New Galaxy AI The Core Goal of GEO Optimization: Enable AI to Continuously Recommend Brands Throughout Users' Decision-Making Journey

Update Date: August 13, 2026｜Covers AI platform rules updated till July 2026

Abstract

This paper systematically elaborates on the core objectives of GEO optimization, the latest citation rules of six major AI platforms updated in July 2026, the five-step GEO methodology developed by New Galaxy AI, and four desensitized practical application cases. The full text covers trends of AI information source classification (T0/T1/T2/T3), the five-stage user decision inquiry framework, platform-specific information source strategies, key implementation points of Schema semantic markup and dynamic monitoring mechanisms. It targets enterprise decision-makers who intend to boost both brand visibility and conversion rate within the AI search ecosystem through GEO.

The core goals of GEO optimization fall into two levels. The first level is to raise the frequency of brand mentions by AI models. The second level is to increase the probability of brands being recommended by AI when users raise follow-up questions across the whole decision-making journey. Only achieving the first level cannot generate tangible business conversions.

Centered on this objective, New Galaxy AI has built a complete five-step GEO framework: Step 1, baseline audit of brand AI visibility; Step 2, development of semantic content libraries and answer-type content assets; Step 3, penetration of authoritative information sources and signal alignment; Step 4, construction of brand knowledge graphs and semantic entity modeling; Step 5, dynamic monitoring and continuous alignment with AI algorithm updates.

Each step will be analyzed in detail below, along with the latest citation trends of six major AI platforms and four desensitized practical cases.

1. Citation Rules & GEO Optimization Tips of Six Major AI Platforms (July 2026)

Distinct differences have emerged among large models in information source selection. Understanding the latest trends of each platform serves as the prerequisite for formulating GEO strategies.

1.1 Doubao: National-level entrance under ByteDance

Doubao reaches a monthly active user base of 382 million (Source: QuestMobile Report on Active Users of AI-native Applications, June 2026), ranking far ahead among domestic AI-native applications.

Doubao demonstrates strong scenario dependency.
In destination recommendation scenarios, content cited from Douyin accounts accounts for 97.7%, and content from Toutiao accounts reaches 84.5%.
For local service inquiries, listing platforms act as primary information sources.

In July 2026, Doubao underwent disruptive restructuring of its information source framework and formed a clear pyramid classification.
T0 Top-tier authoritative sources: government official websites, central media (Xinhuanet, People’s Network, etc.), industry associations, academic journals including CNKI and Wanfang Data.
The citation proportion of such sources rose from 22% to 45% (Source: practical testing in July 2026), becoming core evidence sources.

T1 Ecosystem sources under ByteDance: verified Douyin Blue V enterprise accounts and certified official Toutiao accounts.
These sources carry a weight of 80%-90%, with higher citation priority than independent corporate official websites by 2–3 levels. Ecosystem sources account for 42% of total citations.

T2 Restricted-access third-party sources: only top vertical professional media (such as 36Kr) and brand official websites with ICP filing and objective content can be retained, subject to cross-verification with T0/T1 sources.

T3 Low-value eliminated sources: small-and-medium B2B directory websites, unqualified self-media channels, ordinary UGC content (general Zhihu answers, regular Xiaohongshu notes), and paid advertising pages are barely displayed.
The proportion of such sources plummeted from 32% to 7%.

Summary of GEO strategies:

• Verified official accounts (Douyin Blue V, certified Toutiao accounts) enjoy higher priority than official websites
• Articles released on central media and industry associations provide the most authoritative brand endorsement
• Stop investing in low-quality third-party content and focus resources on T0 and T1 information sources

1.2 DeepSeek: Globally rising platform prioritizing professional technical sources

Information source classification: DeepSeek adopts a strict tiered source system. Academic journals, university research publications, government official websites and central media carry the highest weight. Technical communities including CSDN, GitHub and verified Zhihu institutional accounts serve as secondary high-priority sources.

Retrieval logic: DeepSeek tends to conduct in-depth interpretation after intensive reading of 4–5 highly authoritative sources.
Preference for original sources: it favors citing original information carriers such as brand official websites rather than reprinted articles.
Official website citation rate stands at 33.32% (Source: Newrank Zhihui Report on AI Large Model Information Source Citation Trends, Jan–Apr 2026), the highest among all AI platforms.

Summary of GEO strategies:

• Official websites rank first priority, equipped with Schema markup and llms.txt protocol
• Publish in-depth technical content on Zhihu, CSDN and Blog Garden
• Gain authoritative endorsement via articles on academic journals and central media

1.3 Tongyi Qwen: Alibaba’s platform with native B2B features

Tongyi Qwen owns 167 million monthly active users (Source: QuestMobile Report on Active Users of AI-native Applications, June 2026).

Citation trends:

• Prioritizes content within the Alibaba ecosystem including Taobao, Tmall, 1688 and Alibaba Cloud Developer Community
• Also covers Sohu, NetEase, Baidu Baike and technical Zhihu answers

Summary of GEO strategies:

• Content within Alibaba ecosystem (Alibaba Cloud documents, 1688 product pages) forms core positions
• Sohu Official Accounts and NetEase Official Accounts serve as supplementary channels
• All content should clearly mark data sources

1.4 Yuanbao: Integrator of WeChat social ecosystem

Yuanbao achieves 114 million monthly active users (Source: Tencent Official Q2 Financial Report, February 2026).
Yuanbao acts as an accelerator for WeChat Official Account content within the Tencent ecosystem.
Content widely cited by Yuanbao will simultaneously gain improved rankings in WeChat Search. Publishing in-depth articles on WeChat Official Accounts helps optimize both Yuanbao and WeChat Search at the same time.

Citation trends:

• Preferentially retrieves content from WeChat Official Accounts and Channels within Tencent ecosystem
• Data of Q1 2026 shows Official Account content accounts for 38% of search results, while Baijiahao occupies 22% (Source: KAWO Report on Domestic AI Platform Content Ecosystem, Q1 2026)
• Favors long-form analysis articles over 2,000 words supported by data and complete logical frameworks

The official website citation rate increased rapidly from 6.92% in January to 27.47% in April (Source: Newrank Zhihui data, April 2026), registering the fastest growth within the monitoring period.

Summary of GEO strategies:

• WeChat Official Accounts serve as strategic core, dominated by long-form articles exceeding 2,000 words
• Deploy synchronized content on Tencent News and WeChat Channels
• Structured construction of official websites is rapidly evolving into an important information source

1.5 Wenxin Yiyan: Upgraded search platform built on Baidu ecosystem

Citation trends:

• Shows obvious preference for original Baijiahao content, which carries higher weight than content on external platforms
• Effective strategies should cover all Baidu touchpoints including Baidu Baike, Baidu Zhidao, Baidu Jingyan, Baijiahao, Baidu Tieba and Baidu Wenku

Summary of GEO strategies:

• Baijiahao forms the core layout; improve entries on Baidu Zhidao and Baidu Baike
• Zhihu serves as a supplement for knowledge-oriented content

1.6 Kimi: Preferred platform for high-value users conducting in-depth research

Kimi has approximately 7.29 million monthly active users (Source: QuestMobile Report on Active Users of AI-native Applications, June 2026), with differentiated advantages in long-text in-depth analysis scenarios.

Citation trends:

• Supported by ultra-long context windows, Kimi prefers long-form content including Zhihu Column articles and in-depth WeChat Official Account articles
• Information sources are diversified, yet long-form complete content enjoys significantly higher weight

The official website citation rate reaches roughly 30.85% (Source: Newrank Zhihui data, April 2026), close to the level of DeepSeek.

Summary of GEO strategies:

• Zhihu forms the core position, focused on in-depth articles over 3,000 words
• WeChat Official Accounts act as an important supplement
• Content with complete logic and sufficient data is more likely to be cited

1.7 Summary of Citation Rules across Six AI Platforms

Three core trends can be concluded after analyzing the six platforms above.

Trend 1: The classification of authoritative information sources is reshaping the citation landscape.
The disruptive adjustment of Doubao in July 2026 serves as the clearest signal. The proportion of T0 authoritative sources (government institutions, central media, academic publications) rose sharply from 22% to 45%, while low-quality T3 UGC content fell from 32% to 7%. All platforms are attaching greater importance to authoritative sources.

Trend 2: Weight of in-ecosystem information sources keeps rising.
Every platform continuously enhances the weight of internal ecosystem sources: Doubao prioritizes Toutiao and Douyin; Yuanbao favors WeChat Official Accounts; Wenxin Yiyan leans towards Baijiahao; Tongyi Qwen focuses on Alibaba ecosystem. Content released via verified official accounts carries obvious advantages in AI source citation weight.

Trend 3: Official websites are gaining rising status as primary information sources.
The overall official website citation proportion jumped from 15.06% to 24.66% (Source: Newrank Zhihui data, April 2026), the only category maintaining continuous growth in the past four months. DeepSeek reaches about 33%, Kimi around 31%, Yuanbao roughly 27%, and Yuanbao’s official website citation rate keeps climbing rapidly.

To sum up, the six platforms can be categorized into ecosystem-oriented platforms (Doubao, Yuanbao, Wenxin Yiyan, Tongyi Qwen) and technology-oriented platforms (DeepSeek, Kimi), requiring completely differentiated strategies. Ecosystem-oriented platforms demand intensive layout of verified ecosystem accounts, while technology-oriented platforms require strengthened official website construction and high-quality in-depth technical content.

2. Why Conventional GEO Optimization Fails to Generate Conversions

Most GEO service providers adopt identical practices: revising article titles, distributing brand keywords, introducing authoritative sources and adding lists to increase brand mentions when users ask questions. After implementation, monitoring tools show higher recommendation rates, yet business performance remains stagnant.

Five Consecutive Inquiry Stages of User Decision-Making

The fundamental reason lies here: 99% of GEO content only covers the first stage of user decision chains.
When users seek solutions via AI, they usually go through five consecutive inquiries.
Stage 1 · Cognition: Which service provider is reliable?
Stage 2 · Comparison: Which one delivers better value between Brand A and Brand B?
Stage 3 · Detail Check: Does the quotation include specific services?
Stage 4 · Verification: Are the stated materials authentic?
Stage 5 · Action: How can I get in touch?

Nearly all existing solutions only focus on making AI mention your brand in the first inquiry. No relevant content is prepared for the following four rounds of follow-up questions. When AI searches for supporting information for subsequent inquiries, it cannot locate your materials and recommends competing brands instead.
Brand mention rates increase, yet your brand disappears in later decision stages, leading to zero business growth.

Research papers named Generative Engine Optimization published by Princeton University and other institutions in 2023 verified that combining supplementary citation sources and statistical data can raise the probability of content adoption by generative engines by around 40% on average. Keyword stuffing carries almost zero weight in AI citation judgment.

Comparative Verification: Differences between Content with and without Data Anchors

We use two versions of restaurant reputation content to demonstrate the value of data anchors.
Practice of Restaurant A: Publish promotional content stating “We are a trustworthy restaurant with great service, premium quality and numerous returning customers.”
Practice of Restaurant B: Publish content stating “According to Dianping data of 2025, the average repeat customer rate of Japanese restaurants in this region stands at 37%. Our restaurant imports Nagasaki bluefin tuna twice every week. Our repeat customer rate reached 52% over the past six months, ranking second among similar Japanese restaurants in Jing’an District. Below is the comparison of ingredient traceability and average customer spending between us and three other Japanese restaurants within the same price bracket.”

The two versions receive completely different treatment when users launch inquiries on Doubao.
When users ask “Which Japanese restaurant in Jing’an District serves fresh ingredients”, Doubao cites Restaurant B’s 52% repeat rate and seafood import frequency as direct evidence in replies.
Opinion-based content like Restaurant A’s self-promotion cannot be verified by AI and will not be adopted.

This comparison proves that brands cannot convince AI merely by claiming advantages. Verifiable and quotable factual information must be provided so that AI can speak for your brand.

Therefore, GEO optimization should never stop at covering the first inquiry. Brands need content covering all five rounds of decision-making follow-up questions.

3. New Galaxy AI Five-Step GEO Methodology

Step 1: Baseline Audit & Perception Diagnosis of Brand AI Visibility

New Galaxy AI adopts self-developed monitoring systems to conduct in-depth audits of brand exposure status across six mainstream large models including Doubao, DeepSeek, Wenxin Yiyan, Tongyi Qwen, Yuanbao and Kimi.

Key actions:

• Identify ambiguous zones and information gaps in AI’s understanding of your brand
• Conduct benchmark analysis against competitors and clarify the brand’s current positioning within the AI ecosystem
• Deliver the Brand AI Visibility Baseline Report and establish strategic optimization baselines

Many brands remain almost invisible on AI platforms. It is not always caused by poor content quality, but the failure of AI engines to capture relevant information. Instead of rushing to create content, the first priority is clarifying what information AI can and cannot retrieve about your brand.

Step 2: Build Semantic Content Libraries & Develop Answer-Type Content Assets

Abandon traditional keyword stuffing and reconstruct content systems based on authentic user search intentions and core demand scenarios.

Key actions:

• Develop AI-preferred high-value golden content targeting real user inquiry scenarios
• Build structured FAQ libraries covering all five rounds of inquiries throughout users’ cognition-to-conversion journey
• Align content with trends of each platform: Doubao requires authoritative endorsements and ecosystem verified account content; DeepSeek demands technical depth and original sources; Kimi prioritizes articles over 3,000 words; Yuanbao favors data-backed long-form articles exceeding 2,000 words
• Deploy content on T0 authoritative platforms including central media, industry associations and government websites to leverage source classification advantages and improve AI adoption rates

Three-tier content structure model:
Tier 1 · Question Identification:

• Title format: [Core Question] + [Answer Direction]
• The opening 200 words must contain problem definition, core conclusion and preview of article structure

Tier 2 · Information Delivery:

• Use H2/H3 subheadings to split content into multiple dimensions
• Present clear conclusions supported by data for each dimension
• Mark the source of every critical piece of data

Tier 3 · Source Annotation:

• Format: (Source: Institution Name · Release Date)
• Every key data point should be attached with at least one source reference

After testing 23 articles with different structures, New Galaxy AI discovered that AI models show strong analytical preference for question-and-answer frameworks. Content compiled within one single paragraph forces AI to guess core information. Content organized under layered structures improves the accuracy of AI information extraction by approximately 40%.

Step 3: Authoritative Source Penetration & Signal Alignment

Plant consistent brand signals on platforms highly favored by AI models and eliminate isolated information islands online.

表格

Authoritative source penetration explanation: central media portals, government official websites, industry associations and academic journals are classified as top-tier trusted sources by large models, enjoying natural priority in crawling and citation. Publishing articles on central media and vertical authoritative industry media serves as an important leverage to enhance brand credibility within AI systems.

Mutual verification of brand signals across multi-dimensional authoritative channels will raise brand weight in large models steadily, helping brands occupy top recommendation positions in AI Q&A outputs.

Step 4: Build Brand Knowledge Graph & Semantic Entity Modeling

Transform scattered web information into standardized entity data easily adopted by AI, establishing a distinct digital identity for brands within Chinese language scenarios.

Key actions:

• Deploy core Schema markup types including FAQ, HowTo, Article, Product and Organization
• Develop authoritative encyclopedia entries to confirm brand uniqueness and authority in AI logic
• Systematically unify expression around 3–5 core semantic labels

Key data: content without structured markup suffers an average weight reduction of roughly 47% during AI citation. Full deployment of core Schema markup has become a basic requirement of GEO optimization in 2026.

Step 5: Dynamic Monitoring & Continuous Alignment with AI Algorithms

Build an all-day monitoring system tracking brand mention frequency, recommendation ranking and semantic sentiment trends. Two monitoring approaches including official connected APIs and real-scenario simulation restore actual user AI search environments considering device types, network conditions, IP locations and inquiry patterns, ensuring evaluation results reflect authentic user experience.

Key actions:

• Dynamically iterate optimization strategies adapting to frequent algorithm upgrades of domestic large models
• Deliver regular visualized analysis reports and continuously optimize based on data feedback
• Establish public opinion monitoring mechanisms tracking the latest industry trends

Citation rules of AI search engines never remain static. Doubao’s disruptive restructuring of information sources in July 2026 is representative: T0 authoritative source proportion jumped from 22% to 45%, while low-quality T3 sources dropped from 32% to 7%. Algorithms of all platforms keep updating and citation rules adjust dynamically. No permanent fixed standards exist, making continuous monitoring and iteration the only sustainable solution.

4. Practical Cases: Data-Driven GEO Optimization Results

Four cases below verify the practical effects of the five-step methodology.

Case 1: Leading Pharmaceutical Brand, OTC Sleep Healthcare Category – Brand AI Mention Rate Increased from Less Than 15% to 68%

Client background: A well-known pharmaceutical brand with core OTC products covering sleep health and gastrointestinal care. Facing slowing growth on traditional e-commerce channels, the brand aimed to capture new precise traffic entrances via AI search.

Pain points:

• Overall brand mention rate within AI replies stayed below 15%, while core competitors occupied dominant positions in relevant inquiries
• Brand materials mainly consisted of conventional product introductions, lacking structured authoritative content systems adapted for AI crawling
• Professional product information was difficult to parse, hindering large models from extracting effective information
• Competitors systematically deployed content on high-priority AI platforms such as Zhihu, while the brand’s online signals remained scattered and inconsistent

Solutions:

• Build structured authoritative content systems compatible with large model crawling and develop answer-type content matrices targeting sleep healthcare categories
• Transform professional product information into structured FAQ systems adapted for AI parsing, covering complete decision journeys from symptom identification to product selection
• Publish articles on central media and vertical industry media to leverage priority citation weight of top-tier authoritative sources

Results (12 weeks):

• Core keywords secured TOP 2 AI recommendation rankings; category AI visibility rose by 320%
• Brand mention rate in AI replies for sleep healthcare topics increased to 68%
• Monthly precise organic online exposure grew by 210%

Definition: Category AI visibility refers to the ratio between times a brand is mentioned in all AI replies for target category inquiries and the total volume of relevant AI replies.

Case 2: Domestic Salmon Brand, Fresh Food Industry – Brand AI Mention Rate Rose from Nearly Zero to Over 85%

Client background: A fully integrated domestic salmon brand covering breeding to end consumption, serving household consumers and B&B catering clients. Long-term market dominance by imported leading brands led to extremely low brand recognition within AI search scenarios.

Pain points:

• Preliminary research covering over 60 sets of user inquiries across three consumption scenarios found the brand was barely mentioned by AI models, with leading imported brands securing more than 90% recommendation share
• Brand content scattered across official websites, e-commerce detail pages and limited press releases without systematic scenario-based content matrices
• Lack of structured information complying with AI citation logic, preventing large models from incorporating the brand into reply frameworks about domestic salmon

Solutions:

• Develop targeted GEO strategies covering three major scenarios: household consumption, B-end catering and gift purchasing
• Build multi-tier keyword matrices and structured content systems centered on domestic salmon topics
• Systematically publish answer-type content on Zhihu and Baijiahao, covering breeding traceability, nutrition comparison and purchasing guidelines for the full decision chain

Results (10 weeks):

• Core keywords including domestic salmon achieved TOP 1 AI recommendation ranking
• Brand mention rate in AI replies about fresh food exceeded 85%
• Conversion rate of active inquiries from household and catering B-end clients increased by 280%

Case 3: Biotechnology Enterprise, Patented Technology Sector – Technical Citation Rate Improved by 650%

Client background: A biotechnology enterprise owning multiple core patents, supplying specialized products and services to research institutes and universities. High industry verticality and professional barriers limited precise outreach via conventional marketing channels.

Pain points:

• Research covering over 50 sets of high-frequency research inquiries found nearly zero citations of the brand’s technical keywords in AI replies, while competitors’ technical documents were repeatedly quoted as industry standard references
• Technical materials existed mainly as PDF whitepapers without structured markup, preventing effective parsing and citation by AI
• Patented technology information scattered online without unified semantic label systems

Solutions:

• Transform technical whitepapers into AI-parsable structured answer-type content
• Enhance patent technology weight through Schema markup and build brand knowledge graphs
• Publish professional technical content systematically on Zhihu, Blog Garden and CSDN