By Song Jingwoo 
As Malaysia moves toward digital transformation, many SMEs still rely heavily on social media for marketing. Yet, according to Newpages — Malaysia’s leading AI-driven website platform — having a proper company website is now more important than ever. In the era of ChatGPT, Gemini, and other generative AIs, websites have become the primary way for AI to “see” and understand a business online.
According to OpenAI’s documentation (2024), ChatGPT’s data comes from three major sources:
- Publicly available web data
- Licensed data from third-party partners
- Information provided or generated by human trainers
Among these, the first category — publicly available websites — forms the foundation of ChatGPT’s knowledge. This means that business websites, blogs, news articles, and product pages are the main types of content AI can access.
Large-language models like ChatGPT learn from vast datasets such as Common Crawl, which collects over 3 billion web pages every month. The Stanford AI Index (2024) reports that more than 60% of AI training data comes from open websites. In other words, if a company does not have its own website, AI models have nothing substantial to learn about that business.
Unlike websites, social platforms limit access to their content:
- Facebook prohibits automated scraping (robots.txt).
- Twitter (X) blocks unregistered browsing (The Verge).
- Reddit provides limited, licensed access through its OpenAI partnership (2024).
These examples show that social-media content is not freely crawlable. While platforms remain useful for marketing, AI systems rely mostly on web pages for factual and brand information.
The Meta Llama 2 Technical Report (2023) states that its AI training used only publicly available data — not private content from Facebook or Instagram.
To illustrate this, Newpages compared both Google AI Overview and ChatGPT responses. As the screenshots show, when AI generates answers, the listed “Sources” come almost entirely from public websites — such as news portals, corporate domains, and technical blogs — not social-media posts.
Figure 1 – In Google AI Overview, all references are public websites (news and corporate pages).
.png)
Figure 2 – ChatGPT’s “Sources” also show website URLs, confirming its references come from the web.
These visuals offer clear proof: both Google AI Overview and OpenAI’s ChatGPT rely primarily on website data, not social-media content. For SMEs, this highlights a simple truth — AI can only recommend or mention what it can actually read online.
| Aspect | With Website | Only Social Media |
|---|---|---|
| Can AI find your brand? | High visibility | Low or none |
| Indexed by Google / AI | Yes | No |
| Long-term credibility | Stable and ownable | Depends on platform policy |
GEO, or Generative Search Engine Optimization, is the next stage of digital visibility. It focuses on optimizing online content so that generative AI tools like ChatGPT or Gemini can identify and recommend your business in natural-language answers. The foundation of any GEO strategy begins with your website.
- Build a proper business website with clear company info, products, and contact details.
- Publish expert content that demonstrates knowledge and credibility.
- Apply GEO principles – use brand, service, and industry keywords that AI can understand.
- Ensure crawlability – check your robots.txt and site visibility settings.
- Use social media strategically – as a support tool to drive traffic, not as your main presence.
Both Google AI Overview and ChatGPT prove one thing: AI recognises businesses through websites, not just social posts. For Malaysian SMEs that want to stay competitive in the age of AI, having a company website is no longer optional — it’s essential. Without it, your brand remains invisible to both search engines and AI-driven recommendations.
For a full analysis and original insights, read the complete article by Newpages Malaysia.