Assistants like ChatGPT, Claude, Gemini and Perplexity increasingly answer
shopping questions. When one of them looks at a shop, it gets the same HTML a
browser does: menus, scripts, banners. An llms.txt is a plain text summary at the
root of your domain, written for them instead. This module generates that file
from your PrestaShop and keeps it current on a schedule.
YOUR WORDS COME FIRST
The file opens with three texts you write: the shop name, a short context, and
the rules an assistant must not get wrong — delivery, returns, what you do not
sell, which page decides on price and stock. Everything below that is generated
from your catalogue: categories, products grouped one section per category,
brands, suppliers and CMS pages. The part a model cannot work out on its own is
the part you control, and it is the first thing it reads. The back office shows
you how to write both texts and gives you a full example to copy.
ONE FILE PER LANGUAGE
Every active language gets its own file, built entirely in that language: your
context, the product names, the descriptions, the URLs and the generated section
headings, which ship translated for 23 languages. A language you switch on later
is picked up on the next run. The primary language keeps the plain /llms.txt
name, so the URL you already advertise never moves; the others are llms-de.txt
next to it or de/llms.txt one folder below, whichever you choose. Every file
links to the other language versions.
RUNS BY ITSELF, AND PROVES IT
A cron URL or a CLI command rebuilds everything, and one call covers every
language of every shop. The write is atomic, so a crawler is never served a
half-written file. A button in the back office fetches your files over HTTP the
way a crawler would and reports the status code for each one, because a file can
be on disk and still be unreachable.
AI CRAWLER ACCESS
The module builds a robots.txt block from the file names your shop really
produces, not from a guessed pattern. Sixteen AI crawlers can be allowed or
refused one by one, grouped by what they actually do: assistants that cite live
pages, fetches a person asked for, and collection for training.
MULTISTORE
Each shop gets its own files, its own settings and its own texts, and a shop
served from its own folder writes into that folder.
- Generates llms.txt into the root of your shop, on a cron schedule
- One file per active language, built entirely in that language
- Section headings translated for 23 languages, English elsewhere
- Your project name, short context and additional context, per language, always at the top of the file
- Guidance panel in the back office with a worked example of both texts
- Sections: categories, products grouped per category, brands, suppliers, CMS and built-in pages, plus an optional block
- Every product listed once, under its default category
- Cross links between the language versions of the file
- robots.txt block for 16 AI crawlers, built from your real file names
- Reachability test: fetches every file over HTTP and reports the status code
- Preview before writing, and a per-file view of what is on disk right now
- Size controls: maximum products, maximum per category, description source and length
- Exclusions by product, category, brand and supplier, with tick lists
- Multistore: per-shop files, settings, texts and paths
- HTTP cron endpoint and CLI runner, both with a dry run and a single-language rebuild
- Atomic write, so a crawler never sees a partially written file
