provide funds for next-intl in the long run, then the community loses because I can't keep working on it.

next-intl's funding currently comes from these sources:

1. Primary sponsor Crowdin
2. learn.next-intl.dev
3. Incredibly kind people who volunteer to sponsor me on GitHub

At least the first two rely on website traffic to be discovered. Website traffic for next-intl peaked about one year ago, and while npm downloads grew wildly since then, website traffic has kept decreasing.

So removing more website traffic is unfortunately not really a solution to keep next-intl in a healthy state.

But again, I agree a lot with the convenience aspect of this, the reality to steer the project into a sustainable direction just doesn't align with /llms.txt so well currently. I hope this helps to provide some background and I also hope to find more recurring revenue in the future that is less dependent on website traffic.