Full-Process GEO Optimization Case Study — A Brand's Complete Zero-to-One Journey
The previous article gave the "roadmap," this article gives the "real case" —
How a CRM brand went from zero GEO foundation to being stably cited by AI in 3 months.
Brand background:
- Brand name: QiKeTong (pseudonym)
- Business: SME CRM systems
- Team: 3 people (1 marketing manager + 1 content specialist + 1 part-time developer)
- Starting point: Never mentioned in AI search, official website DAU of 200
This case study isn't fabricated — it synthesizes data and processes from several real GEO projects.
If your situation is similar, follow this process.
I. Pre-Optimization: Week 0 (Baseline Report)
AI Search Current State
- Searching "SME CRM recommendation" on ChatGPT → AI's 3 recommended brands don't include QiKeTong
- Searching "How to choose a CRM system" on Doubao → AI gives generic selection criteria, no brand listed
- Searching "QiKeTong CRM" on Perplexity → AI returns: "I couldn't find detailed information about 'QiKeTong'"
Website Current State
- robots.txt: Configured with
Disallow: /— blocking all crawlers - Schema: No structured data
- Content: Only product pages (3 pages) and 1 blog post ("QiKeTong CRM Feature Introduction")
- Encyclopedia entry: None
- Zhihu: No account
Competitor Current State
- Competitor A: Complete encyclopedia entry, FAQ on website, answered 30+ questions on Zhihu, occasionally cited by AI
- Competitor B: Industry white paper publisher, republished by multiple industry media, cited by AI on "industry trends" topics
- Competitor C: Rich website content, case library, cited by AI on "CRM selection" topics
II. Phase 1 (Weeks 1-2): Foundation Repair
Action 1: Fix Website Crawlability
Problem: robots.txt's Disallow: / blocked all AI crawlers.
Solution:
User-agent: GPTBot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: PerplexityBot
Allow: /
Result: AI crawlers can now access the website.
Action 2: Deploy Structured Data
- Deploy Organization Schema on official website
- Deploy Product Schema on 3 product pages
- Confirm Schema passes validation
Action 3: Create LLMs.txt
Create llms.txt in website root directory, containing brand name, one-line positioning, core products, customer scale, and other key information.
Action 4: Establish GEO Baseline
Record current data:
- AI referral rate: 0%
- Citation share: 0%
- Brand description accuracy: N/A (AI doesn't know the brand)
- Competitor citation shares: Competitor A 8%, Competitor B 12%, Competitor C 15%
Baseline Summary
End of Week 2. The website is now "readable" to AI crawlers — but there's no content for AI to "cite" yet.
III. Phase 2 (Weeks 3-6): Content Building
Week 3: Identify Core Topics and Create Content
Core topics (prioritized):
- "SME CRM recommendation"
- "How to choose a CRM system"
- "Difference between CRM and ERP"
- "Do small companies need CRM"
- "CRM system pricing"
This week's output (5 pieces):
- Article 1: "2026 Top 5 SME CRM System Recommendations" (including QiKeTong)
- Article 2: "CRM Selection Guide: From Needs Analysis to Product Trial"
- Article 3: "CRM vs ERP — Everything You Need to Know in One Table"
- Article 4: "Do Small Companies Really Need CRM? Signs It's Time to Move from Excel"
- Article 5: "2026 CRM System Pricing Comparison"
Week 4: FAQ and Q&A Pairs
FAQ page output: Create FAQ page with 15 Q&A pairs covering users' most common questions. Deploy FAQPage Schema.
Zhihu publishing: Register Zhihu enterprise account, answer 5 CRM-related questions.
Week 5: Deep Content
This week's output (2 pieces):
- Article 6: "How CRM Boosts Sales Team Efficiency? Real Results from Data" (2000-word in-depth article with data + cases)
- Article 7: "7 Common Mistakes SMEs Make When Implementing CRM" (2500 words, with real-world cases)
Week 6: Content Optimization Review
- Conduct "AI-friendliness check" on all 7 published content pieces
- Add "article summary" (first 100 words as core conclusion) to each article
- Add "related articles" internal links to each article
IV. Phase 3 (Weeks 7-12): Growth and Results
Weeks 7-8: First Results Appear
First monthly retest (end of Week 8):
Searching "SME CRM recommendation" on ChatGPT:
- QiKeTong's name appeared in the AI answer, ranked 4th in the recommendation list
- Citation source: The company's Zhihu answer
Searching "How to choose a CRM system" on Doubao:
- AI didn't directly recommend brands, but in the "selection criteria" section, one of the citation sources linked to the company's FAQ page
Data:
| Metric | Baseline (Week 0) | Week 8 |
|---|---|---|
| AI referral rate | 0% | 2% (on "CRM recommendation" topic) |
| Citation share | 0% | 3% |
| Brand description accuracy | N/A | 60% (AI knows "QiKeTong is a CRM brand") |
Weeks 9-10: Optimization Iteration
Issues discovered:
- AI's description is "QiKeTong is a CRM company" — doesn't mention "suitable for SMEs" positioning
- On the "CRM pricing" topic, AI didn't cite QiKeTong's content
Optimization actions:
- Update LLMs.txt to clearly state "QiKeTong specializes in SME CRM"
- Publish new content: "Complete SME CRM System Pricing Guide — From Free to Custom"
- Add pricing-related Q&As to FAQ page
- Answer "How much does SME CRM cost" on Zhihu
Weeks 11-12: Scaled Results
Second monthly retest (end of Week 12):
Searching "SME CRM recommendation" on ChatGPT:
- QiKeTong appeared at position 3 in the recommendation list
- AI described it as "QiKeTong specializes in SMEs with good value for money"
Searching "How to choose a CRM system" across multiple platforms:
- In the "selection criteria" citation sources, QiKeTong's FAQ page was cited
Final data comparison:
| Metric | Baseline (Week 0) | Week 8 | Week 12 |
|---|---|---|---|
| AI referral rate | 0% | 2% | 8% |
| Citation share (core topics) | 0% | 3% | 11% |
| Brand description accuracy | N/A | 60% | 85% |
| Zhihu platform citation count | 0 | 3 | 15 |
| Official website organic traffic change | 200/day | 250/day | 420/day |
V. Case Review: What Worked?
Top 3 Most Effective Actions
TOP1: FAQ Page + FAQPage Schema deployment.
The FAQ page was the most frequently cited content by AI. Questions AI needed to answer like "How to choose a CRM system" and "What's the difference between CRM and ERP" were all covered by the FAQ page.
TOP2: Zhihu matrix.
Zhihu content has extremely high citation rates in Chinese AI search. After answering 2-3 CRM-related questions on Zhihu weekly, Zhihu became the most frequently cited "external source" by AI.
TOP3: Direct first-paragraph answers.
Every article's first sentence gave the core answer. When extracting answers, AI almost exclusively pulled from the first paragraph.
Actions Taken But With Average Results
- Deep long-form articles: Had effect, but not as immediate as FAQ (long-form articles need a longer "accumulation period")
- Internal links: Helped SEO, but direct GEO effect was not significant
Actions Not Taken But Should Have Been
After the Week 12 review, the team agreed that if starting over, they should have done from Day 1:
- Encyclopedia entry: Establishing an encyclopedia entry takes 3-6 months, should have applied on Day 1
- Multi-AI platform monitoring: Only manually tested ChatGPT, Doubao, and Perplexity — missed Kimi and Wenxin Yiyan
The core lesson this case study demonstrates:
GEO "launching" isn't difficult — as long as you do the right things in the right order.
Weeks 1-2 fix infrastructure (let AI in) → Weeks 3-6 build content (give AI something to read) → Weeks 7-12 monitor and iterate (get AI to start citing you).
3 months, from "non-existent" in AI search to "stably cited" — this is a brand's first and most important step in GEO.
You don't need a perfect GEO strategy — you just need the determination to "start executing."