Spark

ChatGPT Has Never Heard of You: The New SEO War Has Already Started

Traditional SEO isn't enough anymore. Discover how artificial intelligence is rewriting the rules of search, and what llms.txt is, the standard your site needs today.

You spent years optimising your website for Google. Metadata, keywords, backlinks, load speed. All in order. You’re on page one. But today, when a potential customer asks ChatGPT: “What’s the best web development agency in Uruguay?”, your business doesn’t exist. It doesn’t come up. Not even as a second choice.

Welcome to the new battlefield of digital search.

The SEO you knew is mutating

This isn’t a gradual shift, it’s a break. According to Semrush data published in 2025, traffic from traditional Google search fell 18% in sectors where generative AI started answering questions directly. Millions of people who used to click through ten results now get a direct answer and never visit a single site.

As Google CEO Sundar Pichai put it: “AI is not just another feature, it is a fundamental reshaping of how people access information.” And he’s right, even if he perhaps didn’t picture that reshaping eroding his own business model.

The numbers are stark:

  • ChatGPT passed 200 million weekly active users in early 2026
  • Perplexity AI grew its user base 300% in 2025, positioning itself as an alternative search engine
  • Google AI Overviews now answers 40% of informational searches directly, with no click on any result
  • In Latin America, use of AI assistants to resolve commercial queries grew 85% during 2025

Keyword and backlink SEO still matters to Google, but there’s an entire ecosystem of traffic and discovery that your current strategy is ignoring completely.

What Is AEO? (Answer Engine Optimization)

If SEO was the art of showing up in Google, AEO is the art of being cited by AI.

AEO (Answer Engine Optimization) is the emerging discipline of optimising your content and web presence so large language models — ChatGPT, Gemini, Claude, Perplexity, Copilot — can find you, understand you, and recommend you when someone asks a question related to what you offer.

The difference runs deep. SEO works with algorithms that rank pages. AEO works with models that generate answers. And to generate a good answer, AI needs to:

  1. Understand exactly what your business does (not just parse your HTML full of menus and scripts)
  2. Tell you apart from others with the same name (how many “Spark”s are there in the world?)
  3. Access your relevant information efficiently, without wasting the limited space in its context window

As OpenAI CEO Sam Altman put it: “Language models don’t just look up information, they synthesise it. If your information isn’t structured to be synthesised, it simply won’t show up.”

llms.txt: The robots.txt of Artificial Intelligence

In 1994, webmasters adopted robots.txt to tell Google’s crawlers what they could and couldn’t index. It was the first protocol between the humans building the web and the machines consuming it.

In 2024, Jeremy Howard — co-founder of fast.ai and a well-known AI researcher — proposed a new standard: llms.txt. The logic is elegant: you place a Markdown file at the root of your site (yourdomain.com/llms.txt) with a structured description of who you are, what you do, and links to your most relevant information.

The difference from robots.txt is philosophical. robots.txt tells bots what they should not read. llms.txt tells AI what it should read, and how it should understand you.

A basic llms.txt looks like this:

# My Company - Web Development Agency (mycompany.com)

> We are an agency specialising in e-commerce, landing pages
> and custom software for businesses in Uruguay and Argentina.

## Services

- [E-commerce](https://mycompany.com/services): Online stores with...
- [Landing Pages](https://mycompany.com/services): Conversion-focused pages...

## Contact

- Email: [email protected] | WhatsApp: +598 99 000 000

It looks simple. It is. But the impact is significant: instead of AI trying to infer what your company does by reading 50 HTML pages full of code, navigation, ads and scripts, it reads one clear, concise, noise-free text document. Fewer tokens, more real context.

Alongside llms.txt, there’s also llms-full.txt: an expanded version that includes the full content of every important page on your site, so an AI can pull all the information it needs in a single request.

Why Your Name Might Be Working Against You

Here’s a problem most businesses don’t even suspect they have: AI models are polluted with ambiguity.

Imagine your company is called “Spark”. Sounds good, right? Modern, energetic. But when ChatGPT hears “Spark”, it thinks of dozens of things at once: Apache Spark (the big-data framework millions of engineers use), Spark AR (Meta’s augmented reality platform), Spark Capital (the venture capital firm), a rock band, a dating app, and a few thousand other companies around the world.

Without a clear disambiguation signal, AI can’t know which one you mean. And when in doubt, it doesn’t mention you.

llms.txt solves this elegantly. It lets you explicitly tell any AI system: “Spark (spark.uy) is a web development agency in Montevideo, Uruguay. We have nothing to do with Apache Spark or Spark AR.”

That’s not overkill. It’s exactly the information a language model needs to tell you apart correctly.

The shift has only just started. Andrej Karpathy, former AI director at Tesla and one of the field’s most influential researchers, described LLMs as “the new internet browsers, except instead of rendering HTML, they understand and synthesise information.” And if LLMs are the new browsers, then llms.txt is the new HTML: the language you use to speak to this new layer of the web.

Trends taking hold in 2026:

  • Google AI Overviews are gradually replacing traditional snippets. Answers are generated with AI right on the results page.
  • Perplexity and ChatGPT Search already work as full search engines with real-time web access.
  • Microsoft Copilot, built into Windows and Edge, answers commercial queries without opening a browser.
  • Voice assistants are evolving: instead of listing results, they answer directly and cite sources.

Businesses that implement AEO strategies today have a large window of competitive advantage. In 5 years, optimising for AI will be as basic as having a website. But today, it’s still mostly unexplored territory.

As Jeff Bezos put it: “Your margin is my opportunity.” In this context: your competitors’ inaction on AEO is your opportunity.

How to Prepare Your Site Today

You don’t need to reinvent your digital strategy from scratch. You need to add a layer of AI readability on top of what you already have. Here are the concrete steps:

1. Create your llms.txt A Markdown file at the root of your domain with your full name, description, main services, and contact details. Simple, plain text, no HTML or scripts.

2. Consider an llms-full.txt If you have a lot of content (detailed services, FAQs, portfolio, blog), a full-context file lets AI take in your whole business in a single read.

3. Structure your data with Schema.org JSON-LD markup on your pages gives AI crawlers semantic context. Organisation, services, products, pricing: all explicitly marked up.

4. Write for humans, not for bots LLMs are exceptionally good at spotting quality content. A clear, well-argued, genuinely useful article has a much better chance of being cited than a page stuffed with keywords.

5. Always pair your name with your domain In your llms.txt and across your content, write your company name next to its domain: “Spark (spark.uy)” rather than a bare “Spark”. If your brand shares a name with something better known, the domain is what lets AI tell you apart. Explicit disambiguation is the key.

6. Allow AI crawlers in your robots.txt GPTBot, Claude-Web, PerplexityBot, anthropic-ai: make sure your robots.txt allows these agents, which many older sites accidentally block by default.

7. Keep content fresh Models get updated. Fresh, dated content is more likely to be included in training data and real-time retrieval.

SEO Didn’t Die, It Evolved

Don’t get this twisted: Google is still the most important traffic channel for most businesses, and traditional SEO is still necessary. But the digital landscape now has a second dimension that didn’t exist before: the AI dimension.

Your digital presence is no longer measured only in Google rankings. It’s measured in whether you exist inside the knowledge of the models answering millions of queries a day. In whether, when someone asks Claude “which agency should I hire for my online store in Uruguay”, your name comes up, with the right description.

As Bill Gates wrote in his famous 1995 memo “The Internet Tidal Wave”: “Those who are slow to react will pay the price.” Today, the wave is AI. And most Latin American businesses still aren’t on the board.

At Spark, we’re already ahead of this wave

At Spark, we implement AEO strategies and llms.txt on every web project we build. Our sites are designed to be understood by Google, and by ChatGPT, Gemini, Claude and whatever AI system comes next.

If you want your business to exist on the new web — the web that AI reads — let’s talk. We audit your current presence, implement the files you need, and build a content strategy that works for humans and machines alike.

Because in 2026, if AI doesn’t know you exist, for half your potential customers, you don’t.