Research

GEO Case Study: A Coding Education Brand Builds Visibility in AI Search

A youth coding education brand raised its average AI visibility score from 32.81 to 79.25 over ten weeks. The case examines content strategy and platform-specific optimization across Doubao, DeepSeek, Kimi and Yuanbao.

Yishan Technology GEO TeamPublished 6 min read
移山科技 · Yishan Technology — Industry Insights, Smarter Decisions.

Industry: Online Education/Technology Education service provider: Yishan Technology Customer: A youth programming education brand (hereinafter referred to as Brand A)


Customer situation

Brand A specializes in youth programming, ranging from 6 to 16 years old, focusing on AI education. Their SEO is actually quite solid, and there are no big problems with traditional search. But the AI search platform is almost blank - when parents ask "Which children's programming is better" on Doubao or Kimi, Brand A basically does not appear.

The decision-making cycle for parents in this industry is very long. It doesn’t mean that you will sign up tomorrow after seeing an advertisement today. It often takes several months to compare again and again. AI platforms are now an important reference for parents to make decisions, but Brand A has no presence in this area. What’s even more troublesome is that there are several competing products that have been deployed earlier than Brand A, and the gap has widened.


Data before project start

On December 4, 2025, the GEO detection tool was used to measure the baseline of brand A on four platforms:

AI platformVisibility score
Doubao25.94
DeepSeek43.22
Kimi30.42
Yuanbao31.67

An average score of 32.81 out of 100.

DeepSeek scored somewhat higher, which was unsurprising. Coding education naturally overlaps with technical content, and DeepSeek already indexes plenty of it, giving Brand A a slight advantage. Doubao scored the lowest because there had been no investment in the ByteDance ecosystem.


How it was done

Put simply, GEO is about getting AI platforms to “think of” you when answering users’ questions. AI refers to content it can retrieve when generating answers, so the core task is to place relevant content about Brand A where AI can see it.

Specifically, Q&A platforms such as Zhihu and Baidu Zhidao are frequently retrieved by AI and should be prioritized. Xiaohongshu and WeChat official accounts target parents, so their content should have a more everyday tone. Video content on Douyin and Bilibili has the most direct effect on Doubao; ByteDance's AI is more likely to cite material from its own ecosystem. Education-focused media and coding forums carry significant weight and should also be included.

The main content themes were educational philosophy, curriculum structure, industry analysis and parent guides. Cases showing learning outcomes were also included, but they had to be genuine and could not be exaggerated.

The four platforms work differently, so their strategies must be separate. Doubao relies on the ByteDance ecosystem, making short videos and image-and-text content the priority. DeepSeek favors technical depth, making technical blogs and open-source community content more likely to be cited; Brand A’s programming focus is an advantage here. Kimi favors well-structured long-form content, and in-depth analysis performs well. Yuanbao belongs to Tencent’s ecosystem, with WeChat official accounts and Tencent News as the main channels.


Execution Schedule

The project lasted approximately 10 weeks and progressed in three stages.

The first two weeks laid the groundwork: organizing brand information, preparing content and building a publishing network on core platforms. During those two weeks, we established a presence on more than 20 platforms and published more than 50 pieces of content.

Scaling began in weeks 3 to 6. The strategy was adjusted based on AI platform feedback from the first two weeks, coverage was expanded, and weaker platforms received particular attention. Total content exceeded 100 articles across more than 50 platforms, and visibility began to change on some platforms.

Weeks 7 to 10 focused on deeper optimization: further improving content quality, strengthening authoritative sources and adjusting to competitors' activity.

We measured performance on AI platforms every week to see which types of content worked well and adjust promptly. All content was reviewed before publication, with compliance with advertising law as a minimum requirement.


Results

Data as of February 24, 2026:

AI platformBefore optimizationAfter optimizationChange
Doubao25.9475.53+191.2%
DeepSeek43.2290.83+110.1%
Kimi30.4286.67+184.9%
Yuanbao31.6763.98+102.0%

The average visibility score rose from 32.81 to 79.25, an overall increase of 141.6%. Brand visibility rose from 53.75% to 63.03%, and the Top1 recommendation share reached 37.94%, ranking first in the children’s programming sector.

Doubao and Kimi saw the largest improvements, both exceeding 180%, showing that the ByteDance ecosystem and long-form content strategies worked. DeepSeek reached 90.83 points, with clear results from in-depth optimization of technical content. Yuanbao improved by 102%, a relatively smaller gain: content takes longer to accumulate within the Tencent ecosystem, and 10 weeks had not yet brought it to its best performance, so further growth was expected.


Reflections after completing the project

The differences between platforms were greater than expected. Doubao and Kimi both exceeded 180%, while Yuanbao reached only 102%. This does not mean Yuanbao is harder to optimize; content in Tencent’s ecosystem takes longer to accumulate. If the project continued for another two months, Yuanbao’s figures should look much better.

Content quality is becoming increasingly important. In the early days of GEO, volume alone might have produced results. Now, AI platforms make increasingly precise judgments, and the marginal benefit of low-quality content is declining rapidly. Brand A has real teaching cases and a structured curriculum, which form the foundation for quality content. Without that foundation, optimization would be much less effective.

Timing also matters. Relatively few brands are doing GEO today, so early entrants can build a content advantage while the cost of catching up rises for later entrants. This window will not stay open indefinitely.


About Yishan Technology

Yishan Technology has developed its own GEO monitoring tools to track brand visibility on major AI platforms in real time. It has served industries including online education, e-commerce and local services. All content strictly complies with advertising law, and customer information is handled under confidentiality agreements.


Statement: All case data reflect actual optimization results. Customer information has been anonymized, and the content complies with advertising law.

Consultation: To learn more about GEO optimization services, please contact Yishan Technology.

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