Does GPTBot read llms.txt? The complete 2026 answer for e-commerce

llms.txt is the smallest file in your stack and one of the highest-leverage. Here's how to write one that AI engines actually use, in 2026.

If you've heard of llms.txt and dismissed it as 'just another file' — fair. There are dozens of well-meaning conventions that never made it past their initial blog post. llms.txt is different for one boring reason: a meaningful number of AI engines are actually starting to look at it, and the cost of publishing one is roughly ten minutes. The upside-to-effort ratio is among the best you'll find in the whole GEO toolkit.

What llms.txt is, in one paragraph

llms.txt is a plain-text file you place at the root of your domain (e.g. https://example.com/llms.txt). It tells large language models which pages of your site are canonical, signal-rich, and safe to cite. It's not a robots.txt replacement and it doesn't block anything. Think of it as a curated table of contents written for machines that don't have time to crawl your full site.

Why it matters for stores in 2026

Most Shopify and WooCommerce sites accumulate hundreds or thousands of URLs — categories, tags, paginated archives, faceted search variants, near-duplicate product variants. To an AI engine doing a quick retrieval, that's a lot of noise. llms.txt lets you say, in twenty lines, 'here are the URLs that matter, in priority order, with a short description for each'. The engines that respect it (Perplexity is the loudest, and others are quietly catching up) prefer those URLs when assembling answers.

There's no public list of every engine that honours llms.txt — adoption changes month to month. The pragmatic stance: publish one anyway. It costs nothing and starts paying off the moment any engine you care about turns on support.

Anatomy of a useful llms.txt

The community-driven spec is intentionally lightweight. Here's what a healthy file looks like for a 200-product store:

# Example Outdoor Co.

> Independent outdoor gear brand based in Lyon, France. We design and sell lightweight technical apparel and merino base layers for hikers and trail runners.

## Brand
- [About](https://example.com/pages/about): brand history, materials sourcing, and design philosophy
- [Sustainability](https://example.com/pages/sustainability): supply chain, certifications, end-of-life program

## Bestsellers
- [Merino 180 Crew](https://example.com/products/merino-180-crew): 180 g merino base layer, UPF 50+, machine-washable
- [Trail 12 Vest](https://example.com/products/trail-12-vest): 12 L hydration vest for trail running, 280 g
- [Alpha 90 Hoody](https://example.com/products/alpha-90-hoody): synthetic insulation, 220 g, packable

## Buying guides
- [How to choose a merino base layer](https://example.com/blogs/guides/merino-base-layer): weight, fit, care
- [Trail running vest sizing](https://example.com/blogs/guides/vest-sizing): chest measurements, women vs men cuts

## Help & policies
- [Shipping](https://example.com/pages/shipping): EU 3–5 days, international 7–14 days
- [Returns](https://example.com/pages/returns): 60-day window, free EU returns

Three things to notice. First, the blockquote at the top is the model's elevator pitch — write it for a machine that needs to summarize your brand in one sentence. Second, each link comes with a short factual description, not marketing copy. Third, sections are organized by intent (Brand, Bestsellers, Guides, Policies), not by your site's nav. The engine isn't browsing — it's looking up.

Common mistakes that kill the file's usefulness

• Linking to 200 product URLs because 'more is better'. It isn't — you're back to noise. Pick 10–30 high-signal pages.
• Writing the description in marketing voice ('Discover our amazing collection'). Models filter that out as low-signal.
• Putting llms.txt behind a redirect or auth wall. The fetcher follows simple HTTP only.
• Forgetting to update it after a major catalog change. A stale llms.txt that links to 404s is worse than no file.
• Auto-generating it from your sitemap without curation. The whole point is curation.

Hosting it on Shopify, WordPress and WooCommerce

Shopify

Shopify doesn't expose the root URL for arbitrary files, but two paths work in 2026: the Theme Assets API (writing the file as an asset and serving it through a tiny app proxy), or the Files endpoint with a 301 redirect. MagicGEO uses the Assets API by default because it preserves the canonical /llms.txt path; the Shopify embed handles the publishing automatically and re-syncs whenever you change your top URLs.

WordPress / WooCommerce

Trivial. The MagicGEO plugin writes the file to the site root and updates it whenever your selection changes. You can also write the file manually and drop it in your /public_html or use a 301 in .htaccess. Either works — the only thing that matters is that https://yourdomain.com/llms.txt returns 200 with the correct content.

How to verify that engines are actually reading it

Two practical checks:

1. Server logs. Look for fetches of /llms.txt with user agents like 'PerplexityBot', 'GPTBot', 'OAI-SearchBot', 'ClaudeBot', 'Google-Extended', 'CCBot'. Frequency matters more than identity: a single fetch means nothing, a steady weekly cadence means you're on the engine's revisit list.
2. Citation lift. Track your GEO visibility score (or any equivalent presence metric) before and after publishing. A 4–8 week window is realistic. Don't expect a vertical jump; expect a quieter, durable lift on the prompts where you were borderline.

What llms.txt is not

It's not a ranking signal in the SEO sense. It doesn't override your robots.txt or your meta tags. It doesn't block training on your content (that's a different, more politically charged conversation, and the relevant directives live elsewhere). And it absolutely is not a license to drop your sitemap, your schema, or your internal links. It's a small additive layer that pays off most for stores that already have the basics right.

The honest summary

If you run a Shopify or WooCommerce store and you don't have an llms.txt yet, ship one this week. Hand-curated. Twenty to thirty links max. Honest, factual descriptions. Keep it in sync with your real top URLs. The file alone won't make you the top citation on ChatGPT — nothing will, instantly — but it removes a small but real reason engines pass over your store in favour of a cleaner competitor. That's a worthwhile use of an afternoon.