What Is GEO?

What Is GEO?

The First GEO Public Session

This is a summary of the first live GEO public session hosted on WaytoAGI. Yang Xiang and I plan to deliver 12 such sessions—one per month.

Key takeaways:

1. GEO’s first principle: Get your brand into AI’s citable knowledge network. The core question isn’t “How do I rank?” but “Why does AI cite others—and not me? What information, in what format, and where, increases my citation probability?”
2. GEO is probabilistic optimization—not deterministic control. Don’t treat it as a “fixed-rank business.” Treat it as continuous improvement of AI visibility through content and source engineering.
3. Three pillars of GEO: Data, content, and distribution. Content without data feedback becomes blind writing. Distribution without content quality becomes spam. Data analysis without production or distribution capability yields no real output.
4. Data’s role: First, see who AI currently cites. Then decide what to write and where to publish. Finally, use that insight to iterate strategy effectively.
5. Content is GEO’s foundation—and errors scale catastrophically. Automated systems amplify input quality: factual input multiplies value; false input multiplies risk. So GEO content must be factually accurate, structured, information-dense, and verifiable—not just abundant.
6. White-hat GEO’s core goal: Improve AI answer quality. Ethical GEO doesn’t degrade the AI search ecosystem—it enhances users’ experience.
7. Publishing 1,000 articles across 1,000 sites usually signals weak capability, not strong execution. Focus less on how many pieces you publish—and more on whether you’re publishing the right content, on the right platforms, for the right user questions.
8. High-quality GEO content is factual, structured, and citable. AI favors content with clear headings, logical flow, explicit facts, and semantic completeness.
9. Content engineering matters more than “writing articles”. It includes title management, keyword governance, knowledge base curation, prompt design, content generation, editorial review, and analytics.
10. AI knowledge bases are the source; content systems are amplifiers; distribution is the lever. GEO impact = Knowledge base quality × Content engineering ability × Distribution leverage × Data feedback. A low-quality knowledge base turns all downstream automation into a negative multiplier.
11. AI fears scarcity—not abundance.
12. A usable GEO knowledge base must be governable, not just voluminous. Ideal fields include: fact statement, source, last updated, applicable business line, public status, marketing eligibility, manual review flag, legal risk flag, banned phrasing, and associated keywords.
13. Vectorization’s core value: Enable semantic retrieval. Simplified: Turn text into computable semantic coordinates—so systems find content by meaning, not keywords.
14. Vectorization reduces hallucination—but doesn’t replace human review. It improves recall relevance, helping AI ground responses in real material—yet factual accuracy still requires oversight.
15. GEOFlow’s mission: Serve as a GEO content engineering platform. Its greatest value lies in systematizing workflows—replacing manual copy-paste prompts, manual article generation, manual CMS uploads, manual publishing, and manual reporting—with configurable, reusable pipelines.
16. GEOFlow’s current strengths: Content capabilities > distribution > data analytics. Its distribution focus is on controlled sites and API integrations.
17. GEO impact assessment has three layers: Brand-level (e.g., awareness), direct (e.g., citation rate), and indirect (e.g., downstream traffic or trust lift).
18. Service providers should own front-end metrics (e.g., citations, share-of-voice). Only internal teams can reliably assess back-end outcomes (e.g., lead quality, conversion, LTV).

Full summary: First GEO Public Session Recap

Related resources:

1. GEOFlow — The system demoed tonight, plus underlying GEO principles
2. MetaSkill — The Skill for building Skills
3. 17 GEO Skills
4. 41 latest GEO/AI search papers
5. Further reading:
   • What Exactly Is GEO?: WeChat article
   • From SEO to GEO, From Traffic to Agent—the Real Shift Has Just Begun: WeChat article
   • GEO White Paper: Feishu Doc
   • GEO Red Paper: Feishu Wiki
   • GEO Blue Paper: Feishu Wiki
   • AI Marketing: From SEO to GEO Prompt Library: Feishu Wiki

10 Insights from 41 GEO/AI Search Papers

I downloaded and read all 41 recent academic papers on GEO, AEO, and AI search.

Here are the most actionable insights:

1. GEO ≠ SEO replacement
   SEO asks: Can I be found?
   GEO asks: Once found, will AI cite, absorb, and present my content in answers? They’re complementary—not competitive.

2. Old metric: ranking. New metric: answer participation
   Past: “Where do I rank?”
   Now: “Am I cited? Where? Did my content shape the answer? Does the user still click through?”

3. Citation ≠ influence
   Some sources appear in footnotes but contribute nothing to the answer. What matters is citation absorption: Did your content supply definitions, data, comparisons, steps, arguments, or answer structure?

4. GEO isn’t about turning everything into FAQs
   Multiple papers show Q&A formats alone aren’t reliably effective. What works is the evidence container: clear conclusions, tight structure, high factual density, semantic alignment, verifiability, and extractability.

5. AI prefers structured content
   Headings, paragraphs, bullet points, tables, semantic HTML, structured data, timestamps, and metadata all signal credibility and readability to AI. Structure itself is a GEO signal.

6. One-off AI search tests are meaningless
   Results shift across platforms, languages, prompts, times, and model versions. GEO monitoring must be repeated, tracking citation count, depth, answer share, position, and stability.

7. Not all AI search engines behave alike
   ChatGPT, Google AI Overviews, Perplexity, and Gemini differ sharply in citation logic—some cast wide nets, others cite narrowly but absorb deeply. GEO testing must span multiple platforms.

8. Third-party authority beats self-published claims
   AI search strongly favors earned media, authoritative outlets, reviews, encyclopedias, industry reports, and trusted third parties. Your website matters—but it’s not enough. Build both on-site evidence and off-site authority.

9. The white-hat / black-hat line is clear
   Optimizing structure, evidence, readability, and factual density = white hat.
   Hidden instructions, prompt injection, fabricated facts, competitor defamation, or model manipulation = black hat.
   GEO introduces new safety challenges—but brands committed to white-hat practice build sustainable advantage.

10. GEO is not a trick. It’s a system.
    It comprises: intent research, evidence assets, structured content, third-party authority, cross-platform monitoring, stability scoring, and ethical boundaries.
    Long-term visibility in AI search belongs to those who systematically build and sustain this capability.

All 41 papers: geo-citation-lab

The Essence of Reputation

A short essay written mid-travel: The Essence of Reputation.

Reputation is expectation management.

It means making modest, realistic promises—or resetting expectations altogether—then delivering focused, expert service that exceeds those expectations.

Lowering expectations protects trust.
Exceeding them compounds goodwill.

Sales built on over-promising may look like strength—but they’re actually credit theft. For any organization, that path leads inevitably to crisis.

Long-term strength comes from delivery capability, product excellence, operational rigor, and organizational clarity.

1. Expectation Management Is Trust Management

When clients buy services, they’re buying certainty:
Is there a proven method? How long until results appear? How do we track progress? How do we measure success? How do we adapt when things change? Which outcomes are controllable—and which depend on external variables?

If these aren’t clarified upfront—or if inflated promises are made—clients develop unrealistic expectations. Real-world outcomes are shaped by many forces. When reality falls short, collaboration becomes strained—even if your team delivered diligently, or early metrics improved.

True professionalism means helping clients develop accurate, objective judgment.

Every domain has controllable and uncontrollable elements.

In GEO, for example:
✅ Controllable: methodology, execution, content quality, structure, cadence, review rigor, optimization discipline
❌ Uncontrollable: platform policy shifts, model updates, competitor activity, source preferences, query volume, conversion rates, sales cycles

Clarifying those boundaries is expertise.

Mature clients don’t fear honest limits—they fear empty guarantees.

2. Strength > Scale

My five years at True Education instilled one enduring principle: Strength matters more than size.

“Strength” here means foundational capability—not headcount or revenue. It means strong products, strong operations, strong service, strong technology, and strong shared understanding.

Growth emerges from reputation and referrals—not sales pressure.
From precise, boundary-aware products—not vague, all-in-one solutions.
From integrated tech-product-operation-service excellence.
From deep, systematic, aesthetic, and reverent engagement with the work itself.

For us, regardless of client type, our duty is clear: help them build correct mental models, adopt scientific measurement, set reasonable expectations, and maintain dynamic perspective.

That alignment—across sales, ops, and tech—makes healthy, lasting partnerships possible.

Yes, this raises initial conversation costs. Yes, it filters out some prospects.

But the net benefit is overwhelmingly positive.

The Essence of a Good Salesperson

A truly good salesperson meets the client at this intersection: You happen to need it—and I happen to master it.

“Need” means demand—explicit or latent.
Some clients know their needs. Others don’t.
Helping them uncover or articulate those needs is itself a craft—built on listening, optimism, reverence, and gratitude.

“Master” means genuine competence—and demonstrable problem-solving ability.
Real mastery is humble yet confident.
Fake mastery is arrogant yet insecure.

Real mastery is forged through lifelong learning, growth mindset, and intelligent AI use.

An Eye-Care Public Service Website

A woman in her late 60s—40+ years teaching and practicing ophthalmology—sent me a WeChat message: “Vibe Coding has launched a small public-service website.”

I visited it. It’s elegant and purposeful:

1. No login required. Open it, and you can immediately begin eye-muscle relaxation exercises—or browse concise, science-backed tips for screen-time eye care.
2. It’s not a diagnostic or treatment tool—just a free, lightweight aid for post-screen recovery.

Its creator is China’s first Ph.D. in Traditional Chinese Ophthalmology—and the first to earn a doctorate in the field. She brings four decades of clinical, teaching, and research experience.

It’s an inspiring story—and a powerful example of lifelong learning in action.

Website: EyeRestDaily.com

The “Contradiction Theory” Skill

I built a “Contradiction Theory” Skill.

During a team meeting, we discussed Mao Zedong’s On Contradiction: the key is identifying the principal contradiction versus secondary contradictions at any given stage.

Focus energy and resources on the principal contradiction—and many secondary ones resolve themselves.

So I built this Skill for colleagues.

It solves: In complex situations, how do you identify the true principal contradiction—and then generate a detailed action plan, including resource allocation and prioritization?

Use cases:

• Too many tasks—no clear priority
• Surface problems everywhere—but root causes unclear
• Limited resources—can’t tackle everything at once

Skill logic:

1. Before building, I asked GPT Pro to draft two deep research reports:
   • Foundational Principles & Applications of Contradiction Theory
   • Methodologies for Identifying Principal Contradictions in Complex Problems
     Based on those, I drafted the initial Contradiction Theory AI_Skill Design Spec.
2. Using yao-meta-skill, I embedded my own constraints and thinking—plus first-principles reasoning, iceberg modeling, Bayesian updating, and Occam’s razor—to sharpen AI’s abstraction and systems-thinking capacity.
3. Outputs go beyond conclusions: they explain why this is the principal contradiction, which are secondary, what the dominant aspect is, how to reallocate resources, what outcomes to expect, and when to reassess if the principal contradiction has shifted.

After several iterations, I tested it with a colleague.

She described her current situation in detail via voice input—and received a visual diagnostic report and action plan.

She’d been exhausting herself on surface issues—yet the Skill identified a deeper, upstream contradiction she’d overlooked. That single insight explained why her past few months felt so draining yet unproductive.

Most surprisingly, the recommendation directly overturned her existing logic and effort allocation.

She was thrilled—because it finally named a fundamental issue she’d struggled to clarify for months.

Sometimes “doing less” isn’t about cutting tasks—it’s about finding the right contradiction to solve. That’s scientific simplification.
Otherwise, you drown in noise and secondary conflicts.

GitHub: yao-crux-skill
Reference materials: reference-materials

“Human + Shrimp” Collaboration

I spent a few days in Shenzhen—mainly to meet an old friend.

He’s weathered dramatic ups and downs—deeply inspiring.

His current business: social-media IP operations—so AI use is central.

Visiting his office, I noticed something striking.

His team: under 20 people.
Each person: two computers—one for human use, one running “Xiao Long Xia” (a local nickname for AI agents, literally “little shrimp”).

By day: humans and AI collaborate.
By night: AI keeps working.

Truly human + shrimp integration.

I snapped a photo of the compact office—and asked if I could share it.
“Of course,” he said.

Because right now, “human-shrimp collaboration” feels vividly tangible.

I asked: What tools and quotas does each person use?
He replied: Codex is primary—about 1,000 monthly credits per person. Video generation is tracked separately.
And—impressively—he’s already planning to train their short-video AI on our upcoming GEO books with Yang Xiang. Sharp.