ai.txt vs llms.txt is a genuine point of confusion, because the two files sound interchangeable and both sit at the root of a domain talking to AI systems. They do opposite jobs. One is a permissions file that tells AI systems what they may do with your content; the other is a content index that tells AI systems what your content is. If you run an online store, you almost certainly want one of them, and you may not need the other at all.

What ai.txt is

ai.txt is a proposed convention — popularized by Spawning and adopted piecemeal — for declaring whether AI systems may use your site's content for model training. It reads like a robots.txt for machine learning: allow or disallow directives, sometimes by media type (text, images, audio), aimed at the crawlers that gather training data. Its spirit is opt-out: it exists mainly for publishers and creators who do not want their work absorbed into models.

Two practical caveats define its real-world weight. First, adoption is inconsistent — ai.txt is not an enforced standard, and there is no guarantee any given crawler reads it. Second, the AI companies that do offer opt-outs mostly implement them through robots.txt user agents (GPTBot, ClaudeBot, Google-Extended, CCBot), which means robots.txt remains the permissions mechanism with actual teeth. If your goal is restricting AI use of your content, the Shopify robots.txt guide covers the directives that are genuinely honored.

What llms.txt is

llms.txt comes from the other direction entirely. Defined by the llmstxt.org proposal, it is a Markdown-formatted index of your most important pages — your store summary, citable facts, and grouped links with one-line descriptions — written so a language model can understand your site quickly and cite the right URLs. It is not about permission; it assumes you want AI engines reading you and hands them the guided tour.

For an online store, that difference is everything. When a shopper asks ChatGPT or Perplexity for a product recommendation, the engines that know your catalog, your shipping terms, and your best guides can name and cite you. llms.txt is the single cheapest signal that feeds them exactly that, which is why it sits near the top of every answer engine optimization checklist.

The comparison, side by side

  • Job: ai.txt sets permissions for AI training. llms.txt curates content for AI consumption and citation.
  • Direction: ai.txt is defensive (keep AI out, or set terms). llms.txt is offensive (help AI in, on your terms).
  • Format: ai.txt uses robots-style allow/disallow directives. llms.txt uses human-readable Markdown with links and descriptions.
  • Enforcement: ai.txt is a courtesy signal with patchy adoption; robots.txt user-agent rules are the enforceable version of its intent. llms.txt needs no enforcement — it works by being useful to any model that reads it.
  • Who needs it: ai.txt suits publishers protecting original work. llms.txt suits any business that benefits from being found and recommended by AI — which is essentially every store.

The two are not rivals and can coexist: a news site might use both (permissions tightened, key explainers indexed), and a store might use llms.txt alone with permissive robots rules.

What an online store should actually do

For a store, the decision tree is short. You want AI engines to know your products, repeat your shipping and return facts correctly, and cite your pages in shopping answers — so ship an llms.txt, keep AI crawlers allowed in robots.txt, and skip blocking-oriented files entirely. Blocking GPTBot to "protect" product descriptions mostly protects your competitors' visibility instead: the engine still answers the shopper's question, it just answers it with someone else's store. That trade-off, and the crawler-by-crawler detail, is covered in AEO best practices for Shopify.

On Shopify specifically, neither file can be dropped at the root through the admin — Shopify owns the storefront root. llms.txt is served through an app proxy or edge worker (the llms.txt generator handles creating a valid one), and training permissions belong in robots.txt.liquid, which Shopify does let you edit. Once served, verify the file actually returns 200 by opening your-store.com/llms.txt — and then check whether the engines are picking you up with an AI visibility probe.

RankEngine generates and serves llms.txt for Shopify stores, keeps it synced to the live catalog, and ships the AI-crawler robots rules alongside it — the offensive half of this comparison, automated and verified.