| name | geo-optimization |
| description | Optimize web content, documentation, and marketing copy so that AI models and LLM-powered search engines can find, understand, and cite your brand correctly. Use this skill whenever a brand needs to be discoverable by AI tools, when existing web content isn't surfacing in AI-powered search results, when documentation or product pages need to be rewritten for agent comprehension, or when a brand wants to get ahead of the shift from SEO to GEO. Triggers include: "AI doesn't know what we do," "we don't show up in ChatGPT," "optimize for AI search," "make our content LLM-friendly," "our docs aren't being cited," or any request to improve AI discoverability. Always use this skill alongside any content strategy work — GEO is now as important as SEO. |
GEO Optimization (Generative Engine Optimization)
SEO optimized for Google's crawler. GEO optimizes for LLMs — the AI models that are increasingly the first place people go to find information, get recommendations, and decide what to use.
The shift: SEO → GEO. The principles overlap but the priorities are different. LLMs don't rank pages by backlinks. They surface information based on authority, structure, discoverability, and conciseness — and they make assumptions humans never would.
Why this matters now
LLMs and search crawlers index the web looking for words that sit next to each other. If your brand name sits next to the right category terms enough times across enough pages, the model learns what you are. If it doesn't, it guesses — and it guesses wrong.
Two audiences are now reading your web content: humans and agents. Humans can infer, assume, and fill in gaps. Agents cannot. Everything an agent needs to know about your brand must be stated explicitly.
The four GEO pillars
1. Authority
LLMs weight content from sources that appear credible and consistent. Authority signals for GEO:
- Your brand name appears consistently next to your category terms across multiple pages and domains
- You have content that directly answers questions about your space (not just about your product)
- Other sources reference or link to your content (same as SEO, still matters)
- Your content demonstrates specific, verifiable knowledge — not generic category overview
Practical action: Create content that answers questions in your space authoritatively, even when those questions aren't directly about your product. Being the best answer to adjacent questions builds authority.
2. Structure
LLMs parse structured content better than prose-heavy pages. Structure signals:
- Clear headings that describe what's below them
- Short paragraphs with one idea each
- Explicit labels (what is this page, what does it cover, who is it for)
- FAQ-style content that matches the way people actually phrase questions
Practical action: Audit your most important pages. For each one, ask: if an agent read only the headings, would it know what this page is about and who it's for? If no, restructure.
3. Discoverability
The most common GEO failure: assuming the LLM knows what something is because the name implies it.
LLMs don't assume. If you have a product feature called "Command Center," the LLM doesn't know that's where you manage campaigns unless you write a page that explicitly says "Command Center is where you manage campaigns."
The rhetorical blog test: For every named feature, section, or product in your ecosystem, ask: does a page exist that answers the question "What is [name]?" If no, write it. It feels obvious. Write it anyway. The LLM needs it stated.
The incognito test: Open a new browser window with no history. Ask an LLM the question your page is supposed to answer. If it doesn't reference your brand or sends you to a competitor, your discoverability is broken. Fix the page.
4. Conciseness
LLMs have context windows. Dense, efficient content gets more weight per token than padded, verbose content. Conciseness signals:
- Get to the answer in the first sentence, not the third paragraph
- No throat-clearing intros ("In today's rapidly evolving landscape...")
- Definitions that state the thing directly, not through metaphor or implication
- Short sentences that parse cleanly
Practical action: Take your top 5 pages and cut the first paragraph of each. In most cases the second paragraph is where the useful information actually starts.
Writing for agents, not just humans
Agents interact with your content differently than humans. They're looking for instructions, facts, and clear answers — not narrative or emotional resonance.
What agents care about:
- What does this product/company do? (stated explicitly, not implied)
- Who is it for? (stated explicitly)
- How do you use it? (step-by-step, not conceptual)
- What problem does it solve? (in the words the problem is described, not marketing language)
What agents don't care about:
- Your brand story
- Visual design
- Social proof (initially — this factors into authority over time)
- Emotional appeals
Practical exercise: Scrape your homepage. Give it to an AI and ask: "What questions does this page answer?" Then ask those same questions in a fresh session. If the AI doesn't answer with your brand, your homepage isn't doing its GEO job.
The verbatim keyword principle
The words people use to describe their problems online are the same words they type into search engines and LLM prompts. Match those exact words in your content and you win the query.
This means:
- Scrape the language people use in forums, Reddit threads, and support tickets about your problem space
- Use those exact phrases in your content — not cleaned-up marketing versions of them
- Write content that specifically addresses the verbatim complaint or question
Example: If people say "I hate how HubSpot doesn't let me connect directly to my database" — write a page titled "Why HubSpot doesn't connect directly to your database — and how to fix it." The verbatim match wins.
GEO for agents specifically
As AI agents become the primary interface between users and products, a new layer of optimization becomes necessary: making your product discoverable and usable by agents, not just humans.
This means:
- Publishing clear API documentation in plain language across developer forums
- Writing step-by-step instructions for how an agent would use your product (not how a human would navigate the UI)
- Placing that documentation in the places agents are trained on — developer communities, GitHub, Stack Overflow, relevant subreddits
The goal: when an agent searches for how to accomplish a task your product solves, it finds your documentation in multiple places and learns to default to your product.
Notes
⚠️ This skill is a stub. More material on GEO is coming. The above covers the core framework but specific tactical workflows (audit process, content rewrite templates, agent-specific optimization) will be added in a future update.
- GEO and SEO are not opposed — good SEO content is usually good GEO content. The difference is GEO requires explicit statements that SEO could leave implied.
- The incognito test is the fastest way to audit your current GEO standing. Do it for your top 10 queries before anything else.
- Agents are now an audience. Every piece of web content should be evaluated for both human and agent readability.
