www.siyou365.com/company/We did it on our own site first
This site was rebuilt against three conditions — crawlable, identifiable, citable: llms.txt, Organization structured data, an entity page, and answer blocks an AI can quote whole.
AI Agents · AI Apps · GEO
Put AI to work for you — and get AI to recommend you
We build AI agents, applications and knowledge bases, and use GEO to get your brand cited by ChatGPT, Perplexity, Gemini and Claude — and DeepSeek, Doubao, Qwen and Yuanbao for the China market. This site is our own GEO case study.
Facts you can check
- 3 GEO sitessiyou365.com / scimpexp.com / tax.siyou365.com
Six practices: put AI to work, and get AI to recommend you
AI Agent Development
Hand repetitive workflows to an AI agent that decides and self-heals
- read requirements and pages
- call tools · decide autonomously
- self-heal on failure · traceable logs
- Repetitive manual workflows eat up team capacity
- Off-the-shelf SaaS tools don't fit internal systems and processes
- Script-based automation breaks widely whenever the environment changes
AI App, Mini-Program & E-commerce Development
Bring AI capability into your mini-program, app and e-commerce systems
You want AI features but don't know where to start
View service
This site is the case study
GEO — Generative Engine Optimization
Get ChatGPT, Perplexity, Gemini and Claude — and DeepSeek, Doubao, Qwen and Yuanbao for the China market to cite you, not your competitors
Buyers can’t find you in ChatGPT, DeepSeek or Doubao, or the AI confuses you with another company
View serviceAI Skill Development
Package your team's expert workflows into reusable AI skills
Expert know-how lives in a few people's heads and is hard to replicate
View serviceEnterprise Knowledge Base
Turn scattered documents into an AI-searchable knowledge base
Experience walks out the door when staff leave, and onboarding is slow
View serviceAI Assistant Development
An AI assistant that actually knows your business context
Generic assistants don't know your internal processes or terminology
View serviceWe are our own case study
We measure it the same way we measure yours
Three baselines on our own domain (B0 / B1 / Round 1), recording for every buyer question whether each AI engine cites us, mentions us, or leaves us out.
CasesOur agents run on us before they run on you
Regression testing for this site is done by our own AI test agent: it reads the cases, generates the scripts, runs them and repairs selectors that drift.
AI delivery case studies
AI Agent
AI Test-Automation Agent
A testing agent that reads requirements and pages, then generates, runs and self-heals end-to-end test cases with a traceable report.
Claude CodePlaywrightPythonGitHub Actions
AI Skill
GEO Skill: our GEO methodology packaged as an agent skill
Our own Claude Code / agent skill: the crawlable → identifiable → citable workflow, site-health scripts and baseline scoring, packaged as one reusable skill we run on our own and client sites.
Claude Code SkillPythonPlaywrightMarkdown
GEO
scimpexp.com: bilingual site build and GEO rebuild
A bilingual site for a food import/export company in Tianjin, rebuilt for GEO: separate Chinese and English URLs, hreflang, entity structured data and answer blocks an AI engine can quote whole.
More delivered projects
Taurus Payment Gateway
Enterprise Systems
Nationwide TLS certificate monitoring platform
Enterprise Systems
Tax Calculator: an online tax estimation tool
Tools
Travelog: travel booking platform website
Web & Platform
Pipixiu: website for a tax and accounting agency
Web & Platform
Qingjian Card-Key: a card-key generation and validation mini-program
Tools
All cases
Three stages, from assessment to running in production
-
1
Assess
- Map your workflows, system interfaces and data boundaries
- Agree on scope, acceptance criteria and a fixed quote
-
2
Build
- Prove feasibility with a POC first
- Development, integration and internal acceptance
-
3
Launch & operate
- Deployment, training, source code and documentation
- Free maintenance period and iteration on request
Pricing, timelines and delivery
How much does an AI app, mini-program or e-commerce app cost?
Cost depends on scope, model integration and deployment. We quote a fixed price after a requirements review; MVP ranges are below, full versions are quoted per module.
| Type | MVP timeline | Notes |
|---|---|---|
| Mini-program | 3–5 weeks | Incl. admin |
| AI app (LLM) | 2–4 weeks | Cloud API or on-prem |
| E-commerce app | 6–12 weeks | iOS/Android + payments |
How long do AI agent, knowledge base and GEO projects take?
Timelines vary by complexity: an AI agent runs from a POC to full delivery over weeks to months; a standard knowledge base ships within weeks; GEO foundation work usually takes 2–4 weeks.
| Project | Timeline |
|---|---|
| AI agent POC | 1 week |
| AI agent full version | MVP 2–4 weeks / full 6–12 weeks |
| Knowledge base (standard) | A few weeks |
| GEO foundation | 2–4 weeks |
Do we get the source code? Who owns the IP?
You receive all source code and deployment docs at delivery; IP belongs to you. We keep reuse rights only to generic components and will sign an NDA.
#source-ipHow long is free maintenance after delivery?
Delivery includes a free maintenance period covering defects caused by our code, with the term set in the contract; annual maintenance can follow, and new features or changes from third-party API updates are scoped and quoted separately.
#maintenanceWhat is GEO?
GEO (Generative Engine Optimization) makes AI engines such as ChatGPT, Perplexity, Gemini, Claude, DeepSeek, Doubao and Qwen find, correctly identify and cite your site. It rests on three conditions: crawlable, identifiable, citable.
#geo-whatHow is GEO different from SEO?
SEO optimizes for search-engine ranking and clicks; GEO optimizes for whether an AI engine cites you directly in its answer. They complement each other but measure and fix different things.
| Dimension | SEO | GEO |
|---|---|---|
| Goal | Ranking & clicks | AI citation |
| Measurement | Keyword rank | Prompt citation rate |
| Key actions | Backlinks, keywords | Entity, schema, llms.txt |
How long until GEO shows results?
Crawler hits typically appear 2–6 weeks after indexing submission; answer visibility lags crawling. We measure with B0 / B1 / Round 1 baselines, with the first re-test about 2 weeks after launch.
#geo-timeIs our knowledge-base data secure?
On-prem data always stays inside your own environment, with access control and audit logging; cloud plans use encrypted transport and storage, and we can sign an NDA.
#kb-securitySee how AI describes your company right now
A free GEO baseline audit: 12 buyer questions across 3 AI engines, reported within 3 business days, including crawl, entity and structured-data checks.