The rise of AI-powered search engines like ChatGPT, Perplexity, and Google’s Gemini represents a fundamental shift in how customers discover businesses online. But while companies invest heavily in traditional SEO, many remain completely invisible to these AI systems without realizing it.
The problem isn’t content quality or keywords. It’s how AI bots actually read web pages – and the answer challenges everything most businesses assume about their digital presence.
The Mythology of AI Optimization
Social media and industry forums are filled with conflicting advice about optimizing for AI search. Some consultants recommend restructuring websites with semantic HTML tags. Others suggest adding extensive schema markup. A few even propose serving content in Markdown format to “save tokens” for AI processing.
The reality is simpler and more frustrating: most of this advice is based on guesswork, not evidence.
Recent testing using controlled experiments reveals exactly how AI bots process web content – and the findings overturn many widely held assumptions about AI optimization.
What AI Bots Actually See
When ChatGPT or Perplexity visits your website, it doesn’t experience your carefully designed interface. It doesn’t see your navigation menus, your styled headings, or your semantic HTML structure. It sees plain text.
Analysis of the top 1,000 websites on the internet revealed a stark pattern: the average site sends 174,000 tokens of HTML code to deliver just 3,100 tokens of actual readable content. That’s a signal-to-noise ratio of 56:1 – fifty-six words of code for every word of value.
AI systems solve this problem by stripping everything away. They pull the text content from HTML tags and discard the rest. Your semantic classes, your carefully structured divs, your HTML5 elements – none of it matters. The bot sees a linear flow of headings, paragraphs, and lists. Nothing more.
Controlled testing confirms this behavior. Pages where critical information was placed exclusively in HTML semantic classes were tested with both ChatGPT and Gemini. Both systems failed to find the information or hallucinated answers based on the URL alone.
The Schema Markup Paradox
This next finding surprises many digital marketers: AI user bots don’t process schema.org markup.
Schema markup is structured data embedded in web pages using JSON-LD format. It’s clean, it’s machine-readable, and it should be perfect for AI systems. Major search engines like Google and Bing parse it extensively for rich results.
But AI user bots ignore it completely.
Testing placed product prices and inventory data exclusively inside JSON-LD blocks. ChatGPT couldn’t see it. Gemini couldn’t see it. In one test, Gemini hallucinated a product SKU because the real one was hidden in schema markup the bot never accessed.
Does this mean businesses should remove schema markup? No. Schema remains valuable for Google and Bing, which do use structured data. The distinction is important: traditional search crawlers that index content may process schema differently than real-time AI user bots that fetch pages on demand.
Keep your schema for Google. Just don’t rely on AI bots to read it directly.
The JavaScript Execution Problem
The third critical issue affects businesses running modern web applications: AI bots don’t execute JavaScript.
Many business websites use React, Vue, Angular, or similar frameworks. Content loads dynamically after the initial page renders. For human visitors using Chrome or Safari, this works perfectly. For AI bots, it means the content doesn’t exist.
Testing confirmed zero JavaScript execution across ChatGPT, Perplexity, and even Google’s Gemini. Despite modern User-Agent strings that suggest browser capabilities, these bots function as simple HTTP clients. They fetch raw HTML and stop.
The business impact is direct. If your product descriptions, pricing information, or key selling points load through JavaScript, AI search engines can’t see them. When potential customers ask ChatGPT or Perplexity about your products, the AI has nothing to work with.
The Timeout Problem
Adding to the JavaScript challenge is a hard time limit. AI bots have zero patience.
Testing revealed exact timeout thresholds: Gemini abandons requests after 4 seconds. ChatGPT gives websites 5 seconds. If your server is slow, if your page is large, or if network latency eats into response time, the bot moves on. It doesn’t wait. It uses your competitors’ information instead.
For businesses with international customers, this creates a hidden disadvantage. A website hosted in Michigan serving local customers loads quickly for human visitors. But if an AI bot accesses that same site from a distant data center with 200 milliseconds of network latency, the timeout window closes fast.
What This Means for Business
These findings have practical implications for companies investing in digital marketing:
Audit What Bots Actually See
Don’t assume AI bots see what your customers see. Free tools like JSBug.org let businesses check exactly what content bots extract from their pages. Compare the bot view against your actual page. If key information is missing, you’ve identified the problem.
Ensure Critical Content Exists in HTML
Product descriptions, pricing, specifications, contact information – anything that matters for AI search must exist in the initial HTML response. Businesses running JavaScript-heavy applications face a choice: migrate to server-side rendering or implement pre-rendering for bot traffic.
Focus on Text, Not Structure
Semantic HTML and elaborate page structure don’t help AI bots understand your content better. They see linear text. Make sure your most important information appears early in that text flow.
Don’t Over-Invest in AI-Specific Optimization
There’s no evidence that special formatting, custom markup, or “AI-optimized” content structures provide any advantage. Focus on clear, comprehensive text content. That’s what works.
The Two-Path Solution
For businesses running modern web applications with JavaScript-dependent content, two practical solutions exist:
Server-Side Rendering (SSR)
Frameworks like Next.js, Nuxt, or Astro can pre-render pages on the server before sending them to visitors. This produces complete HTML that both human browsers and AI bots can access. This approach requires development work and may involve rebuilding parts of your application, but it’s the cleanest long-term solution.
Edge-Based Pre-Rendering
This approach intercepts bot requests at the content delivery network level and serves pre-rendered HTML without requiring changes to your application code. Tools like EdgeComet detect AI bots and serve them fully rendered content within milliseconds, solving both the JavaScript execution problem and the strict timeout limits these bots impose.
The right choice depends on your technical resources and timeline. What matters is recognizing that bots don’t see JavaScript-rendered content and taking concrete steps to address it.
Verifying Your AI Visibility
The simplest way to check if AI bots can access your business information is to test it directly. Ask ChatGPT or Perplexity specific questions about your products or services that require pulling information from your website.
If the AI can’t retrieve the information, returns incorrect data, or says the content isn’t available, you’ve confirmed a visibility problem. The cause might be JavaScript rendering, slow server response, or content buried too deep in HTML markup.
This test takes minutes and reveals problems before they affect customer acquisition.
The Bottom Line
AI search represents a significant shift in how customers discover businesses online. But these systems operate under strict technical constraints: no JavaScript execution, aggressive timeouts, and text-only content extraction.
Companies investing in digital marketing need to verify that their websites work within these constraints. The good news is that the same optimizations that help AI bots – fast response times, complete HTML content, and clear text presentation – also improve the experience for traditional search engines and human visitors.
The businesses that will succeed in AI search aren’t the ones using exotic optimization techniques. They’re the ones ensuring their core content is actually visible to the bots doing the crawling.
Understanding what AI bots can and can’t do is the first step. Taking action to ensure your content is accessible is what separates companies that benefit from AI search from those that remain invisible to it.
For Michigan businesses and companies worldwide competing for customer attention, AI search visibility is no longer optional. It’s a fundamental requirement for digital presence in 2026.
About the Research: Testing methodology and open-source tools referenced in this article are available at OpenSeoTest.org and JSBug.org. Additional technical documentation is available at EdgeComet.
