---
title: LLMS.TXT + Webflow | Sygnal Technology | Blog
published: 2025-05-19T19:57:52.000Z
public-url: https://www.sygnal.com/blog/llms-txt-webflow
domain: www.sygnal.com
webflow-site-id: 59b8d49f7fdf9700017d780f
webflow-page-id: 59d3f382f210ca00015829d6
webflow-collection-id: 59d3f382f210ca00015829d5
webflow-collection-item-slug: llms-txt-webflow
---

April 14, 2025

# LLMS.TXT + Webflow

## Is AI Optimization ( AIO ) the new SEO?

[## #ai](https://www.sygnal.com/tags/ai.md)[## #seo](https://www.sygnal.com/tags/seo.md)

![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/67fcf69ae305347f45ba23ff_ChatGPT%20Image%20Apr%2014%2C%202025%2C%2011_49_17%20PM.avif)
![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/67fcf69ae305347f45ba23ff_ChatGPT%20Image%20Apr%2014%2C%202025%2C%2011_49_17%20PM.avif)

**In May 2025,** [**Google announced**](https://blog.google/products/search/google-search-ai-mode-update/) **a number of initiatives surrounding AI integration into its search platform.**

In fact, most everything [announced at Google I/O](https://blog.google/technology/ai/google-io-2025-all-our-announcements/) appears to be AI-related.

The team at Sygnal has been watching this trend since mid-2024, and wanted to know...

* Is the nature of SEO changing?
* How does AI affect our SEO strategy?
* How do we keep client sites relevant?
* Is sites getting "absorbed" by LLMs a good thing, or a bad thing?

*And we decided to find out.*

## The Experiment

Just over a month ago, we began an experiment to see whether some of the newer standards for exposing web content to AI were actually being used.

In particular we wanted to know if LLMS.TXT and Markdown actually make a difference to how your site is indexed. We'll discuss more about these standards below.

On *this* site, **sygnal.com**, we implemented two things-

1. An `/llms.txt` markdown page as a controlled directory of the content we want LLMs to focus on.
2. A markdown page for every page on our site, generated from that page's content.

We launched this.

We made zero announcements to Google, Google Search Console, or any other search engine.

*And we sat back and watched.*

## 30 Days Later

Here's our traffic over the past 30 days;

* **Blue** is our standard webpage traffic for the past 30 days.
* **Red** is LLMS.TXT traffic.
* **Green** is the page-specific Markdown traffic.

You can see spikes as crawlers absorb this content.

![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/682fe95fc966c709f3efde83_ph.png)

And here are the stats totals;

![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/682fe93933d6bf1573241c40_stats.png)

### Who is Visiting?

Well, everyone. Here are the top 5 crawlers appearing in that data;

![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/682feb70d7dc3908909e32da_ua.png)

In terms of page crawl count over the past 30 days, that's;

1. META / Facebook
2. OpenAI's GPTBot
3. Googlebot

### Did This Impact Human Traffic?

> Now we need to determine whether this change has impacted our visibility in LLMs like Google Search AI, ChatGPT and Claude.

While we haven't studied this heavily, we can definitely see new traffic with **utm\_source** querystring indicators, two in particular;

1. chatgpt.com ( 53 referrals )
2. perplexity ( 8 referrals )

### Did LLMS.TXT make a difference?

If we look at all **utm\_source**'d traffic and extend the timeperiod to 180 days, you can see a clear jump in April on the blue ( chatgpt-referred ) and purple ( perplexity-referred ) lines, *after* we implemented LLMS.TXT and page-specific Markdown.

![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d783a/682ff361be47c145aa9e2f76_refer.png)

*Well, that excalated quickly.*

## What Did We Learn?

Here's what we know so far...

* META, OpenAI, and Googlebot are all actively looking for `/llms.txt`
* All 3 are crawling links referenced there, extensively
* The addition of LLMS.TXT and per-page markdown correlates with a significant increase in identifiable human traffic from LLM's - ChatGPT and Perplexity in particular.

With Google's announcements this week we hope to observe visible traffic increases there as well.

**Now let's look at the tech, and Sygnal's own implementation of LLMS.TXT.**

‍

## What is LLMS.TXT?

Conceptually, you can think of `llms.txt` like a `sitemap.xml` for Large Language Models ( LLMs ), but instead of simply containing links, it also contains structured organized markdown content that the LLM can immediately use.

**It's an abstract if your site's most important content.**

Here's an example, Webflow's API documentation...

[https://developers.webflow.com/llms.txt](https://developers.webflow.com/llms.txt)

When an LLM processes your LLMS.TXT, it can know your products, services, hours of operations, recent blog articles, and much more. In most cases, LLMS.TXT also acts as a directory to your most valuable content, with links to other markdown files that represent your individual products and service offerings.

For an LLM such as Claude, Gemini, or ChatGPT, this means a much more efficient way to find the content that you need. It starts with `/llms.txt` and then can continue to examine other files later as needed, in response to the user's prompt.

## How does it benefit Website owners?

In truth, only time will tell.

In theory, these are some key benefits a good LLMS.TXT implementation can give you.

### AI Optimization ( AIO ), the new SEO

Your website becomes much easier for AI's to process and make sense of. This makes it more likely that they'll find the right information quickly, and that likely means you content will simply be shown more.

Think of the early days of Google SEO, we're at that same stage. It's good to be in the SERPs.

### Better User engagement

Today, LLMs are still largely used by a geekier crowd that favors technical experts and business professionals. However, it won't be long before LLMs are helping people with much more general tasks;

> Help me plan my vacation and book my hotels & tours

> Organize my grocery list, here's a photo of what's in the fridge, we need chili and salad for 4 people tonight

> I want to find the right car, used is ok, low miles within 3 years new. Help me search dealerships and auction sites for the right car.

This means that the LLM will need not just general information, but specific information and agents to perform tasks with. Getting prices, ordering groceries, checking business hours.

**The truth is, modern websites *aren't ready for this* .**

But LLMS.TXT is one of the first steps here, specifically designed to connect websites to LLMs. The ones who embrace it first will be the ones that are most likely to get LLM attention, and visibility.

### Better Integration Support

Have a content-heavy site, full of information, product catalogs, and support docs?

LLMS.TXT can be used by your directly integrated systems too. Power an AI chatbot on your website, or an internal customer sales and support desk, all from your site's content and CMS.

And done right, it's always *perfectly* current.

## Do you need an LLMS.TXT?

**It's an exciting future, but today LLMS.TXT is *not useful*  for all sites.**

It works best when;

* Your site is text-heavy, with valuable content
* All of that content is public, not paywalled or login-gated
* The content of your site is highly valuable in answering questions that an LLM user might have.

> This is why currently, technical documentation is a popular use for LLMS.TXT.

As of April 2025, LLMS.TXT is probably not that useful for a photography portfolio, or a media site.

## How to Setup Webflow + LLMS.TXT

Currently, Webflow's hosting platform doesn't offer any form of native LLMS.TXT or markdown generation support.

Here are two ways you can add it.

### Limited Static Support

If you just want basic support, you can hand-create your LLMS.TXT markdown content, or use an LLM like ChatGPT to analyze your key pages and help you write one.

Once you have a text file ready with the markdown content you need, you can then upload the TXT file to Webflow, and setup a Webflow redirect from /llms.txt to your file.

This is relatively easy.

* Upload your LLMS.TXT file to your Webflow site's assets
* Create a redirect from /llms.txt to that asset file
* Republish your site

> Unfortunately, this only solves for the LLMS.TXT file itself, and you must repeat this manual process any time you want to update it. More importantly, the other pages on your site will not have markdown extracts, which is a substantial part of making your site AI-Optimized.

### Full, Dynamic Support

Another approach is to use a solution like Sygnal's Hyperflow LLMS.

Our solutions is built as a reverse proxy which runs on top of your Webflow-hosted site. It *dynamically*, create LLM-friendly markdown content you need directly from your Webflow site's current HTML.

With this approach, you can deliver;

* An auto-generated, auto-updated `/llms.txt` file
* An auto-generated markdown page for every page on the site. For example, your `/contact` page would be given an equivalent `/contact.md`.
* Fully dynamic content. Publish a new blog article or product, and it's instantly visible in your LLMS.TXT as well.

Sygnal's Hyperflow LLMS provides full LLMS.TXT support for Webflow-hosted websites.

All content of the LLMS.TXT and page-specific markdown extracts is directly controlled within the Webflow designer, so that you can fully customize it to your needs.

Once it's setup, it works automatically with no special user actions. Any changes to the site, like new blog entries, or a pricing list change, are automatically picked up and exposed in the llms.txt.

As a demonstration, you can see a fully-functioning implementation on Sygnal's site.

[https://www.sygnal.com/llms.txt](https://www.sygnal.com/llms.txt.md)

> Note the `.md` extension on all links in this file. If you copy-paste those into your browser, you can see the individual page markdown extracts as well.  
> e.g. [https://www.sygnal.com/index.md](https://www.sygnal.com/index.md.md)

If you're interested in Sygnal's **Hyperflow LLMS** for your Webflow-hosted site, [contact us](https://www.sygnal.com/contact.md) to learn more.

## The Future of LLMS.TXT

This is *all* conjecture, but I expect that the LLMS.TXT standard will evolve in a number of ways.

* Standardization and improved use of dates, to indicate current content more effectively.
* Key information structures, like;
  * Important alerts
  * Special offers and new products
* Expanded support for media such as photos and video

### Querying Support

In my view, a key missing feature of LLMS.TXT is that an LLM has no clear way to query your site for specific information... such as to find a particuar product, or news about a specific topic.

Soon I anticipate the ability for the LLM to pass keywords in on a querystring and get a markdown search response efficiently.

### Specification for Page-Specific Markdown

Currently when LLMS.TXT links to a page's Markdown, there is no specification for how that Markdown should be formed.

This should cover details like;

* The public human-readable URL to link to when citing content from that page.
* Licensing and reuse. Is this open source content or copyrighted? Public domain or creative commons? These matter more as AI is being used to assemble content and needs to correctly cite sources.
* Attribution. Who must be attributed in the citation?

Other standards like EXIF already support this substantial metadata for images, and Open Graph and JSON+LD offer specific views of web content for specific uses.

> Personally, I advocate for a frontmatter section that can be used to identify these specific details.

### Agents

Extending this further, it seems quite possible that a kind of Website-integrated MCP could exist that allows for specific LLM interactions, such as;

* Quote generation
* Shopping cart management and purchasing
* Account creation and management

### Adoption into Web Publishing Platforms

Web publishing platforms like Webflow will likely add;

* Native support for LLMS.TXT ( or whatever it evolves into ).
* Support for MDX-like capabilities, components-as-markdown that allow the dynamic embedding of code blocks, tables, graphs and other content

## Where is this all going?

**Maybe we should call this... Web 4.0**

The team at Sygnal has a very strong information engineering background. We anticipate that *eventually* *most* websites will shift towards an "agent" model. Sites will continue to perform their current functions like information-sharing and order capture, but under the direction of a LLM.

This makes things much easier for users. Rather than digging through Amazon.com, your AI's will know what you like, and recommend new products and the best, fastest, and cheapest suppliers to choose from.

For web designers, this will mean radical changes to the way we think of design and user interfaces, and a lot of websites will simply evaporate to the point that they have no user interface at all.

This is a long way away. Before then, there will be a complete rethink on UX's, protocols, directory services and search engines, and technical infrastructure.

For now, you're safe.

*But the first steps have already been taken.*

30 years ago, we designed websites exclusively for *people*.

15 years ago, we added search engines and SEO to the mix as a second critical "user" of our website content.

Today, it's time to think seriously about what AI and LLMs need to best represent our content to users and deliver the best results for our businesses.

‍

## Discussion

Want to support our team?

*Passion* drives our long hours and late nights supporting the Webflow community. Click the button to *show your love*.
[![](https://cdn.prod.website-files.com/59b8d49f7fdf9700017d780f/635c70196c0ae0320d4eeeb5_beer.svg)

Buy us a beer](http://buymeacoffee.com/sygnal)

## Need Help?

Click a related service to learn more.

No items found.

Learn

>
[Sygnal's Blog](https://www.sygnal.com/blog.md)

>

LLMS.TXT + Webflow