＃65

Your Website Is Winning the Old Game. That's the Problem.

Author: Kzone Chen / KYORYX Team 
Category: Marketing Strategy / Business Growth 

Listen to this article:

A few months ago, I sat with the second-generation owner of a mid-sized precision metal parts manufacturer in central Taiwan. On his desk were two documents: a freshly issued carbon fee invoice under Taiwan's new 2026 levy, and an export report showing the damage from U.S. reciprocal tariffs. He'd spent a six-figure sum the year before on a digital agency that stuffed his website with keywords like "precision fasteners" and "Taiwan manufacturing upgrade." The result, months later: not one new international inquiry call, and the click-through rate on a page that used to rank #1 on Google had fallen by more than half.

He wasn't a naive buyer. He'd done everything the old playbook told him to do. The problem wasn't his effort — it was the map he was using. The terrain had already changed underneath him.

If you run a traditional manufacturing business in Taiwan — fasteners, machine tools, molds, metal components, industrial materials — there's a good chance you're standing exactly where he was: you know digital transformation matters, you've probably already spent money on it, and you're still not sure why it isn't converting into orders. This piece is my attempt to explain, plainly and without the jargon, what actually changed in how buyers find suppliers — and to give you a sequence you can act on without hiring an army of consultants.

---

1. The Two-Speed Economy Is Not in Your Head

Before we talk about websites, it's worth naming the economic backdrop, because it explains why this matters now rather than in some comfortable future.

Taiwan is currently running two economies at two different speeds. One is roaring. The other is being quietly squeezed.

The Numbers Behind the Divide

• Taiwan's overall GDP growth reached roughly 7.4% in 2025, driven almost entirely by semiconductors and AI server exports.
• In the first quarter of 2026, real GDP growth spiked to around 13.7% — the fastest pace in nearly four decades.
• That boom is concentrated almost entirely in the AI and high-tech supply chain. It has not reached most traditional manufacturers.

Four Pressures Compounding at Once

If you're in petrochemicals, metal fasteners, mid-to-low-end machine tools, or similar sectors, you are likely absorbing several of these simultaneously:

• Tariff exposure without protection. China's suspension of roughly 134 ECFA tariff concessions removed a longstanding cushion, and most traditional goods were never covered by exemption frameworks like the ITA in the first place.
• U.S. reciprocal tariffs eating into margin. Taiwanese machine tool exporters have reported export declines of around 30% to the U.S. as a direct result.
• A currency working against you. The tech boom has pushed the New Taiwan Dollar higher, which quietly erodes the competitiveness of every USD-denominated quote you send abroad.
• The first year of carbon fees. 2026 marks Taiwan's first year of carbon fee collection, at a base rate of roughly NT$300 per tonne — adding real cost pressure to decarbonize supply chains that were never built with carbon accounting in mind.

Here's the number that should really get your attention: of Taiwan's 1.716 million SMEs, only about 7.4% have meaningfully adopted AI in their operations. Roughly 70% are still using only basic digital tools. That gap is not a reason to feel behind — it's the actual opportunity. When almost nobody in your category has moved, the ones who move first don't need to outspend giants. They just need to move correctly.

---

2. Why the SEO Playbook You Paid For Stopped Working

This is the part most business owners get wrong, and it's not their fault — the ground genuinely shifted underneath the entire industry within the last two to three years.

From Search Engines to Answer Engines

For twenty years, the model was simple: a buyer types a keyword into Google, sees a page of blue links, clicks one, and lands on your website. Your job was to rank high enough to get clicked.

That model is breaking down. Google AI Overviews, ChatGPT, Perplexity, and Gemini now increasingly answer the question directly, pulling fragments from across the web, cross-referencing multiple sources, and synthesizing a complete answer with citations — without the buyer ever leaving the chat window.

Picture a procurement manager asking an AI tool: "Who makes ISO 13485-compliant, high-strength stainless steel fasteners?" The AI doesn't hand them ten blue links to evaluate. It gives them a synthesized shortlist — and if your company isn't in the source material it drew from, you were never in the running. You didn't lose a competitive bid. You never got invited to bid at all.

The Zero-Click Reality, in Numbers

• Traditional search already produced a "zero-click" rate — searches where the user never clicks through — of roughly 34% to 60%.
• In the AI search era, that figure has climbed to somewhere between 80% and 93%. The overwhelming majority of buyers now get their answer without ever visiting a website.
• The click-through rate for the #1 organic Google result has fallen from around 7.3% to roughly 1.6%.

The practical translation: if your company isn't showing up inside the AI's synthesized answer, it's the modern equivalent of being buried on page five of Google — except now there often isn't a "page two" for the buyer to scroll to. This is why two new metrics matter more than your old keyword rankings: Share of Model (SoM) — how often AI models recommend you — and AI Share of Voice (AI SOV) — how consistently your brand is mentioned across AI-generated answers in your category.

---

3. The Genuinely Good News: This Actually Favors You

Here's where I want to change the tone, because most of what gets written about AI search is designed to scare business owners into buying expensive software. The underlying research says something more useful, and more hopeful, for a company your size.

The Equalizer Effect

Researchers at Princeton and other institutions, publishing at KDD 2024, studied what's now called Generative Engine Optimization (GEO) — the practice of structuring content so AI systems are more likely to cite it. Their most important finding wasn't a tactic. It was a pattern:

Websites that ranked lower on traditional Google saw their AI visibility improve by as much as 115% after applying GEO principles. Websites that were already top authorities saw almost no additional gain.

Read that twice. The system that used to reward twenty years of domain authority and marketing budget is, structurally, far more forgiving to a company that has neither. AI systems selecting the best answer weigh factual density, data precision, and structural clarity more heavily than raw domain reputation. A well-organized, fact-dense page from a 40-person fastener manufacturer can out-cite a page from a company fifty times its size — if the content is built the right way.

Five Moves That Actually Move the Needle

The same research isolated which content changes produced measurable gains in AI citation rates, and which didn't. Vague, persuasive marketing language and keyword stuffing actively hurt visibility. What helped:

Optimization Tactic	Measured Lift in AI Citation	What It Looks Like in Practice
Statistical data injection	+37% to +41%	Replace "excellent durability" with "maintains zero stress-corrosion cracking after 10,000 hours at 250°C"
Cited, authoritative sources	+30% to +40%	Add clear attribution and external references, not just claims
Direct expert quotations	+28% to +40%	Include a named engineering lead or third-party researcher, quoted directly
Improved fluency and clarity	+15% to +30%	Shorten dense sentences; make each claim easy to extract cleanly
Objective, encyclopedic tone	+10% to +20%	Write like a technical reference, not a sales brochure

Notice what's absent from that list: budget, agency size, and years in business. This is a content-craft problem before it's a technology problem — which is exactly why it's solvable by a company your size, in-house, without a large team.

---

4. A Four-Phase Roadmap You Can Actually Run

I generally advise clients against trying to overhaul everything at once. Pick 10 to 15 of your highest-value pages — the ones tied to your actual revenue-generating product lines — and work through these four phases in order.

Phase 1: Open the Door for AI Crawlers

Nothing else in this roadmap matters if AI systems physically cannot read your site.

• Check your robots.txt file and confirm it isn't blocking GPTBot, ClaudeBot, PerplexityBot, or other major AI crawlers.
• If your site runs on client-side JavaScript rendering, understand that AI crawlers strongly prefer server-side rendering with a first-byte response time under 200 milliseconds — heavy, slow-loading pages get skipped.
• Add structured JSON-LD markup (Organization, LocalBusiness, FAQPage schemas) so machines can parse who you are and what you offer without guessing.

Common obstacle: older template-based websites that don't allow direct code edits. If that's you, this may be the trigger to finally address the underlying platform, not just the content on it.

What success looks like: AI crawler visit logs return to normal, and your schema markup passes validation testing.

Phase 2: Rewrite for Answers, Not Applause

This is where most manufacturers' existing copy actively works against them. Traditional B2B copy builds up to a pitch. AI systems want the conclusion first.

• Use an inverted-pyramid structure: deliver a direct, 40–60 word answer within the first 200 words of the page.
• Apply a "definition-first" sentence pattern immediately after every subheading: "[Subject] is a [category] that [core differentiator or use case]." For example: "A five-axis machining center is a high-precision subtractive manufacturing system used for complex surface machining in aerospace and medical device components."
• For product specs and competitive comparisons, use structured HTML tables rather than paragraphs of prose.

Common obstacle: your sales team, who often drives copy decisions, is used to a more persuasive style and may resist writing in this plainer, more clinical tone.

What success looks like: a measurable increase in the fact-density of your core pages, particularly in the first 30% of each page — the section AI systems weight most heavily when extracting content.

Phase 3: Build Authority Beyond Your Own Website

Roughly 85% of the brand recommendations AI engines make are drawn from sources outside your own website. This phase solves what's known in GEO research as the "ghost citation" problem — where AI knows the general facts but never attaches your name to them.

• Actively pursue genuine brand mentions in local industry forums, trade association websites, and partner sites — not link farms, but real third-party recognition.
• Publish an llms.txt file in your site's root directory — a plain Markdown file that gives AI agents a structural map of your site, reducing the chance they hallucinate incorrect details about you.
• Regularly publish original data: internal R&D findings, first-hand test results, or documented progress on decarbonization — anything that didn't already exist somewhere else on the internet.

Common obstacle: earning genuine external media or industry mentions takes longer and requires real relationship-building, not a purchase order.

What success looks like: when AI tools are asked to recommend suppliers in your category, your brand's mention rate and citation frequency rise noticeably.

Phase 4: Monitor and Adjust

You cannot manage what you don't measure — and unlike traditional SEO rank tracking, this requires actively querying the AI tools yourself.

• Compile 15–20 real prompts that mirror how a buyer actually asks questions at each stage of their decision process.
• Run those prompts weekly against ChatGPT, Gemini, and Perplexity, and log when and how your brand appears.
• In Google Analytics 4, track referral traffic originating from chatgpt.com and perplexity.ai — this is the "AI dark traffic" that's easy to miss if you're only watching organic search numbers.

Common obstacle: most companies at this stage lack a systematic tracking process and rely on ad hoc, occasional checks.

What success looks like: your AI Share of Voice climbs steadily over time, and qualified inbound inquiries increasingly cite "I asked ChatGPT" or "I found you through an AI search" as their origin story.

---

5. Three Mistakes That Will Quietly Sabotage Everything Above

I've watched capable, well-intentioned manufacturers undo months of good work by falling into one of these three traps.

• Keyword stuffing, rebranded. Cramming terms like "machine tools" or "stainless steel" repeatedly into page footers doesn't fool AI systems the way it once fooled early search algorithms. Large language models read for semantic meaning, and repetitive, meaningless text is flagged as low-quality — actively reducing your visibility rather than helping it.
• Content with zero informational increment. Rewriting or lightly paraphrasing information that already exists widely online gives the AI nothing new to attach your name to. It gets absorbed as generic background knowledge, and your brand gets no credit — the "ghost citation" problem again, this time self-inflicted.
• Ignoring your footprint beyond your own domain. Focusing exclusively on your own website while neglecting your Google Business Profile, industry directories, and consistent brand mentions on partner sites prevents AI systems from cross-verifying that your company is a real, established entity — and inconsistent or absent third-party signals can knock you out of consideration entirely.

---

6. Your Self-Assessment: Is Your Website Ready to Be an AI's Answer?

Before you spend another dollar, spend fifteen minutes running through this checklist honestly.

#	What to Check	Why It Matters	Priority
1	Does robots.txt allow GPTBot and ClaudeBot to crawl your site?	If crawlers are blocked, nothing else in this list matters	P0 — Highest
2	Is your server's first-byte response time under 200ms?	Slow, JavaScript-heavy pages are frequently skipped by AI crawlers	P0 — Highest
3	Does the first 200 words of each key page contain a 40–50 word direct summary?	This is the section AI systems weight most when extracting answers	P0 — Highest
4	Do your product pages include specific data, dates, and sourced figures?	Concrete statistics are cited far more often than generic claims	P1 — High
5	Is JSON-LD structured markup deployed in your page headers?	Removes ambiguity about who you are and what you make	P1 — High
6	Does your site have an llms.txt file at the root directory?	Gives AI systems a clean, direct map instead of forcing a guess	P1 — High

Your First Move — This Week, Not Next Quarter

• Check and release your AI crawlers. Spend ten minutes reviewing robots.txt to confirm GPTBot and PerplexityBot aren't blocked. This is the single lowest-cost, highest-leverage action available to you — it costs nothing and directly determines whether anything else you do can even be seen.
• Rebuild one flagship page. Choose your single most important service or product page and rewrite its opening with a 40–50 word objective definition backed by a real statistic.
• If you bring in outside help, ask the right question. Don't evaluate an agency on whether they promise "top keyword rankings." Ask whether they understand your specific manufacturing category, whether they can actually implement RAG-aware technical changes, and whether their GEO plan includes both JSON-LD deployment and off-site authority building — not just on-page copywriting.

---

Frequently Asked Questions

Q1: How much budget does GEO actually require to get started?

Enterprise-grade GEO platforms and AI-visibility monitoring tools typically run anywhere from $99 to $5,000 a month in subscription costs. But for a resource-constrained SME, none of that is necessary at the outset. A "lightweight GEO" approach — dedicating 4 to 8 hours of internal time to robots.txt cleanup and rewriting your core pages in an answer-first structure — carries close to zero direct technical cost. The constraint isn't money. It's whether someone commits the focused time to do it properly.

Q2: How long before we actually see results?

Technical changes — fixing robots.txt, improving load speed — are typically crawled and indexed by AI engines within about 30 days. Once you've also rewritten your core pages in an answer-first structure and added original data, most companies begin to see measurable increases in AI citation and brand mention frequency within 60 to 90 days of consistent prompt testing.

Q3: Our products are highly commoditized contract manufacturing. How do we even pick an entry point?

Don't try to compete for a broad, generic term like "best stainless steel contract manufacturer" — you'll be fighting every competitor at once for a term that's expensive to win and vague enough that it rarely converts. Instead, target the narrow, high-intent question your best buyers are actually asking at the decision stage, and that most competitors haven't bothered to answer in depth. For example: "precision fastener manufacturing process analysis for EU carbon border tariff compliance with ISO certification and low process emissions." Own that specific, dense answer with real data, and you become the internet's de facto answer for that narrow niche — which is exactly where high-intent buyers are searching.

---

The Bottom Line

Taiwan's manufacturing sector is in a genuine survival window, caught between a booming tech economy it can't fully participate in and a set of external pressures — tariffs, currency, carbon costs — it didn't create. But the shift from search engines to answer engines is, unusually, a shift that favors companies exactly your size over companies with ten times your marketing budget.

The right entry point isn't a bigger spend on the old playbook. It's a smaller, more disciplined spend on the new one: open the door for AI crawlers, rewrite your best pages to answer questions directly, build genuine authority beyond your own site, and measure what's actually happening in AI search rather than guessing.

Start with the ten-minute robots.txt check this week. It costs nothing, and it's the one action that determines whether everything else you do can even be seen.

Is your company's website currently being blocked from AI crawlers — or have you hit a different wall in your digital transformation efforts? I'd genuinely like to hear where you're stuck.

Kyoryx → GaaS → Healthcare Growth → Fractional CMO → Growth Strategy → Case Studies → Evidence of Results