In the last months, models like ChatGPT, Gemini, and Perplexity have made something very clear: when users search for information about a brand, the official website is often the most influential and most reliable source the AI uses.
This isn’t about classic SEO dynamics anymore. This is AI visibility.
If a brand doesn’t maintain its website with complete, up-to-date, authoritative content, it leaves the field open to third‑party sources: less accurate, less controllable, and rarely aligned with the brand’s identity.
In this article, we summarize the data from Giada Birbitello’s analysis (Head of Customer Success at Kleecks) to show how and when brand websites truly influence LLM responses.
Where do LLMs get their information? The 4 main source categories
According to the analysis, generative models draw from four macro‑categories of data:
1) Books, industry reports, academic texts
- ChatGPT: ~25–30%
- Gemini: ~30–40%
- Perplexity: ~30–40%
Reliable, validated, ideal for historical, conceptual, or strategic knowledge — but not always up to date.
2) General web text (news, blogs, forums)
- ChatGPT: ~40–50%
- Gemini: ~40–50%
- Perplexity: ~25–30%
The “online noise”: broad, current, but mixed in quality.
3) Technical repositories (GitHub, StackOverflow, etc.)
- ChatGPT: ~10–20%
- Gemini: ~5–10%
- Perplexity: ~10–15%
Relevant only in technical contexts.
4) Official websites, documentation, policies
- ChatGPT: ~20–40%
- Gemini: ~5–15%
- Perplexity: ~20–25%
Here’s the key point: even if the average percentage looks modest, in certain queries this contribution rises up to 100%.
When brand websites become the primary source (even up to 100%)
The weight of the brand website varies dramatically depending on the category of question.
a) Product information (specs, pricing, warranty, details)
- ChatGPT: 70–90%
- Gemini: 80–100%
- Perplexity: 70–80%
For LLMs, the official site is the single source of truth.
b) Brand statements, mission, policies, press releases
- ChatGPT: 90–100%
- Gemini: 90–100%
- Perplexity: 60–70%
Models do not improvise here — they fetch the original material.
c) Comparisons, reviews, user opinions
- ChatGPT: 10–30%
- Gemini: <5%
- Perplexity: 10–15%
Dominated by third‑party sources; the brand site provides only factual grounding.
d) General knowledge or brand heritage
- ChatGPT: 10–20%
- Gemini: 15–40%
- Perplexity: 25–30%
Heritage pages contribute, but models cross‑validate heavily with external sources.
e) Analytical or speculative questions (“Why is this brand successful?”)
- ChatGPT: <10%
- Gemini: 10–30%
- Perplexity: 10–15%
Here, broader market analysis and independent sources dominate.
The real takeaway: low average contribution, very high contribution where it matters
The average contribution (5–40%) may look small.
But on the queries that matter most for brand perception and conversion (product details, prices, values, identity, corporate statements) the official site should ideally be the primary source, and sometimes even the exclusive one.
In reality, however, very few brand websites today are structured and enriched enough to truly serve as the primary source for all these points. Many sites are increasingly minimal or sparse in content, so when the official pages are incomplete or unclear, AI ends up relying on older or less reliable third‑party sources instead.
Why this is critical for brands in 2026
Here’s why a website with limited informational depth has become a strategic risk in an AI-first world:
1) LLMs evaluate both qualitative and quantitative signals
A rich, updated, well‑structured website increases the likelihood that the model uses it.
2) Official information gets replicated across AI answers
If the brand doesn’t provide it, AI models fill the gaps through external sources.
3) Brands lose control of their narrative
Users are shifting from Google to ChatGPT, Perplexity, and Gemini.
Wrong or outdated information inside those models quickly becomes a brand liability.
4) Competitors with better websites gain an objective advantage
LLMs reward freshness, structure, clarity, and authority.
Conclusion
Giada’s analysis brings to light a point that should be obvious, yet many brands still overlook it.
Everyone agrees the official website is one of the primary sources LLMs rely on. But in practice, entire sectors are stripping their sites of informational depth in favor of ultra-minimal experiences.
The result? AI models have less to learn from, and the brand becomes less visible in search and AI-driven answers.
With Kleecks, brands can ensure their websites are structured, rich, and continuously updated, making it far more likely that LLMs treat them as the authoritative source.
In today’s AI-first ecosystem, substantial, accurate content isn’t optional anymore. It’s the key to controlling how your brand appears across AI-driven platforms.