ARTICLE

robots.txt marks what should stay unfetched; llms.txt instead highlights the material most worth understanding. One file sets limits, the other provides direction: as AI takes over more search, both files should be ready. For a custom site moving toward AI-Ready, this index is now standard equipment.

In planning talks, decision-makers often ask whether this is just another new label. That reaction is fair, because the spec is still young. Longer observation suggests: technologies adopted early usually yield the larger gain. sitemap.xml also moved from optional to expected. Few sites publish this file today; once it becomes routine, the early advantage is gone.

The notes below cover the idea, the format, and where to place the file, in a sequence you can follow. For the wider build process, also read building an AI-Ready site from the ground up.

What llms.txt Means

Answer.AI founder Jeremy Howard proposed this site convention in 2024. The premise is:

A model does not need every markup tag on a site. It needs a precise summary of what matters. llms.txt is that summary.

The contrast looks like this:

Major tools such as Anthropic Claude, OpenAI, and Perplexity have started consulting this file. It is not mandatory yet, but a common result after publishing it is: citations stay closer to the site’s own wording. Questions about brand positioning or service scope are often answered straight from the file.

Standard File Format

The body is Markdown. The skeleton is fixed; the wording is flexible. Example:

# Brand name> A web-building team founded in 2002, focused on smart sites that AI can interpret correctly.> It serves more than 3,000 companies and aims for official sites that AI search can read and cite.## Core services- [AI-Ready site package](https://example.com/services/ai-ready): a full build that is friendly to AI- [AI visibility check](https://example.com/services/diagnostic): a free half-minute look at how often AI cites the site## Key articles- [What an AI-Ready site is](https://example.com/insights/ai-ready-website-building-guide): the definition and three requirements- [AI Overviews return answers, not ranks](https://example.com/insights/ai-overviews-vs-traditional-search): how AI search selects content- [Schema.org markup notes](https://example.com/insights/schema-markup-web-design-guide): seven markup types designers should know## Company facts- Founded: 2002- Clients served: 3,000+ companies- Base: a major metro area

Keep four writing points in mind:

• Top heading: site or brand name
• Opening blurb (>): three to five sentences that define the site
• Second-level sections: important pages grouped by topic
• Link plus short note: a colon and one clarifying line after each URL

Five Setup Steps

Step 1: Choose Pages

First select the 10 to 30 URLs you most want a model to cite. Typical picks:

• Brand and company overview
• Core service or product pages
• In-depth articles covering main keywords
• Case write-ups
• Question-and-answer pages

Do not list the entire site: this file is a shortlist, not a sitemap. Too many entries dilute the point.

Step 2: Write the Index

Draft it with the skeleton above. Watch three details:

• Open with definition sentences a model can quote directly
• The note after each link is the model’s first impression of that page, so make it exact
• Pair the brand name with measurable facts (founding year, client scale) to strengthen credibility

Step 3: Add a Full File

You can also prepare a full-text companion that places important articles in one file:

# Full brand content file[Site intro matches the index file]## Article: What an AI-Ready site is[Full text, about 1,500 to 3,000 words]---## Article: AI Overviews return answers, not ranks[Full text, about 1,500 to 3,000 words]---[Further articles...]

The gain is that a model can read the essentials in one pass instead of fetching many URLs, so citations get sharper.

The cost is size: the file can grow to hundreds of kilobytes or past 1MB, and updates take more effort.

Practical split: smaller sites keep both the index and the full file; larger sites keep the index and limit the full file to the 10 to 20 most important pieces.

Step 4: Place at Root

Both files belong at the site root:

https://yoursite.com/llms.txthttps://yoursite.com/llms-full.txt

Do not put them in a subdirectory (for example /docs/llms.txt), or models usually will not find them.

With Laravel, you can serve the file like this:

// routes/web.phpRoute::get('/llms.txt', function () {    return response(view('llms-txt'), 200, [        'Content-Type' => 'text/plain; charset=utf-8',    ]);});

A simpler option is a static file at public/llms.txt.

Step 5: Keep It Updated

When an important article or service is added, write it back into the index. Three rhythms help:

• Run a routine check once a month
• When a new article ships, decide immediately whether it belongs
• Each quarter, drop dead links and items that are no longer central

Advanced Practices

Practice 1: Core Definitions

Give definitions their own block in the file:

## Core definitions**AI-Ready site**: a site structured so AI search can read and cite it correctly. The essentials are structured data, clear definition paragraphs, and authority signals.**AISO (AI Search Optimization)**: site optimization aimed at AI search, and an extension of traditional SEO.**Hub and cluster**: one topic hub linked to several deeper subtopic pages, forming a knowledge network.

Models can reuse these sentences when answering related questions.

Practice 2: Add FAQs

Put the questions visitors ask most often into the file as well:

## Frequently asked questions**Q: Must an AI-Ready upgrade rebuild the whole site?**A: Usually no. The work is adding Schema, the index file, and restructuring content.**Q: How soon do results show?**A: Often within one to three months, as the AI citation rate rises.

Practice 3: Pair with Schema

The index can also point to important Schema entry points (implementation detail is in a practical guide to structured-data markup):

## Structured dataSchema.org types used across pages include:- Organization (organization facts)- Article (articles)- FAQPage (question blocks)- BreadcrumbList (path navigation)Markup overview: https://example.com/schema-overview

Four Common Mistakes

• ❌ Too many links: this is a shortlist, not a sitemap; a hundred links only blur the focus
• ❌ Relative paths: every URL must be absolute and include https://
• ❌ Stale pages: a listed URL that returns 404 clearly hurts trust
• ❌ Written only for models: if people can also read it smoothly, the structure is actually clear

What Changes After Launch

Teams that placed both the index and the full file at the root have reported:

• Cited text stays closer to the brand’s own definitions
• Core phrases such as site building and AI-Ready websites are cited more often
• Organic traffic rises indirectly as well

Early Movers Gain More

The spec is still new, but adoption may move quickly, much as sitemap.xml shifted from optional to expected. Sites that publish it now can take the early-mover gain: few sites offer the file, so models tend to weigh those that do more heavily.

The bar is low and the return is clear: one engineer can produce a first version in about a day, and later upkeep is usually under half an hour a month. If the site still needs content marketing, search visibility, or brand recognition, it is worth starting now.