> ## Documentation Index
> Fetch the complete documentation index at: https://docs.topify.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Source metrics
> Definitions and formulas for citation count, used %, avg citations, and source authority — computed at both domain and URL granularity.
## Overview
Source metrics track how AI models use and cite content sources when generating responses. These metrics reveal which domains have **content authority** -- the ability to influence what AI search engines say.
Sources are analyzed at two levels:
* **Domain level** -- aggregated across all URLs from a domain (e.g., all pages on `techcrunch.com`)
* **URL level** -- individual page performance (e.g., `techcrunch.com/article-1`)
***
## Citation count: total references to a source
**Definition**: Total number of times a source is explicitly referenced in AI responses.
### Formula
```
Citation count = Total citations of the domain (or URL)
```
Each citation in a response counts separately. If one response cites two different URLs from the same domain, that's two citations for the domain.
```
Domain: techcrunch.com
Time period: Last 7 days
Response 1: Cited techcrunch.com/article-1, techcrunch.com/article-2 (2 citations)
Response 2: Cited techcrunch.com/article-1 (1 citation)
Response 3: Cited techcrunch.com/article-3 (1 citation)
Response 4: No citations from techcrunch.com
Citation count = 2 + 1 + 1 = 4
```
Content from techcrunch.com was explicitly cited 4 times across 3 responses.
***
## Average citations per response
**Definition**: The average number of times a source is cited per AI response where it appears.
### Formula
```
Average citations = Total citations of source / Responses where source was cited
```
Higher values indicate deeper trust -- when an AI model uses the source, it tends to cite multiple pages or cite it multiple times.
```
Domain: techcrunch.com
Time period: Last 7 days
Response 1: 2 citations from techcrunch.com
Response 2: 1 citation from techcrunch.com
Response 3: 1 citation from techcrunch.com
Total: 4 citations across 3 responses
Average citations = 4 / 3 = 1.33
```
When AI uses techcrunch.com, it cites an average of 1.33 unique URLs from the domain per response.
***
## Used: how often a source appears
**Definition**: The share of AI responses that cited this source within the selected time window. The same definition applies at both domain and URL level.
### Formula
```
Used (%) = (Responses that cited the source / Total responses analyzed) x 100
```
Citation depth — how many URLs from the same domain show up when it's cited — is reported separately as **average citations per response** (defined above).
```
Domain: techcrunch.com
Time period: Last 7 days
Response 1: Cited 2 URLs from techcrunch.com
Response 2: Cited 1 URL from techcrunch.com
Response 3-100: No citations from techcrunch.com
Responses citing techcrunch.com = 2
Total responses analyzed = 100
Used = (2 / 100) x 100 = 2%
```
techcrunch.com appeared as a source in 2% of AI responses for this project. Across those 2 responses, it averaged 1.5 unique URLs cited per response (reported separately as **avg citations**).
```
URL: techcrunch.com/article-1
Time period: Last 7 days
Response 1: Cited this URL
Response 2: Cited this URL
Responses 3-100: Did not cite this URL
Used = (2 / 100) x 100 = 2%
```
This specific article was cited in 2% of responses, regardless of how many times it was cited within each one.
The "used" metric is your share-of-voice equivalent for sources: it tells you how often AI reaches for this source when answering questions in your space.
***
## Source categories
Topify.ai classifies cited domains into categories to help you understand the source landscape:
| Category | Examples | Description |
| ----------------- | ----------------------------------- | --------------------------------------------- |
| **Corporate** | apple.com, tesla.com | Official brand and company websites |
| **Editorial** | nytimes.com, techcrunch.com | News outlets and journalism |
| **Institutional** | .gov, .edu, who.int | Government, academic, and institutional sites |
| **Reference** | wikipedia.org, stackoverflow\.com | Knowledge bases and reference platforms |
| **UGC** | reddit.com, youtube.com, medium.com | User-generated content platforms |
| **Other** | Edge cases, parked domains | Everything that doesn't fit the above |
These categories help you understand whether AI providers are citing authoritative sources, user-generated content, or your competitors' sites when answering prompts related to your brand.
***
## Using sources data to improve your strategy
Sources data helps you answer questions like:
* **Which of your pages are already being cited?** Look for your domain in the top domains list. High citation counts mean AI providers trust your content.
* **Where are the gaps?** If competitor domains are cited frequently but yours isn't, you need content that covers those topics.
* **What content types perform best?** Compare citation rates across source categories to understand what AI providers prefer to cite.
* **Which pages should you optimize?** URLs with high "used" counts are already influencing AI responses -- improving them could increase your visibility further.
## FAQ
### How do I get my own pages cited more often?
Look at the domains AI currently cites most for your prompts and study what they have in common: structured headings that match the question, statistics and named numbers, citations to authoritative sources, and a confident tone. Then write content that does those things better. The AI agent's `geo_optimize` skill applies these techniques to existing pages automatically — see [Using the AI agent](/guides/using-ai-agent).
### My domain isn't in the source list — why?
Either AI providers haven't picked up your pages yet, or they have but cite other sources first when answering your tracked prompts. Both are content problems, not data problems: you can fix them by publishing pages targeted at the prompts where your domain is missing, or by improving the pages you already have so they're more citable.