Aizii Learn · AI Search

[← Back to Learn](https://aizii.ai/learn)

# Common llms.txt mistakes

An llms.txt file is a short Markdown file that gives an AI reader a map of your site: your brand name, one short summary, and sections of links to the pages that matter most. The proposal at llmstxt.org defines that shape and nothing more. An H1 with the site name is the only required part. A blockquote summary comes next, then optional context with no headings, then H2 sections of link lists. Most mistakes in llms.txt files come from adding things the proposal never defines, or from treating the file as a place to hide content that belongs on a live page.

Say this first, because it shapes every fix in this guide. No engine is documented as reading llms.txt. Google says Google Search ignores the file, and keeping one will neither harm nor help visibility or rankings in Google Search. The proposal is a community proposal, not a vendor requirement and not a standard. Build the file because it is cheap and because some agents and tools read it, not because it buys a place in an answer. Nothing in this guide claims otherwise.

Also say this plainly wherever the full file comes up. The proposal defines llms.txt only. It does not define llms-full.txt. The full file is a companion convention from documentation platforms, not part of the proposal, not a vendor requirement, and not documented as read by any engine.

## Where it goes wrong

These are the mistakes that show up most often, and each one has a specific fix.

-   The file is at the wrong address. The file lives at a subpath such as a blog folder or a docs folder when it is meant to cover the whole store, or it lives at the root when it only covers one section. The proposal says a file covers the pages under its own path, and where more than one file applies, the most specific one should be used. A store-wide file belongs at the site root. A file that only covers the help center belongs under the help center path.
-   The file has no H1, or the H1 is not the brand name. The H1 is the only required section, and it carries the name of the project or site. A file that opens with a slogan, a keyword phrase, or a section heading has no anchor.
-   The summary is missing, or it is marketing copy. The blockquote after the H1 should carry the key facts needed to understand the rest of the file: what you sell, who it is for, and where you ship or serve. A slogan tells the reader nothing.
-   Links point at pages that do not exist, or at the wrong version of a page. A link to a retired collection, a preview domain, or a page behind a login sends the reader to a dead end. The proposal says links in the file should point to content a reader can use, and for a store that means live public pages. Check every link by opening it.
-   The file lists everything. A file that links to every product, every filter page, and every tag page is a sitemap written in Markdown, not a curated map. The proposal says the file stays small enough to fit in context, with the detail living behind the links. List the collections, the guides, and the policy pages. Leave the full catalog to the category pages and the sitemap.
-   Prices and stock status sit in the file. A price printed in a static text file goes stale the day the price changes, and the file is rebuilt less often than the product page. Prices belong on the product page and in the structured data on that page. The file should link to the product page, not repeat its price.
-   The file invents fields the proposal never defines. Lines such as key and value pairs, allow and disallow rules, priority scores, or training permissions do not exist in the proposal format. A parser built to the proposal reads Markdown headings, a blockquote, and link lists. Invented fields are ignored at best and confusing at worst. If you want to control crawling, that is the job of robots.txt, not llms.txt.
-   The file contradicts the live site. The summary says you ship to one region while the shipping page says another. A guide link carries a note that describes a product you no longer sell. A reader given two answers trusts neither. The file must match the live pages word for word on the facts.

## The steps

**Step 1. Open the live file and read it as a stranger would.** The marketer opens the file in a browser at the site root followed by llms.txt and reads it top to bottom. Does the H1 name the brand exactly as customers know it. Does the blockquote say what you sell and who it is for in one or two sentences. Does every section heading describe the group of links under it. If the file reads like a keyword list instead of a map, rewrite it before you touch anything else.

**Step 2. Fix the shape first.** The developer checks the file against the proposal order: an H1 with the site name, a blockquote summary, optional context with no headings, then H2 sections of link lists in the form of a Markdown link followed by a colon and a short note. Remove any invented field, any allow or disallow line, and any front matter block at the top of the file. The proposal format is plain Markdown, and anything outside it does not belong.

**Step 3. Cut the list down to what earns a place.** The marketer goes through every link and asks one question: would a reader who knows nothing about the store need this link to understand what we sell or how to buy. Keep the collections, the buying guides, and the shipping and returns pages. Cut individual product pages, sale pages that will expire, filter and search pages, and anything behind a login.

**Step 4. Check every link and every fact against the live site.** The marketer opens each linked page and confirms it loads, shows the right content, and matches the note beside it in the file. Fix any link that lands on a retired page or a wrong address, and any note the live page no longer states. Remove every price and stock statement from the file itself, and replace each one with a link to the product page where the current price lives.

**Step 5. Fix the address and the serving.** The developer confirms the store-wide file sits at the site root and serves as plain text over HTTPS with a 200 status. If the file is meant to cover only one section, confirm it sits at that section path and covers the pages under it. If the file moved, remove the old copy or redirect it, and update any link that pointed at the old address. Publish it and verify it loads.

**Step 6. Put the rebuild on a routine.** The marketer and the developer agree a trigger: the file is rebuilt whenever a linked guide changes, a collection is added or removed, or a policy page is rewritten. A file that lags the site teaches the reader the old store. If it has grown past what a reader can use in one document, split it by section instead of shipping one oversized file.

## The machine-readable piece

This is a broken version of a store file, with the mistakes left in so you can see them. Do not copy this version. Each line below breaks a rule from the proposal or from this guide.

```
# Best candles cheap buy now
> Buy the best candles and soap. Huge sale. Limited time.
## Products
- [All products](https://[yourdomain.com]/collections/all): Every product we sell, all in one place.
- [Old winter collection](https://[yourdomain.com]/collections/winter-last-year): Winter scents, now 40 percent off.
- [Gift sets](https://[yourdomain.com]/products/gift-set): Gift sets from 25 dollars, in stock now.
priority: high
allow-training: yes
```

The H1 is a search phrase, not the brand name. The blockquote is a sales pitch, not a summary. One link points at a retired collection. One note prints a price and a stock claim that will go stale. The last two lines invent fields the proposal never defines. A parser built to the proposal has no use for them.

This is the same file fixed. Copy this shape, replace the placeholders with your own pages, and keep the notes short and factual.

```
# Example Candle Supply
> Example Candle Supply sells hand-poured candles and bar soap, made in small batches and shipped across the country.
## Shop
- [Candles](https://[yourdomain.com]/collections/candles): Hand-poured candles, listed with scent, size, and burn time.
- [Bar soap](https://[yourdomain.com]/collections/bar-soap): Bar soap, listed with scent and ingredients.
- [Gift sets](https://[yourdomain.com]/collections/gift-sets): Candle and soap gift sets, listed with contents and box size.
## Guides
- [Candle care](https://[yourdomain.com]/pages/candle-care): How to burn a candle cleanly, trim the wick, and store it.
- [Soap ingredients](https://[yourdomain.com]/pages/soap-ingredients): What goes into each bar, and which bars suit sensitive skin.
## Help
- [Shipping and returns](https://[yourdomain.com]/pages/shipping-returns): Shipping regions, delivery times, and the returns policy.
```

Check the fixed file after every rebuild. Open your domain followed by llms.txt and confirm the H1 names the brand, every link loads, and no line invents a field.

## What it looks like when it works

On the page, nothing looks different to a human reader. This is a file for machines, not a page for shoppers. The change is under the surface: the file opens with the brand name, the summary states what the store sells in one sentence, and each section points at live pages with notes that match those pages. A reader that starts from the file can reach the collections, the guides, and the help pages without wading through the whole catalog.

Say you run a candle and soap store, selling hand-poured candles and bar soap in small batches, plus a candle care guide and a soap ingredients page. The marketer opens the file and finds the mistakes at once: the H1 is a search phrase, the summary is a sales pitch, one link points at a retired winter collection, and two invented fields sit at the bottom. The developer removes the invented fields and rewrites the H1 with the brand name. The marketer rewrites the blockquote, cuts the retired collection, and removes the price and stock claim from the gift set note. When the winter collection retires again next season, the file is rebuilt the same day the collection closes. A shopper asks an assistant how to keep a candle from tunneling. The assistant can follow the file to the candle care guide and cite the store with the steps as written.

### The fundamentals

These mechanics are shared across every engine and covered in full in their own guides:

-   •The llms.txt format and a complete sample, see our [guide to writing a good llms.txt file](https://aizii.ai/learn/ai-search/how-to-write-a-good-llms-txt-file)
-   •The markup AI systems read to verify facts, see our [product schema guide](https://aizii.ai/learn/ai-search/how-to-add-product-schema-markup-for-ai-search)
-   •The complete crawler rule reference, see our [robots.txt guide for AI search](https://aizii.ai/learn/ai-search/how-to-write-robots-txt-for-ai-search)

## What this guide does not cover

This guide covers the common mistakes in an llms.txt file and how to fix them. It does not cover the full technical audit that checks every crawler and every page, the transaction layer that lets an agent complete a purchase, or the per-engine playbooks for getting each assistant to find you. It does not cover where the file goes on a site with several sections, which has its own guide. That is a separate job, and this guide is the map, not the whole trip.

## Check your own store now

Run the free scan. One click, no signup, and it shows whether AI crawlers can read your store today: aizii.ai/aeo-scan

[Run the free scan](https://aizii.ai/aeo-audit)

## Sources

-   llms.txt proposal (specific): [https://llmstxt.org](https://llmstxt.org/)
-   Google Search Central, Optimizing your website for generative AI features on Google Search (specific): [https://developers.google.com/search/docs/fundamentals/ai-optimization-guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
-   Google Search Central, Product structured data (specific): [https://developers.google.com/search/docs/appearance/structured-data/product](https://developers.google.com/search/docs/appearance/structured-data/product)
-   Schema.org, Product type (specific): [https://schema.org/Product](https://schema.org/Product)

Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
