Skip to main content
案例分析

【AI Visibility Case Study】Strong SEO, Invisible to AI — 23/100

Published on March 24, 2026

Case Overview

This case involves a Taiwan-based manufacturer specializing in power supply units and thermal management solutions, serving industrial, consumer, and server markets. When we conducted an AI visibility audit for the company, we uncovered a striking contradiction that caught our attention immediately.

On traditional SEO metrics, the company's website scored an impressive 92 out of 100 — a clear sign of sustained investment in conventional search optimization. PageSpeed performance came in at 81, reflecting solid site infrastructure. Yet when we shifted the lens to GEO (Generative Engine Optimization) — the technical framework that determines how well AI search engines can read, interpret, and cite a website — the score collapsed to just 20 out of 100. The final composite AI visibility score landed at 23/100, placing the company in the "AI Visibility Potential: Undeveloped" tier.

This case is a textbook example of a growing disconnect in the search landscape: excelling at traditional SEO no longer guarantees relevance in the AI search era. The rules have changed, and the scoreboard looks very different.

Composite Score Breakdown

The overall AI visibility score is a weighted composite of three dimensions. A serious weakness in any single dimension can drag down the entire score — and with it, a brand's chances of appearing in AI-generated answers.

Evaluation Dimension Score Status
AI Brand Mention Rate 69 / 100 ⚠️ Foundational — gaps remain
GEO Technical Audit 20 / 100 🔴 Primary bottleneck
Website Performance (PageSpeed) 81 / 100 🟡 Good, but room to improve

The bottleneck is clear: a GEO technical score of 20 is the single biggest drag on the company's overall AI visibility. Even when AI platforms can identify and mention the brand in some query contexts, a weak underlying technical architecture prevents AI crawlers from reliably reading and indexing the site's content. The result is a brand whose AI citability is fragile — dependent on historical reputation rather than deliberate optimization.

AI Search Visibility Testing

To gather objective data, we ran live queries across four major AI platforms, simulating how a procurement engineer or technical buyer might search for power supply solutions. A total of 16 industry-relevant queries were executed to test whether the company appeared in AI-generated responses.

Claude

Claude delivered the strongest and most consistent performance of the four platforms tested. Across all 4 queries, the company received a positive mention — a 100% mention rate. This indicates that Anthropic's training data contains sufficient brand and product information to recognize the company as a relevant recommendation within the power supply category. This is an advantage worth protecting and building on.

ChatGPT

OpenAI's ChatGPT also achieved a 100% mention rate, with the company referenced positively in all 4 queries. This reflects a meaningful digital footprint across English and Chinese technical communities, review sites, and industry forums — content that has been captured in GPT's training data. However, being mentioned is not the same as being prioritized. The absence of structured data may limit how prominently the company surfaces in competitive recommendation scenarios.

Gemini

Google's Gemini showed a notable performance gap. Out of 4 queries, the company was mentioned positively in only 2, while it failed to appear at all in the other 2 — a 50% mention rate. Given that Gemini is deeply integrated with Google's search index, technical deficiencies on the company's website — such as missing Sitemaps, Schema markup, and OG Tags — likely affect Google's crawl quality directly, and that degraded crawl signal feeds through into Gemini's citation behavior. This is a clear and measurable technical-to-AI-visibility feedback loop.

Perplexity

Perplexity produced the weakest results across all four platforms. Only 1 out of 4 queries returned a positive mention — a 75% non-mention rate. Since Perplexity relies heavily on real-time web retrieval, pages that lack structured semantic markup and clear content signals are at a significant disadvantage. Without those signals, AI crawlers struggle to determine a page's relevance and authority, making citation unlikely.

Across all 16 queries combined, the company was mentioned in 11 instances — an overall AI visibility mention rate of approximately 69%. While this provides a baseline foundation, the severe inconsistency across platforms reveals that the brand's AI presence is largely the product of accumulated reputation rather than intentional optimization. That's a fragile position to be in as competitors begin investing in GEO.

Competitive Landscape in AI Search Results

Within the AI-generated responses where the company appeared, it was consistently listed alongside brands such as Seasonic, Corsair, EVGA, Super Flower, be quiet!, FSP, and Cooler Master. These competitors generally share several common strengths: broader multilingual content ecosystems, active third-party review coverage, and more complete structured data deployment across their websites.

A consistent pattern emerged in how AI platforms approach power supply recommendations: they tend to prioritize and cite brand pages that include specific technical specifications, certification data, and third-party validation. When the company's website cannot clearly communicate product specs, application contexts, and certifications to AI crawlers, there is a real risk that even a brand with solid market presence gradually gets pushed to the periphery of AI recommendation lists — not because the products are inferior, but because the content architecture fails to communicate their value in AI-readable terms.

This competitive dynamic makes the urgency of improving AI visibility very concrete: the window for first-mover advantage is open, but it won't stay open indefinitely.

GEO Technical Audit Results

The GEO technical audit measures how "readable" a website is to AI crawlers — the foundation of sustainable AI visibility. This audit covered 9 critical technical checkpoints. The company's website passed only 2 out of 9 (approximately 22%), revealing significant infrastructure gaps.

Technical Checkpoint Status
Schema JSON-LD Structured Data ✗ Not implemented
XML Sitemap ✗ Not implemented
Meta Description ✗ Not implemented
OG Tags (Open Graph) ✗ Not implemented
Canonical URL ✗ Not implemented
HTTP/2 Protocol ✓ Enabled
Title Tag ✗ Not implemented
H1 Heading Tag ✗ Not implemented
PageSpeed Performance Threshold ✓ Pass

The most critical failures cluster at the semantic layer. The absence of Schema JSON-LD means AI crawlers cannot structurally identify what product categories, specifications, or brand entities are present on the page — they are left to guess. Missing H1 and Title Tags strip each page of its primary topical signal. A missing Sitemap means crawlers have no systematic roadmap to discover all pages on the site.

These may appear to be basic technical items, but in the GEO era, they are the fundamental building blocks of AI visibility. Without them, even a brand with strong market recognition becomes structurally invisible to large portions of the AI search ecosystem.

Website Performance Analysis

Website performance functions as a secondary — but meaningful — influence on AI visibility. The company's website achieved a composite PageSpeed score of 81, which is generally solid. The SEO sub-score reached 92, confirming a strong foundation in conventional search optimization practices.

However, the Performance sub-score came in at 69, indicating room for improvement in Core Web Vitals metrics including LCP (Largest Contentful Paint), FID (First Input Delay), and CLS (Cumulative Layout Shift). Image loading optimization and server response time are the most likely areas for improvement.

The relevance to AI visibility is indirect but real: AI crawlers factor technical quality signals into their content evaluation processes. Faster load times and more stable rendering contribute to a stronger quality signal, which can reinforce how AI models assess the reliability and authority of a website's content over time. Given that the company's website carries valuable technical content — product specification sheets, thermal performance curves, certification data — ensuring that content is delivered quickly and cleanly is not just a UX concern; it's an AI visibility concern.

Expert Diagnostic Recommendations

Based on the audit data, we identified three structural issues that are most directly suppressing the company's AI visibility. Below is a summary of our diagnostic findings.

Recommendation 1: The GEO Technical Foundation Is Nearly Absent

Only 2 of 9 technical checkpoints passed — and both of those are passive infrastructure items (HTTP/2 and page speed). None of the active semantic signals that tell AI crawlers "who we are and what we offer" are in place. Schema markup, Sitemap, H1 tags, and Title Tags are all missing. In practical terms, this means every time an AI crawler visits the company's website, it must essentially guess at the page's content and relevance — dramatically reducing the probability of accurate citation.

Closing this gap requires more than checking technical boxes. It calls for a semantic architecture strategy built around AI readability as a core design principle.

Recommendation 2: AI Mention Rates Are Too Dependent on Luck

The current 69% AI brand mention rate is largely a reflection of historical brand equity — the accumulated digital footprint the company has built over the years. This is not a sustainable competitive position. As rivals begin deploying deliberate GEO strategies, that passive advantage will erode. The Perplexity result — a 25% mention rate on a platform that relies on live web retrieval — is an early warning signal that AI platforms are already beginning to lose sight of the company in real-time search contexts.

Recommendation 3: The Hidden Performance Contradiction Needs Attention

An overall PageSpeed score of 81 looks acceptable at a glance, but the gap between the Performance sub-score (69) and the SEO sub-score (92) points to a specific problem: the company's visual and technical assets — product spec images, thermal performance charts, certification graphics — are likely not optimized for AI crawler processing efficiency. In the power supply and thermal solutions industry, these assets are among the most persuasive pieces of technical evidence a brand can offer. If AI crawlers cannot quickly parse and process them, the company is effectively leaving its most valuable content on the table.

AI Search Trends in the Power Supply and Thermal Solutions Industry

Purchasing decisions in the power supply and thermal management space have never been impulsive. Whether it's a systems integrator comparing products before issuing a quote, a data center engineer designing a cooling architecture, or an ODM/OEM procurement team searching for Taiwan-based supply chain partners — the entire buying process is deeply reliant on technical specification cross-referencing and credibility assessment.

Traditionally, these buyers would search Google for terms like "80 PLUS certified power supply comparison" or "1U rack thermal solution Taiwan manufacturer," then spend time browsing through dozens of individual pages. But as AI search adoption accelerates, the first touchpoint in the buying journey is shifting rapidly. Engineers and procurement professionals are increasingly turning directly to ChatGPT or Perplexity with questions like: "Which Taiwan power supply manufacturers offer industrial-grade solutions?" or "What are the best thermal management options for high-density server environments?"

This behavioral shift creates an entirely different set of requirements for AI visibility. Traditional SEO optimizes for keyword rankings. AI search optimization is about citability — the ability of an AI model to clearly identify your product categories, application scenarios, technical specifications, and market positioning, and then accurately surface your solutions to prospective buyers in a conversational context.

There are several AI citation opportunity windows specific to this industry that deserve strategic attention. First, product certification data — 80 PLUS ratings, EMC certification, UL/CE compliance — are strong credibility signals that AI models use when evaluating which brands to recommend. Without structured markup, these credentials cannot be reliably extracted and cited. Second, thermal solution use cases are highly vertical; AI models handling specialized queries about edge computing node cooling or high-wattage GPU server heat management will preferentially cite pages with explicit scenario-specific descriptions. Third, Taiwan manufacturers hold meaningful differentiation in global supply chains — ODM capability, customization flexibility, lead time advantages — but if these narratives exist only in PDF catalogs or sales presentations, AI crawlers have no way to access them.

At this moment, the majority of Taiwan-based power supply manufacturers have yet to invest meaningfully in AI visibility infrastructure. For companies willing to act now, this represents a genuine first-mover window — an opportunity to establish a durable competitive position in AI search before the rest of the field catches up.

Disclaimer

This article is based on anonymized audit data from an actual AI visibility assessment. All information that could identify the specific company has been removed. AI platform responses are non-deterministic; results may vary across different query sessions and time periods. Technical audit scores and performance metrics reflect a point-in-time snapshot.

Find Out Where Your Business Stands in AI Search

AI visibility gaps are rarely discovered until a competitor has already closed them. We offer two ways to help you understand your current position quickly:

🔍 Use our free AI Visibility Self-Assessment Tool to get an initial diagnostic report in under 3 minutes.

📋 Or schedule a free results consultation with one of our GEO strategists, who can walk you through what the data means for your competitive position and outline a concrete optimization roadmap tailored to your industry.

For more industry case studies and GEO optimization insights, visit the Joseph Intelligence Case Study Index — continuously updated with AI visibility analysis across industries.

FAQ

Why does a high SEO score not guarantee good AI visibility?
Traditional SEO and AI visibility (GEO) measure very different things. SEO scores reflect factors like keyword optimization, backlink profiles, and page speed — all relevant to ranking in conventional search engines. AI visibility, on the other hand, depends on whether AI crawlers can semantically interpret your content: structured data markup, clear entity signals, descriptive metadata, and well-defined page hierarchy. A site can score 92 on SEO while scoring 20 on GEO if it lacks the semantic infrastructure that AI models rely on to understand and cite content accurately.
What is GEO, and how does it differ from traditional SEO?
GEO stands for Generative Engine Optimization — the practice of structuring website content so that AI-powered search engines like ChatGPT, Gemini, Claude, and Perplexity can accurately read, interpret, and cite it. While traditional SEO focuses on ranking signals for keyword-based search, GEO focuses on AI citability: helping AI models understand who you are, what you offer, and why you are a credible source. Key GEO elements include Schema JSON-LD structured data, clear H1/Title Tags, XML Sitemaps, and scenario-specific content that matches how AI models process queries.
Why did the company perform so differently across Claude, ChatGPT, Gemini, and Perplexity?
Each AI platform sources and processes information differently. Claude and ChatGPT rely primarily on training data, which captured the company's brand presence through historical web content. Gemini integrates closely with Google's live search index, so gaps in the company's website technical setup — missing Sitemap, Schema, and metadata — directly degraded Google's crawl quality, reducing Gemini's ability to cite the company. Perplexity uses real-time web retrieval, making it the most sensitive to current on-site technical quality. The result was a 100% mention rate on Claude and ChatGPT, 50% on Gemini, and just 25% on Perplexity.
How can power supply and thermal solution manufacturers improve their AI visibility?
The highest-impact improvements for manufacturers in this space start with the GEO technical foundation: implementing Schema JSON-LD to identify product types, specifications, and certifications; adding XML Sitemaps so crawlers can discover all pages; and ensuring every page has proper Title Tags and H1 headings. Beyond technical fixes, content strategy matters: product certification data (80 PLUS ratings, UL/CE compliance), specific application scenarios (server cooling, industrial power systems), and supply chain differentiators (ODM capability, customization flexibility) should all be published as structured, crawlable web content — not locked inside PDFs or slide decks.
How quickly can AI visibility improvements take effect after implementing GEO fixes?
Timeline varies depending on the AI platform and the nature of the changes. Technical fixes like Schema markup and Sitemap submission can improve AI crawler indexing within weeks, particularly for platforms like Gemini that draw on Google's live index. For training-data-dependent platforms like Claude and ChatGPT, improvements in citability may take longer to reflect, as they depend on model update cycles. That said, building a stronger GEO foundation now creates compounding benefits: better-structured content is more likely to be cited, syndicated, and referenced by third parties, which feeds back positively into future model training data.

Ready to Elevate Your Digital Marketing?

Let our AI-driven solutions help your business grow