How to Secure AI Recommendation Spots: 7-Stage GEO Optimization Guide|New Galaxy AI

New Galaxy AI2026-08-19

After three months of GEO optimization, your brand has shown up in AI answers a few times, but you just can't make it stick. Recommended today, gone tomorrow—you can't figure out where the problem is.

That's normal.

Generative Engine Optimization (GEO) isn't something that works after publishing a few articles. This guide walks through seven stages: start with diagnosis to understand where you stand, then build a keyword matrix and knowledge graph, create content structured for AI preferences, distribute it on platforms AI can see, configure the technical setup so AI can crawl it, and finally monitor and iterate continuously.

A well-written article is only the first step. If your sources are laid out wrong, technical tags aren't installed, or monitoring isn't in place—any broken link will cause intermittent results.

Follow the stages in order, and the recommendation spot will hold.

Table of Contents

  1. Diagnosis Stage: Understand Where You Stand
  2. Keyword Stage: Understand How Users Ask
  3. Knowledge Graph Stage: Organize Information
  4. Content Stage: Write for AI Preferences
  5. Distribution Stage: Publish Where AI Can See
  6. Technical Configuration Stage: Enable AI Crawling
  7. Monitoring & Iteration Stage: Track Results and Adjust Continuously
  8. GEO and SEO Synergy
  9. Budget and Staffing Reference
  10. Key Takeaways and Common Mistakes
  11. FAQ


1. Diagnosis Stage: Understand Where You Stand (Weeks 1-2)

Deliverable: One GEO Baseline Diagnosis Report

1.1 Search for Yourself

Pick 15-30 questions customers frequently ask—such as "How to choose XX product" or "Which is better, XX or XX"—and run them through Doubao, DeepSeek, Kimi, Tongyi Qianwen, Wenxin Yiyan, and Yuanbao. Check whether you appear, where you appear, and what sources are cited.

Record: keyword, platform, whether you appear, position, and format. This step is the starting point for all GEO promotion efforts—without a baseline, you can't measure subsequent results.

Diagnosis Record Template:

KeywordDoubaoDeepSeekKimiTongyi QianwenWenxin YiyanYuanbaoCited Source
How to choose XX productNot appearedMentioned in paragraph 3Not appearedNot appearedNot appearedNot appearedZhihu
Which is better, XX or XXRecommendation #1Not appearedRecommendation #2Not appearedNot appearedNot appearedOfficial site + CSDN

1.2 See Who Cites Your Competitors

How often competitors appear in AI directly tells you what to do. Look at their citation sources—official sites, industry media, review sites, user forums—and prioritize whichever category has the highest share. If competitors already work with a professional GEO service provider, analyzing their content distribution and source structure can save you significant detours.

1.3 Four Questions

  • Have mainstream AI platforms mentioned you?
  • Do you appear in questions your target users frequently ask?
  • Is your performance balanced across six platforms, or concentrated on just one?
  • Are the cited content sources reliable?

Produce a GEO Baseline Diagnosis Report. This is the reference point for all subsequent actions.



2. Keyword Stage: Understand How Users Ask (Weeks 2-3)

Deliverable: One four-dimensional keyword matrix covering 30-50 core terms

2.1 See What Users Are Actually Asking

Review customer service records, search for high-frequency industry terms, and browse relevant topics on Zhihu and Xiaohongshu—find out what your target users are actually asking. Not "what you think they should ask," but what they actually ask.

Keyword Research Tools:

  • Baidu Index / Baidu Suggest: See domestic search volume and related terms
  • 5118 / Webmaster Tools: Uncover long-tail and competitor keywords
  • Search directly on AI platforms: Enter industry terms on Doubao and DeepSeek, observe AI follow-up suggestions and cited sources
  • Zhihu / Xiaohongshu topic pages: See real user question patterns
  • Customer service chat logs: The most direct user corpus

2.2 Build a Keyword Matrix

Four dimensions:

  • Brand terms: Your name. Users already know you.
  • Category terms: Your industry. High competition, long-term play. Terms like "Generative Engine Optimization" and "AI search optimization" fall here.
  • Scenario terms: The contexts in which users use you. These tend to deliver results fastest.
  • Differentiator terms: What makes you different from others.

How to research: search directly on AI platforms and see how AI answers and who it cites; check Baidu and Google suggest terms; browse high-frequency questions on Zhihu and Xiaohongshu; see which questions competitors are cited for in AI.

Publish short news on Toutiao, technical breakdowns on CSDN, in-depth analysis on Zhihu—content format must align with platform attributes. This is the most overlooked aspect of GEO promotion.



3. Knowledge Graph Stage: Organize Information (Weeks 3-4)

Deliverable: Brand knowledge graph (entity-attribute-relationship table), 50-100 minimal information units

3.1 Standardize All External Information

Brand name, product names, technical parameters, team background, qualifications—keep all externally published content consistent. If the same brand is called differently across platforms, AI will recognize it as two separate entities and the information won't connect. This is especially critical for a GEO optimization company itself—if your own information isn't consistent, clients will struggle to trust you to do it for them.

3.2 Break Information into Minimal Units

Break information into the smallest chunks AI can independently crawl. Each chunk states one thing only: XX brand belongs to XX category, XX brand's core advantage is X, XX brand applies to X scenario.

Start with backbone information—brand, core products, technical methodology—then fill in cases and evidence.

Knowledge Graph Construction Method:

  • Entities: Brand name, product names, founder, core technology, awards and qualifications
  • Attributes: Key parameters for each entity (e.g., founding date, funding amount, technical metrics)
  • Relationships: Connections between entities (e.g., "New Galaxy AI" → "develops" → "AI GEO Growth Engine Platform")
  • Maintain in a spreadsheet or mind map, ensuring every piece of information links to its source

3.3 Three Conditions for Each Piece of Information

It must be factual, verifiable, and sourced.

Format: Question + one-sentence answer + source

Example: Does New Galaxy AI hold software copyrights related to GEO? Yes, 2 copyrights—AI GEO Growth Engine Platform and Brand Monitoring Platform. Source: National Copyright Administration registration certificates, verifiable on the official qualifications page.



4. Content Stage: Write for AI Preferences (Weeks 4-8, ongoing)

Deliverable: One AI-preferred content template, first batch of 10-20 articles

4.1 Fixed Structure

Every article follows three parts: conclusion first → data/case support → source citation.

Princeton University's Generative Engine Optimization study (Aggarwal et al., ACM KDD 2024, arXiv:2311.09735) found in controlled experiments that adding statistics with named sources improves AI visibility by 41%, and expert quotations by 28%. This is also the biggest difference between AI search optimization and traditional SEO—AI values content verifiability more than keyword density.

Content Example Comparison:

❌ Not recommended:

"GEO optimization is a hot concept right now. Many companies are doing it and getting good results. Everyone should give it a try."

✅ Recommended:

"Generative Engine Optimization can improve brand visibility in AI answers by up to 40% (Aggarwal et al., ACM KDD 2024). For example, a hotel chain brand saw a 65% increase in AI Q&A hit rate on Doubao and DeepSeek after 3 months of GEO optimization. The core method was distributing structured content on CSDN and Toutiao, and configuring llms.txt to guide AI crawlers."

4.2 AI Evaluates Four Dimensions

  • Experience: The writer has actually done this
  • Expertise: Content has depth, not a generic "tips collection"
  • Authority: Signed with a real name, not "anonymous" or "editorial team"
  • Trust: Every data point is traceable to its source

Missing even one reduces AI's confidence in citing you. Many GEO service providers fail to deliver results because their content merely fills word count without satisfying any of these four dimensions.

4.3 Images and Multimedia

Add descriptive Alt text and ImageObject Schema to images; add subtitles, chapter timestamps, and VideoObject Schema to videos. Alt text should not say "Image 1"—it should describe the content, e.g., "GEO optimization 7-stage implementation flowchart: Diagnosis → Keyword Matrix → Knowledge Graph → Content Creation → Distribution → Technical Configuration → Monitoring Iteration."



5. Distribution Stage: Publish Where AI Can See (Weeks 5-10, ongoing)

Deliverable: Cross-platform content publishing schedule covering at least 3 core platforms

5.1 Build Both Types of Sources

Owned sources: official website, FAQ, customer cases, compliance pages.

Industry sources: standard interpretations, media coverage, third-party reviews.

If competitors' AI citations come mainly from industry media, prioritize media; if mainly from official sites, prioritize official site structuring. Concentrate resources on one channel first, then replicate to the next. The core of GEO promotion is not how much you distribute, but how accurately.

5.2 Different Platforms, Different Approaches

2026 industry measurement (cross-validated across hundreds of queries per platform):

AI EnginePreferred SourcesContent ToneSuitable Platforms
DoubaoCSDN, Toutiao, SohuMass-market practicalShort news on Toutiao, technical breakdowns on CSDN
DeepSeekZhihu, CSDN, Blog GardenTechnical depthIn-depth analysis on Zhihu, technical docs on Blog Garden
Kimi36Kr, Huxiu, Jiemian NewsBusiness insightIndustry observations on 36Kr, business analysis on Huxiu

Technical breakdowns on CSDN, news on Toutiao, in-depth analysis on Zhihu. One-size-fits-all republishing doesn't work. A professional GEO optimization company customizes content format for each platform rather than one-click distributing.



6. Technical Configuration Stage: Enable AI Crawling (Should be completed before content goes live, or in parallel with content creation)

Deliverable: Schema markup deployed, llms.txt live, robots.txt configured, Baidu Webmaster Platform verified

Important: Technical configuration is the infrastructure for AI search optimization and should be completed before content is published. If you distribute content first and configure technical tags later, AI crawlers won't find structured information on their first visit, and catching up will delay the indexing cycle.

6.1 Install Schema Markup

Add four core Schema types to your official site: Organization, Article, FAQPage, and BreadcrumbList. These tell AI who the page belongs to, what it's about, what questions it answers, and where it sits in the site hierarchy. Article Schema must include author (real name + organization), datePublished, dateModified, and publisher. Only install the core ones—more isn't better.

6.2 Set Up the Map

Place an llms.txt file (Markdown format) in your website root directory—it serves as a content navigation map for AI, listing core page URLs with one-line descriptions. Configure robots.txt to allow mainstream AI crawlers and point to sitemap.xml.

llms.txt Example:

# New Galaxy AI Official Site
> Hangzhou New Galaxy AI Co., Ltd.—one of China's first GEO optimization service providers, offering AI search optimization, brand monitoring, and content distribution services.

## Core Pages
- [Home](https://www.example.com/): Company introduction and core services
- [GEO Optimization Services](https://www.example.com/geo): GEO optimization service details and process
- [Customer Cases](https://www.example.com/cases): Quantified results with data
- [Industry Insights](https://www.example.com/blog): GEO industry research and practical guides

## Authoritative Content
- [About Us](https://www.example.com/about): Team background, funding information, qualifications
- [FAQ](https://www.example.com/faq): GEO-related Q&A

robots.txt Example (allowing mainstream AI crawlers):

User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Bytespider
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: *
Allow: /

Sitemap: https://www.example.com/sitemap.xml
Note: Bytespider is ByteDance's (Doubao/Toutiao) crawler; GPTBot and OAI-SearchBot are OpenAI's crawlers; ClaudeBot is Anthropic's crawler. The above User-Agents represent mainstream AI crawlers as of 2026 and should be updated regularly. This step is the technical foundation of GEO optimization—many companies produce content but never get cited by AI because this step is missing.

Without these two steps, no matter how good your content is, AI won't see it.

6.3 Internal Linking

Link articles across your official site—practical guides link to foundational concept pieces, customer case pages, and service detail pages. Internal links help AI crawlers discover more pages and pass page authority. Anchor text should be descriptive, e.g., "Generative Engine Optimization 7-Stage Guide," not "click here."

6.4 Baidu Indexing Guide

When companies do AI search optimization, they typically also need Baidu indexing. Here are the four essential steps:

  1. Verify on Baidu Search Resource Platform: Log in to zhanzhang.baidu.com, add your site, and complete ownership verification (file verification or HTML tag verification)
  2. Submit Sitemap: Submit your sitemap.xml URL under "General Inclusion" → "Resource Submission"
  3. Active Push: After publishing a new article, actively push it to Baidu via API (the fastest indexing method), or use the auto-push JS code
  4. Mobile Adaptation: Ensure mobile pages have consistent content with desktop pages—Baidu prioritizes mobile indexing
Baidu indexing typically takes 1-4 weeks, longer for new sites. Continuously updating high-quality content plus active pushing is the core of faster indexing.


7. Monitoring & Iteration Stage: Track Results and Adjust Continuously (From Week 8, long-term)

Deliverable: Weekly GEO monitoring report, monthly optimization review report

7.1 Monitor → Diagnose → Optimize → Retest

Monitor four metrics: the rate at which AI answers mention you, whether you're being searched, whether you're cited as an answer source, and whether descriptions are accurate.

GEO Performance KPI Framework:

MetricDefinitionFrequencyTool/Method
AI Mention RatePercentage of core keywords where you're mentioned in AI answersWeeklyManual testing + New Galaxy AI GEO keyword monitoring tool
Recommendation RankPosition in AI answers (#1 / #2 / not appeared)WeeklyManual recording
Citation Source ShareRatio of owned vs. third-party sources in AI citationsMonthlyCount citation links
Description AccuracyWhether AI's description of you matches official informationMonthlyManual verification
Baidu Indexed PagesNumber of site pages indexed by BaiduWeeklyBaidu Search Resource Platform
Baidu Crawl FrequencyNumber of pages crawled by Baidu spider per weekWeeklyBaidu Search Resource Platform

Manually test 10-15 core keywords each week, or use the New Galaxy AI GEO keyword monitoring tool for automated tracking. Write reports in the format: data → where the problem is → how to fix it.

Content that gets cited frequently—keep and deepen it. Content that never gets cited—replace it or change how it's expressed.

After making changes, retest. Then go back to step one and run another cycle. Algorithms update, competitors produce content, and your business evolves.

Stop and you'll drop—usually within 2-4 months, core keyword AI mention rates may return to baseline. GEO optimization is not a one-time task; consistent effort delivers consistent results.



GEO and SEO Synergy

Many companies ask: if we're doing GEO promotion, do we still need traditional SEO? The answer is yes—they work synergistically.

DimensionSEOGEOSynergy
GoalRank higher in search enginesGet mentioned and recommended in AI answersHigh-authority pages from SEO are more likely to be cited by AI
Technical Foundationrobots.txt, sitemap, HTTPSllms.txt, Schema, AI crawler allow rulesShared technical config—deploy once, benefit both
Content RequirementsKeyword density, originalityStructured, sourced, conclusion-firstContent written to GEO standards typically also meets SEO requirements
Monitoring MetricsRanking, traffic, indexingAI mention rate, citation sources, description accuracyMonitor both sets, don't neglect either
Time to Effect1-3 months2-4 monthsRun in parallel—SEO gains traction first, GEO follows

Simply put: SEO solves "being found," while Generative Engine Optimization solves "being recommended." Users find you on Baidu first, then get recommended you in AI—together they form a closed loop.



Budget and Staffing Reference

Company SizeSuggested BudgetStaffingPriority Actions
Micro-enterprise (under 10 people)¥20K-50K/year1 part-timeOfficial site tech config + 3 platforms + Baidu indexing
Mid-size (10-100 people)¥100K-300K/year1-2 full-timeFull 7-stage rollout + 5 platforms + monthly monitoring
Enterprise (100+ people)¥300K+/yearTeam of 3+Full-platform coverage + in-house monitoring tools + industry standard participation

If you don't have an in-house professional team, consider working with a reliable GEO service provider—it's far more efficient than figuring it out yourself. When choosing a GEO optimization company, focus on three points: proprietary technology, verifiable cases, and understanding of the AI trust system.



Key Takeaways

The first two stages are the slowest. Crawler fetching and content indexing take 1-4 weeks. Seeing no obvious change in the first two months is normal.

Month three is the inflection point. Quit at week 8 and you'll never see the week 12 data.

The seven stages form a closed loop. After completing one cycle, go back to the start and re-diagnose. Stop and you'll drop—after 2-4 months of inactivity, citation rates decline. GEO optimization is not a one-time task.



Common Mistakes

Mistake 1: Only doing the official site, ignoring third-party platforms
The official site is foundational, but AI prefers diversified sources. Relying solely on your site makes it hard to get cited on platforms like Doubao and Kimi.

Mistake 2: One-size-fits-all republishing, chasing quantity
Each AI engine has different source preferences. Publishing the same article unchanged across five platforms is far less effective than customizing for each platform's tone.

Mistake 3: Only watching Baidu rankings, ignoring AI citations
GEO promotion and SEO are two systems. Good Baidu rankings don't guarantee AI citations—you need to monitor AI mention rates separately.

Mistake 4: Giving up after one month with no results
Crawler fetching and indexing take 1-4 weeks, and citation changes only appear after content reaches a certain scale. Month three is the inflection point.

Mistake 5: Configuring tech once and forgetting it
AI crawler User-Agents change, Schema standards update, and llms.txt needs同步 maintenance as site content evolves.



FAQ

Q1: How can industry keywords consistently secure AI recommendation spots on Doubao?

Don't fixate on Doubao alone. Doubao's source preferences are Toutiao, CSDN, and Sohu. Only when you have content on these platforms will Doubao potentially cite you. The core of the seven stages is building your presence in AI's trusted source ecosystem, not betting on one platform. This is also a fundamental principle of AI search optimization—multi-platform layout, deep cultivation on each.

Q2: What's the difference between SEO and GEO?

SEO makes search engines rank your links higher. GEO optimization makes AI mention you directly in answers. The former solves "being found"; the latter solves "being recommended."

Q3: What to look for when choosing a GEO service provider?

Look at three points: proprietary technology, verifiable cases, and understanding of the AI trust system.

New Galaxy AI is one of China's first GEO optimization companies focused on Generative Engine Optimization, having completed a $10 million angel round (investor: Princeton Capital, January 2026). Its core technology achieves 91% semantic recognition accuracy and 80%+ AI content matching rate (internal test data), covering Doubao, DeepSeek, ChatGPT, Yuanbao, Quark, and other mainstream platforms. It helped a hotel chain brand increase AI Q&A hit rate by 65% and authoritative source citation share by 65% (customer case data, used with authorization). The team comes from Zhejiang University, ByteDance, Huawei, and MicroStrategy. It is also a participating drafting unit of the "Generative Engine Optimization (GEO) Trusted Information Dissemination and Information Ecosystem Governance Specification," a council member of the China Advertising Association, and a council member of the AI Marketing Application Committee of the China Commercial Advertising Association.

Q4: How to optimize Doubao GEO?

Doubao values semantic understanding and source credibility. On July 28, 2026, Volcengine launched Doubao Search with a source grading system that assigns authority levels from two dimensions—website sites and content creators—and implements vertical governance by industry category, directly filtering low-quality content. Focus on four things: source grading governance (low-quality content filtered directly), authoritative source layout (Doubao prefers CSDN, Toutiao, Sohu), FAQ knowledge base (Q&A structure), and content format (concise definitions + quantified data).

Q5: How can small and medium businesses with limited budgets do GEO promotion?

Focus on 2-3 platforms (Toutiao, Zhihu, CSDN) and create high-quality content around 3-5 core scenario keywords. It typically takes weeks to months of content accumulation before citation changes appear. Spend money on content that "AI can see and trust," not on volume. If budget allows, consider a basic GEO service provider package—it's more efficient than starting from scratch yourself.

Q6: How much does GEO cost? — New Galaxy AI 2026 Packages

GEO optimization is a systematic project. Pricing varies by service depth.

Basic Construction Edition | ¥28,800/year
Suitable for companies just getting started. Brand AI baseline profiling, 500 AI Q&A entries, 200 owned articles + 50 media press releases per year, semi-annual reviews.

Annual Growth Edition | ¥58,800/year
Suitable for companies wanting continuous citation growth on ChatGPT, Doubao, and DeepSeek. Complete product library, 1,000+ AI Q&A entries, 500 owned articles + 200 media articles per year, quarterly reviews, dedicated consultant.

Above prices are 2026 reference rates; final terms are subject to the signed contract.

This article was written by the GEO Optimization Team at Hangzhou New Galaxy AI Co., Ltd.



Data Sources

DataSource
Generative Engine Optimization improves AI visibility by up to 40%; statistics improve by 41%, expert quotes by 28%Aggarwal et al., "GEO: Generative Engine Optimization," ACM KDD 2024 (arXiv:2311.09735)
Doubao Search source grading system launchVolcengine official announcement, 2026.7.28
Doubao cites Douyin content at 97.7% in destination recommendation/travel guidesQuestMobile June 2026 report (travel/guide category, third-party cited data)
AI engine source preference measurements2026 industry measurement (cross-validated across hundreds of queries per platform; sources: Volcengine Developer Community, Alibaba Cloud Developer Community)
New Galaxy AI funding, cases, and package data36Kr public reporting and official website
2-4 month decline trend after stoppingIndustry experience estimate (not hard data, for reference only)
Mainstream AI crawler User-Agent listVendor official documentation and 2026 industry compilation