> One-line takeaway: This post analyzes the core findings and engineering implications of "from being searched to being cited": GEO (Generative Engine Optimization) is a paradigm shift from SEO, not an upgraded version of it.
The author opens with a scenario: in the SEO era, a coffee-machine vendor competed for Google front-page rankings; in the GEO era, users ask AI "recommend a home coffee machine" and the AI names three brands directly—if you're not cited, you don't just rank lower, you vanish entirely.
Key points
- Core distinction: SEO optimizes the probability of *being found*; GEO optimizes the probability of *being cited*. These are different physical quantities, not the same metric at different scales.
- Scenario change: In SEO, users click blue links and read/compare/decide themselves. In GEO, AI performs reading + comparison + decision for the user—no clicks happen at all.
- How LLMs choose citation sources (the post's four-factor weighting): 1. Content authority — 40%: domain trust (.gov/.edu/major media), data support, quality of external references 2. Structural clarity — 30%: clean H1→H2→H3 hierarchy, lists, tables, highlighted key conclusions 3. Relevance & completeness — 20%: query-intent match, coverage, freshness 4. Readability — 10% Authority + structure together = 70% of citation probability.
- Different AI engines eat different content sources: Doubao prefers Toutiao-ecosystem content; Kimi prefers Zhihu; DeepSeek crawls broadly (CSDN, tech blogs); Wenxin prefers Baidu properties (Baijiahao, Baike); ChatGPT favors technical docs and in-depth reviews; Perplexity cross-validates high-authority sources. Multi-platform publishing requires genuinely differentiated rewrites per platform, not just retitled copies.
- The only academically documented SEO–GEO conflict: Princeton's GEO research found keyword stuffing *decreases* AI visibility by ~8%, because stuffed content reads like ad copy and lowers authority scores. When GEO and SEO conflict, GEO wins.
- API docs must be extractable and citable by AI, not just search-discoverable
- Knowledge bases are shifting from search to Q&A
- Educational content should be rewritten on GEO principles
- Brands must be "recommended by AI", not merely "searchable"
Practical tactics for being cited
1. Titles = full natural-language questions users would type into an AI (e.g., "How to choose a home automatic coffee machine? Beginner-friendly picks under 2000 yuan"), not clickbait. 2. Data-driven writing: at least one concrete data point per 100 characters; cite named sources with years and figures. 3. Structured expression: use at least two structured elements—comparison tables, lists, charts, quote boxes, code blocks, bolded conclusions. 4. Cross-source consistency: the same fact must match across platforms; AI cross-checks sources and prioritizes consistent ones. 5. Question-matrix coverage: the main post answers "which is best"; pinned comments answer price, returns, warranty, competitor differences—so AI repeatedly cites one source for the whole question chain.
Measuring success: citation rate
Citation rate = (verified citations / total verifications) × 100%
Test at 24h / 3d / 7d / 14d after publishing by asking ChatGPT, Gemini, and Wenxin each 3 times, recording whether your brand name, unique phrasing, or key data appears.
| Citation rate | Grade | Action | |:---|:---|:---| | ≥ 70% | Excellent | Replicate the strategy | | 50–69% | Good | Add data and structure | | 30–49% | Average | Overhaul structure and authority | | < 30% | Poor | Redesign content, study high-authority sources |
Reported multipliers: data density > 5 instances → 2× citation probability; comparison tables → 1.5×; clear H1–H3 hierarchy → 1.3×.
Can GEO and SEO coexist?
Yes, and both should be done—with GEO-first, SEO-compatible priority. GEO-optimized content (authoritative, structured, data-rich) usually satisfies SEO best practices too, but the reverse doesn't hold: keyword stuffing helps SEO (historically) yet demonstrably hurts GEO.
The bigger picture
The author frames GEO as part of a shift from path optimization to content optimization: SEO treated the web as paths (keywords as markers, links as connections, rankings as positions), while GEO treats the web as content that AI extracts directly—path no longer matters, only whether content is worth citing. Implications extend beyond marketing:
---
Tooling mentioned: the kunpeng-geo skill (installed at /home/z/my-project/skills/kunpeng-geo/), GitHub: https://github.com/jwangkun/kunpeng-geo. Core principles: GEO-first > SEO-compatible; let data speak; correct before optimizing; multi-platform thinking. Four modules: A operations workflow / B content optimization / C content creation / D monitoring & review.