How Vercel Responded When 10% of New Signups Started Coming From ChatGPT

Most websites are written with two readers in mind, the person browsing the page and the search engine crawling it. There is now a third reader to plan for, which is the AI assistant pulling content into a chat answer. The way that third reader finds and parses your site is changing fast, and the brands paying attention are starting to publish a new kind of file built specifically for it. Vercel is one of the clearest examples of why that matters.

The Vercel Example

Vercel is a deployment and hosting platform used by a large share of modern web development teams. In late 2025, Vercel reported that roughly 10% of its new signups were coming from ChatGPT referrals. That is a meaningful chunk of their pipeline arriving from a channel that did not exist as a measurable acquisition source two years ago. In response, Vercel published a small file called llms.txt at the root of its website, and Vercel Academy now ships an “Add llms.txt” guide that points teams at the same pattern.

A Vercel-style llms.txt looks roughly like this:

# Vercel

> Vercel is the platform for frontend developers. Deploy and host modern web applications with zero-config builds, preview deployments, and a global edge network.

## Get started

- [Quickstart](https://vercel.com/docs/getting-started-with-vercel): Deploy your first project in under a minute.
- [Frameworks](https://vercel.com/docs/frameworks): Next.js, SvelteKit, Nuxt, Astro, and 30+ supported frameworks.

## Core features

- [Deployments](https://vercel.com/docs/deployments): Preview deployments, production deployments, and rollbacks.
- [Domains](https://vercel.com/docs/projects/domains): Custom domains, DNS, and SSL.
- [Edge Network](https://vercel.com/docs/edge-network): Global CDN, caching, and edge functions.

## Compute

- [Functions](https://vercel.com/docs/functions): Serverless and edge functions.
- [Cron jobs](https://vercel.com/docs/cron-jobs): Scheduled invocations of your functions.

## Reference

- [CLI](https://vercel.com/docs/cli): The Vercel command-line tool.
- [REST API](https://vercel.com/docs/rest-api): Programmatic access to projects, deployments, and teams.

llms.txt is essentially a cheat sheet for AI tools. It lists the pages on a site that the brand most wants surfaced and cited when a user asks an AI assistant a question about that brand. The file is written in markdown, which is a simple plain-text format, and includes a one-line description of the site, a short summary, and a list of links to key pages with a brief note explaining what each one covers. Vercel’s version points AI tools to the help pages, setup guides, and developer references that customers most often ask about. The goal is straightforward. When ChatGPT, Claude, or another AI assistant is asked a Vercel question, the model should land on the right pages first.

We covered the format itself, and why a major AI lab adopted it without an official endorsement, in our piece on Anthropic’s llms.txt. Stripe extended the same idea by organizing the file by product line. Vercel is the example that ties the format to a measurable acquisition channel.

Why a File Like This Matters

Modern web pages are full of navigation, ads, popups, and scripts that make it hard for an AI tool to extract the substance of the content. A short markdown file with the key pages and clear descriptions removes that noise and gives the AI a direct path to the content the brand wants cited. There is also a related file called llms-full.txt that takes the idea further by combining the actual content of the listed pages into a single document, which is most useful for documentation-heavy sites where an AI tool can pull the full content in one read.

The file lives at the same kind of address as robots.txt, the older file websites have used for years to communicate with search engines. The difference is that llms.txt is written in language and structure designed for AI assistants rather than search crawlers.

Why This Matters for a Business

The honest answer is that the format is still early. No major AI company has officially confirmed that their systems use these files when reading websites, and adoption is uneven. There is no published click-through study or traffic report that ties llms.txt directly to AI citations.

What is documented is the trend Vercel is responding to. AI assistants are becoming a real referral channel, and the brands that show up first in those answers are the ones with content that is easy for a model to parse and cite. Whether the route to that visibility runs through llms.txt, structured data, clean HTML, or all three is still being sorted out. Shipping the file is a low cost hedge. It takes a developer about an hour, costs nothing to host, and positions a brand to be indexed first if broader support lands.

The same hedge logic now applies in service categories that look nothing like Vercel. We have written about the early dental practice adopters in our dental piece, the med spa version of the same shift, and the local-market gap in Austin.

Where This Fits in an AEO Strategy

Answer Engine Optimization, or AEO, is the practice of structuring a website so it is clear, citable, and machine readable for the AI tools that now generate answers instead of returning a list of blue links. Traditional SEO is concerned with whether your pages rank in search results. AEO is concerned with whether your pages are clear and structured enough to be confidently summarized and cited by AI systems. llms.txt is one of several signals that fit under the AEO umbrella, alongside structured data, entity clarity, and clean information architecture.

The Vercel example is useful because it shows what motivates a serious team to ship the file. The motivation is not proof of impact. It is the recognition that the channel is real, the cost of shipping is low, and brands that move first will be indexed first if broader support lands.

See Where Your Site Stands

If AI assistants are starting to send you traffic and you have no idea which pages are getting cited, get an AEO audit. We will look at your llms.txt readiness, your structured data, your entity signals, and your overall machine readability, then walk you through what to fix first to claim more of the AI referral channel.