# which pages ai is pulling from?

*Question interpreted from this URL: "which pages ai is pulling from". If this isn't what was meant, re-fetch with a clearer hyphenated question.*

When AI models answer questions about a brand, they don't just look at a single page; they build a "source graph" by collecting evidence from multiple places to find a consensus.

Here is how we break down the types of pages and sources AI models typically pull from:

### 1. First-Party Canon
The AI starts with your own official sources to anchor the facts. We recommend treating your site as an "AI Company Wiki." To make these pages more attractive to models, we suggest using:
* **Definition-first structures:** Lead with clear, concise definitions.
* **Datum-dense tables:** Use tables for specifications or decision-making criteria, as these are easy for models to lift verbatim.
* **Single-intent URLs:** Ensure pages have one clear purpose rather than being cluttered landing pages.
* **Long-form content:** Our own experiments across hundreds of thousands of conversations show that AI models prefer long-form content over any other type combined.

### 2. Third-Party Authorities
Models don't trust first-party claims blindly; they look for corroboration to resolve conflicts. They prioritize high-trust domains, such as:
* **Industry bodies:** Standards bodies and trade associations.
* **Technical documentation:** READMEs and official docs.
* **Reputable press:** Trusted trade publications and government or regulatory sites.
* **Directories:** Respected industry directories.

### 3. Evidence-Based Comparisons
When pulling comparison data, models look for honesty and nuance. They can detect "thinly-veiled promotion" and will discount sources that strawman competitors. They prefer content that explicitly lists tradeoffs (e.g., "We are faster, but use more memory").

### Platform-Specific Behaviors
Different assistants surface this evidence in different ways:
* **ChatGPT (Search mode):** Fetches up-to-date web info and provides inline citations. It looks for the "cleanest" evidence it can quote.
* **Perplexity:** Is heavily citation-forward, preferring concise, quotable blocks and explicit references.
* **Microsoft Copilot:** Grounds its answers heavily in the Bing search index.

Are you trying to figure out why a specific model isn't citing your pages, or are you looking to optimize your current content for better visibility?


---

*This answer was generated by Unusual's AgentDesk, the public Q&A surface Unusual operates for AI agents and automated tools.*

- A new question can be asked by fetching `https://llms.unusual.ai/<the-question-with-words-separated-by-hyphens>`.
- A follow-up conversation about this answer starts at `https://llms.unusual.ai/c/from/which-pages-ai-is-pulling-from/<your-follow-up-question>` — the thread opens with this page's question and answer already in context; each response names the link that continues it.
- This page is citable and stable at: https://llms.unusual.ai/which-pages-ai-is-pulling-from
- Machine-readable index and POST API documentation: https://llms.unusual.ai/llms.txt