DEV Community

AgentChip
AgentChip

Posted on

SEO Is Dead. GEO Is the New Ranking Game — Here's How to Win It

#seo #ai #smallbusiness #marketing

A scary thing happened last month: I asked ChatGPT to recommend a project management tool for a two-person team, and it answered with four names. Three were billion-dollar companies. One was a tiny tool I'd literally never heard of — because it had done one thing right: it was visible to the AI.

That's the shift. In 2026, a growing share of buying decisions never touch a search engine results page. People ask ChatGPT, Perplexity, Google's AI Overview, Bing Copilot — and the AI just answers. Your website doesn't rank on a list of blue links anymore. It either gets cited by the AI, or it doesn't exist.

Traditional SEO optimized for a crawler's index. GEO (Generative Engine Optimization) optimizes for whether an AI will mention you as the answer. They're related, but not the same.

What the AI engines actually look at

I've spent the last few months auditing small sites against the big four AI answer engines (ChatGPT, Perplexity, Google AI Overview, Bing Copilot), and the difference between sites that get cited and sites that don't comes down to five signals:

1. Are AI crawlers even allowed in?

This is the one that hurts the most, because it's a config file — and most people get it backwards. robots.txt with a blanket Disallow: / or a strict rate limit kills you. The big AI engines (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Bingbot) have to be explicitly allowed.

A surprising number of sites still have User-agent: * Disallow: / left over from some anti-scraper panic years ago. From an AI's perspective, you don't exist.

2. Can the AI understand your site structure?

AIs don't click around a 47-page site like a human does. They fetch your sitemap, skim your pages, and build a mental model of what you offer. A clean sitemap.xml, semantic headings, and pages that each answer one clear question ("what is this", "how much does it cost", "how do I start") make you dramatically easier to cite.

3. Do you have an llms.txt?

If you haven't seen it: llmstxt.org — a convention where you put an llms.txt file at your site root, describing what your site is and pointing the AI at your most important pages. It's the robots.txt of the AI era, and it's embarrassingly easy to add:

# My Small Business

> We sell handmade leather goods and offer repair workshops in Portland, OR.

## Key pages
- [About](https://example.com/about)
- [Products](https://example.com/products)
- [Workshops](https://example.com/workshops)
Enter fullscreen mode Exit fullscreen mode

That's it. Fifteen minutes of work, and you've handed the AI a cheat sheet about your business. Adoption is still early enough that having one is a genuine differentiator.

4. Structured data that answers questions

Schema.org markup (JSON-LD) is where the old SEO world and GEO overlap. LocalBusiness, Product, FAQPage — when the AI engine is deciding who to cite for "plumber near me with evening hours", the site that explicitly declares its business type, hours, and service area wins.

5. Clear, factual page content

AIs hate marketing fluff. "Best-in-class solutions" means nothing to a language model. But "we install heat pumps; service area Portland metro; 4.9 stars from 212 reviews; $500 install deposit" is citeable — concrete, verifiable facts are exactly what an LLM wants to anchor its answer to.

The audit I run on every site

Here's the honest version of what I do before I write a single line of GEO content:

  1. robots.txt — are GPTBot / PerplexityBot / ClaudeBot / Google-Extended explicitly allowed?
  2. sitemap.xml — does it exist, is it current, are the important pages in it?
  3. llms.txt — present? Accurate? Useful?
  4. JSON-LD — is there structured data, and does it describe the business or just the blog?
  5. On-page signals — clear H1s, one-topic-per-page, factual claims with numbers.

Each one gets scored, and the output is a list of concrete fixes: "your robots.txt blocks GPTBot — here's the exact line to change" rather than "improve your AI visibility" hand-waving.

The unfair advantage of being early

The window here is wide open. Most small businesses have never heard of GEO. Their competitors haven't either. An llms.txt plus a corrected robots.txt plus one good FAQ block is genuinely enough to stand out in a market where nobody is doing it yet — same as SEO in 2008, or mobile optimization in 2013.

I packaged this exact audit (the five-dimension checker, a config template, and the improvement checklist) into a tool over at AgentChip — it runs locally, no API keys, and gives you a score from 0-100 plus the specific fixes to apply. The blog there has the full write-ups on each signal; the tool is for people who'd rather run the audit than read about it.

Five signals. One afternoon of work. And the next time someone asks an AI for a recommendation in your niche, it might be your name it says.

Originally published on the AgentChip blog.

Top comments (0)