02 · Structure · Open source
llms.txt for Magento 2
angeo/module-llms-txt
Generates spec-compliant llms.txt, llms-full.txt and streaming JSONL so an AI reads a structured summary of your catalogue in one request instead of crawling thousands of rendered pages and guessing.
Free and MIT-licensed. No licence key, no account, no telemetry.
What it does
Four things worth knowing before you install.
Three formats, one pipeline
A short llms.txt index, a full llms-full.txt body, and streaming JSONL for large catalogues — generated in a single pass over the data rather than three separate exports.
Safe on large catalogues
Batched, cursor-based queries and a resolved-once stock lookup. Generation runs on cron and CLI, never during a storefront request.
Markdown page mirrors
Storefront pages can be served as clean Markdown on request, which is what most LLM fetchers prefer over parsing a themed HTML page.
Atomic writes and an async admin UI
Files are written atomically, so a crawler never reads a half-generated file. Generation can be triggered from the admin without blocking the request.
Extensible by design
Entity providers and format renderers are interfaces, so you can add a custom entity type or output format without patching the module.
Perplexity is the clearest beneficiary. PerplexityBot demonstrably reads llms.txt. For ChatGPT Shopping the authoritative source is a registered product feed, not this file — so treat llms.txt as content structure, not as a substitute for the feed.
Requirements
| Magento | Open Source / Adobe Commerce 2.4.x |
| PHP | 8.2+ |
| Output | llms.txt · llms-full.txt · JSONL |
| Scope | Per store view |
| Licence | MIT |
Questions
Which AI engines actually read llms.txt?
Adoption is uneven and worth stating plainly: Perplexity reads it, others treat it as one hint among many. It structures your content for any fetcher, but it is not a switch that turns on AI visibility by itself.
Will generation slow down my store?
No. It runs on CLI and a dedicated cron group, not on page render, and uses batched queries so memory stays bounded on large catalogues.
Does it handle Page Builder content?
Yes — Page Builder markup is reduced to readable text rather than being emitted as raw layout HTML.
Can I control what goes in?
Yes. Entity types and scope are configurable, so you can exclude categories or content you do not want summarised.
Works with
Each module is independent, but these three are the ones most often installed alongside it.
Not sure whether you need this one?
Run the free scanner on your domain, or install the CLI audit. Both report which signals your store currently fails, so you install only what closes a real gap.