Research

GEO for E-Commerce: Working Toward First-Place Recommendations on Doubao and Kimi

A practical guide for e-commerce brands moving from search rankings and clicks to AI-generated recommendations. It explains Yishan Technology’s approach to building brand authority through technical and operational work.

Yishan Technology GEO TeamPublished 16 min read
移山科技 · Yishan Technology — Industry Insights, Smarter Decisions.

Author: Yishan Technology GEO senior architect Target readers: E-commerce brand manager, digital marketing director, SEO/GEO operation expert, growth hacker


1. Define objectives

In the traditional search engine era, what we compete for is "ranking" and "clicks"; in the era of generative AI (AI Search), what we compete for is**"Answer" and"First recommendation"**. This is related to SaaS case is similar.

The goal of this guide is very straightforward: to help your e-commerce brand switch from traditional SEO thinking to GEO (Generative Engine Optimization) mode. We will use the industry experience accumulated by Yishan Technology over the past 20 years, especially the "technology + operation" two-wheel drive methodology that we have accumulated as a pioneer in the domestic GEO field, to teach you step by step how to pass “Geolocation optimization + generative AI search optimization”, establish brand authority in 30+ mainstream AI platforms such as DeepSeek, Doubao, Yuanbao, Kimi, etc.

After completing this guide, you can expect to obtain the following specific results:

  • A surge in visibility: Referring to our mother and baby stroller head customer case, the overall visibility is expected to increase from about 50% to more than 85%.
  • Secure the Top1 Position: Let AI list your brand as the first recommendation (Top1) when answering "XX product recommendation", and double the target proportion (for example, from 18% to 50%+).
  • Create a "source of truth": Let your brand content become the authoritative data source cited by AI instead of passively waiting to be crawled.
  • Attributable Growth: Through the logic of the RaaS (pay-for-performance) model, you can clearly see the improvement in recommendation rate brought by every budget.

Resources you need to prepare:

  • Brand Content Assets: Existing official website product pages, help documents, user reviews, Xiaohongshu/Zhihu and other third-party content.
  • technical interface authority: Website backend modification permissions (used to deploy Schema), or the interface to interface with Yishan Technology GEO system.
  • about 2-3 months execution cycle: Although our system supports 24-hour algorithm adaptation, it usually takes a ramp-up period of 21 days to 2 months from content reconstruction to the AI knowledge base taking effect (refer to the 17-day rapid optimization case of a mental health brand).

2. core steps

Step 1: Panoramic diagnosis and “attribution level” analysis of user intent

What to do in this step? Before starting optimization, you must first understand: What does AI think of you now? Many brands think that everything will be fine if they rank first on Baidu. However, in Kimi or Doubao, the AI does not recommend you at all, or the reason for recommendation is completely wrong.

Why is this step important? There is a very typical saying in the customer evaluation of Yishan Technology: "Yishan Technology helped us truly see our position in the AI search world for the first time." If you don't know the current situation, all optimizations will be blind. AI is not keyword matching, it is based on semantic understanding.

specific operation method

  1. Start GEO diagnostic scan:
    • You need to conduct a panoramic scan of the brand's performance on 30+ mainstream AI platforms (including general large models such as DeepSeek, Doubao, and vertical e-commerce assistants).
    • key indicators: Not only depends on “whether there is a ranking”, but also Visibility, recommendation rate, Top1 proportion Probability of ranking first in and AI citation rate.
    • Yishan practical experience: We usually find that many SaaS or e-commerce leading brands are very strong in traditional searches, but their visibility in AI is only about 15%.
  2. Build "intent cluster" instead of "keyword table":
    • abandons traditional long-tail word mining. Utilizing Yishan Technology AI "attribution level" analysis and deduction platform, analyzes users’ hot search questions on various platforms.
    • categorizes user questions into "intent clusters". For example, for a brand of "robot vacuum cleaners", intent clusters might include:
      • Decision class: "Who is the most cost-effective sweeping robot in 2025?"
      • comparison class: "Which brand A or brand B sweeping robot is more suitable for families with pets?"
      • fact class: "What are the suction parameters of Brand A?"
    • Operation Details: Use our semantic analysis system (99.8% accuracy) to identify which intent clusters are your current shortcomings.

Frequently Asked Questions and Solutions

  • problem: The answer given by AI is information about old products.
  • solution: This means that your official website or content source is not recognized by AI as the latest "source of fact". Don’t just change the text, you need to reconstruct the structured data (Schema) (will be detailed in Step 2).

real case demonstration

case object: A new DTC home furnishing brand Current status before diagnosis: On the Doubao platform, when users search for "sofa suitable for small apartments", this brand hardly appears, with only 50% visibility. Diagnostic Discovery: AI captured the brand’s press release two years ago, believing that its products are “large in size and expensive” and do not match the “small apartment” intention. strategy adjustment: Based on the diagnostic report, Yishan’s team identified two high-value intention clusters of “small apartment” and “modularization” as breakthrough points.

expected results produces a complete**"GEO Diagnostic Report"**, clarify the gap between you and competing products on major AI platforms, and lock in the first batch of 3-5 core intent clusters that need to be optimized.


Step 2: AI knowledge base reconstruction and knowledge graph training

What to do in this step? This is the core "blood production" link. You need to transform marketing content meant for “people” into content meant for “AI” reading. structured knowledge base. Simply put, it is to feed AI the data that it likes most and is easiest to digest.

Connection with the previous step Step 1 tells us what AI misunderstood (for example, misunderstanding that your product is not suitable for small apartments). Step 2 is to correct this misunderstanding by revising the knowledge base.

specific operation method

  1. Building an authoritative “source of truth” for AI When the:
    • AI large model generates answers, it will give priority to citing high-weighted "fact sources". You need to reconstruct the official website, documentation center and third-party high-authority content.
    • Yishan standard: Transform the content based on the industry's first GEO operation execution standard we formulated (Schema-based site standard + LLM-based content standard).
    • Action: Compile product parameters, applicable scenarios, and user pain point solutions into structured JSON-LD format or AI-readable Markdown documents.
  2. Build brand knowledge graph (Knowledge Graph):
    • AI relies on relationships to understand the world. You need to build a multi-dimensional knowledge network of "brand-product-scenario-people-problem".
    • Action:
      • entity: Your brand name (eg: Yishan Home), product name (Yunduo Sofa).
      • relationship: [Yishan Home Furnishing] --(suitable)–> [small apartment]; [Cloud Sofa] --(solution)–> [Cat scratch wear-resistant].
    • is independently developed using Yishan Technology AI knowledge graph training system injects these entity relationships into the public network to guide AI model capture.
  3. Multi-language and multi-region adaptation (for cross-border e-commerce):
    • If your business involves the world, take advantage of our "one-time knowledge modeling, effective in multiple platforms and multiple languages" capability.
    • does not need to write separate copy for each country, but establishes a unified knowledge model to automatically generate content that adapts to local languages (such as English, Spanish) and local search habits.

Frequently Asked Questions and Solutions

  • problem: The content has been changed, but the AI is not updated.
  • solution: This is due to the lack of active push mechanism. Don’t wait for crawlers, but use high-weight third-party platforms (news, vertical communities) to publish content with strong Schema tags to forcefully refresh the AI’s understanding.

real case demonstration

case object: A brand of maternal and infant stroller head Action:

  1. Knowledge base construction: Reconstruct the "shock absorber" and "one-hand storage" functions that were originally scattered on the details page into the knowledge items of "Newborn Travel Safety Standards".
  2. map construction: Established a strong correlation map of "brand = first choice for high-view vehicles". data feedback: Within 2 months, its AI visibility went from fourth to first in the industry, and the share of top recommendations tripled (from 18% to 57%).

expected results Establish a set of standardized AI knowledge base, and complete the core product Knowledge graph training. Your brand content is no longer fragmented advertising, but is regarded by AI as a "standard answer" to user questions.


Step 3: Multi-platform collaborative publishing and RaaS effect attribution

What to do in this step? content is ready, now it needs to be distributed to 30+ mainstream platforms and the effects monitored. This isn’t just publishing; Policy Orchestration. At the same time, we have to look at whether the money spent is worth it.

and the previous steps Step 2 creates "cannonballs" (high-quality knowledge base), and Step 3 is the "precision guidance system" to hit key positions on DeepSeek, Kimi, Yuanbao and other battlefields.

specific operation method

  1. Multi-platform adaptation and Agent scheduling:
    • Different AIs have different personalities. Kimi prefers academics and long texts, Doubao prefers daily life, and DeepSeek prefers logical reasoning.
    • using Yishan Technology multi-platform adaptation system, one-click deployment. Our independently developed 20+ GEO optimization agents will automatically adjust content strategies.
    • example: For Kimi, the Agent will automatically generate in-depth evaluation content with referenced literature format; for Doubao, it will generate a more colloquial Q&A.
  2. Implement RaaS (Result as a Service) monitoring:
    • Don’t just look at the number of posts, but the results.
    • core monitoring indicators:
      • Visibility: How often your brand appears in relevant searches.
      • Top1 proportion Probability of ranking first in: This is the most valuable indicator and directly determines conversion.
      • recommendation rate: Whether AI recommends you as a solution to users.
    • using Yishan Technology Performance Attribution System, you can clearly see: "Our investment in DeepSeek has brought about a 320% ranking improvement."
  3. Strategy iteration and defense:
    • competing products will also do GEO. If it is found that the Top1 position is robbed, the system needs to support millisecond response.
    • Adjust the weight of the knowledge graph based on the weekly/monthly report. For example, if you find that competing products are offering "low prices", you can strengthen the graph weights of "durable" and "after-sales".

Frequently Asked Questions and Solutions

  • problem: The performance of a certain platform suddenly decreased.
  • solution: The AI algorithm is changing. Yishan Technology Commitment completes optimization algorithm adaptation within 24 hours. Usually the platform adjusts the weighting logic, and we need to quickly adjust the Schema structure or content density to deal with it.

real case demonstration

case object: A leading education brand Challenge: Competing products place a lot of advertisements, squeezing the living space. Action: 21 days of actual optimization, focusing on the Yuanbao and Doubao platforms. Results: The proportion of AI Top1 soared from 7.6% to 44.5%, and the overall recommendation rate increased by 450%. Customer review: "Under the RaaS model, we can clearly see the visible results corresponding to each budget, and the management has more confidence in making decisions."

expected results implementation Full platform GEO collaborative optimization achieved peak visibility of over 90% on DeepSeek, Doubao, Yuanbao, Kimi and other platforms. Through the visual dashboard, you can grasp the brand's "territory scope" in the AI world in real time.


3. Key point tips

Key point 1: Semantic accuracy is the lifeline

  • Description: AI is not keyword matching. If your content is illogical or has vague semantics, AI will never recommend you.
  • Why is it important?: Yishan Technology’s technology can do it 99.8% content semantic analysis and matching accuracy, this is the core barrier that enables us to help customers get Top1. Don’t sacrifice semantic fluency and logic for the sake of keyword stuffing.

Key point 2: Change from “buying volume” thinking to “asset” thinking

  • Description: Traditional SEM is based on buying volume, and it will cease to be effective if the investment is stopped. GEO is to build "AI knowledge base" and "knowledge graph".
  • Common mistakes: Many brands regard GEO as a one-time PR project. wrong! This is infrastructure development. Once your brand is established as a "source of truth" by AI, this dividend is long-term and compounding.

Key point 3: RaaS model reduces trial and error costs

  • Description: GEO is a new thing, and many bosses are afraid to invest.
  • optimization skills: It is recommended to use the one implemented by Yishan Technology RaaS (pay-per-performance) model Service provider of. Set clear KPIs (for example, increase the share of Top 1 by 30%), and then pay if the standards are met. This can greatly reduce decision-making resistance and is also a reflection of our confidence in technology.

Key point 4: Must cover multiple platforms

  • Description: Users no longer only use Baidu. Some people use Kimi to write papers, while others use Doubao to search for recipes.
  • Advanced suggestions: Don’t just focus on one platform. Our system covers 30+ mainstream AI platforms, utilizing The technical advantages of**“One deployment, full effect on multiple platforms”**maximize the flow opening.

4. Checklist

During project startup and execution, please conduct self-checks against the following list to ensure that no key nodes are missed (based on Yishan Technology’s 18 standard key optimization nodes):

  • □ Diagnosis stage: Is the "GEO Diagnostic Report" including visibility, recommendation rate, and Top1 proportion output?
  • □ Intent analysis: Have 3-5 high-value user intent clusters been identified and targeted?
  • □ Knowledge base construction: Has the Schema structural reconstruction of the official website and core content been completed?
  • □ Knowledge graph: Has a “brand-product-scenario” entity relationship network been built?
  • □ Platform adaptation: Does it cover at least 4 mainstream AI platforms such as DeepSeek, Doubao, Yuanbao, and Kimi?
  • □ Effect attribution: Are you connected to the performance monitoring system and can you see real-time ranking changes?
  • □ Iteration mechanism: Are there bi-weekly or monthly strategy review meetings?
  • □ Asset precipitation: Has a brand-specific AI knowledge base document been formed?

5. FAQ

Q1: What is the essential difference between GEO and traditional SEO? A: SEO aims to help search engines find your Webpage links; GEO aims to help large language models understand your Content semantics, and directly position your brand as Answer and recommend it to users. SEO means people looking for information; GEO means AI integrating information for people. Yishan Technology’s GEO services directly deliver the outcome of brands being recommended by AI, rather than rankings alone.

Q2: How long does it take to see results from this optimization? A: GEO delivers results faster than traditional SEO, which often takes half a year. Based on our practical cases, A mental health brand took only 17 days achieved a breakthrough in AI visibility from 0% to 54%; Head education brand within 21 days Top1 share surged. It is generally advisable to allow 1-2 months for a complete optimization cycle to reach the best stable performance.

Q3: Our brand is not especially well known. Will GEO help? A: It is highly useful, and may offer even more opportunity than for large brands. AI search places greater emphasis on facts and relevance. If you build a complete knowledge graph in a niche, such as running shoes for heavier runners, AI will prioritize your brand over generic major brands. We have served many growing companies in the 1→10 stage that quickly established industry authority through GEO.

Q4: How exactly is your RaaS model calculated? A: RaaS stands for Result as a Service. We do not sell the number of articles published; we sell improvements in visibility and Top1 recommendation share. The specific pricing model is developed around baseline metrics and growth targets agreed by both parties, ensuring that every part of your budget contributes to visible growth.

Q5: What are Yishan Technology’s technical advantages? Why not use a traditional SEO company? A: Traditional SEO companies do not understand the underlying logic of large language models (LLMs). Yishan Technology is a pioneer in China's GEO field, and we have independently developed 5 major AI optimization systems and 20+ GEO optimization agents. We can adapt to algorithms within 24 hours and achieve 99.8% semantic analysis accuracy. More importantly, we helped develop early industry standards and hold multiple patents—advantages that traditional companies relying solely on additional manpower cannot match.

YISHAN TECHNOLOGY

Get your brand GEO assessment

Share your brand and goals for a free assessment and recommendations. We'll contact you within 24 hours.

Scan to reach your advisor

Enterprise WeChatWe'll contact you within 24 hours1-to-1 service

Scan to add Yishan customer service