# Which sources do AI answers lean on for your questions?

> For the complete documentation index, see [llms.txt](https://attensira.com/llms.txt).
> Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.

> Ask your assistant which domains AI answers cite for your tracked prompts: the Attensira MCP call, the split by engine, and what source share does not mean.

- Canonical page: https://attensira.com/mcp/use-cases/see-which-sources-models-cite
- Family: mcp-use-cases
- Last updated: 2026-10-04
- Author: Karl-Gustav Kallasmaa (https://x.com/kallasmaa)
- Claims as data: https://attensira.com/claims.json?slug=see-which-sources-models-cite

**In short:** Ask your assistant to call get_analytics with area=source, days=90 and order=best_first on the Attensira MCP server. It ranks the domains AI answers cited for the prompts you track, by share of all citations, per engine. Then open the top domain with citation_timeline, and check where your own pages land with area=page.

## What it showed on attensira.com (read 2026-10-04)

Asked in Claude Code: "Which domains do AI answers cite most when they answer our prompts?"

Call: `get_analytics area=source days=90 order=best_first`

OpenAI's and Google's own help pages lead. Our domain is seventh, behind three tracked rivals.

| Row | Value | Note |
| --- | --- | --- |
| help.openai.com | 5.1% | of 97,149 citations · 23 of 91 days sampled |
| developers.google.com | 4.3% |  |
| semrush.com | 2.3% |  |
| otterly.ai | 2.1% |  |
| attensira.com | 1.4% | 1,442 citations |

_Source share for the questions our workspace tracks, ChatGPT and Perplexity._

## Which call answers it?

```json
{ "tool": "get_analytics", "arguments": { "area": "source", "days": 90, "order": "best_first", "limit": 50 } }
```

## What is the chain behind it?

1. **Pull source share over a long window** Call get_analytics with area=source, days=90, order=best_first and limit=50. Source share is a distribution, and short windows make a distribution lumpy in ways that look like findings.
2. **Repeat the call per model** Pass filter with one platform at a time. Answer surfaces lean on visibly different domain mixes, and a category-wide average hides the one difference that would change what you do next.
3. **Locate your own domain in the same window** Call get_analytics with area=page and days=90. It shows where your own URLs land against the same collected answers, which is how you find out whether you are in the mix at all.

## How do I read what comes back?

| What you see | What it means |
| --- | --- |
| One domain taking a large share of all citations | The models treat it as the reference for your category, and being absent from it costs you. |
| A long flat tail of many small domains | No consensus source exists. Breadth of coverage matters more than any single placement. |
| High source share together with an early mean position | Cited often and cited early. The strongest signal this reading can give you. |
| Your own domain absent from the list | You were not cited in the answers collected in this window. That is weaker evidence than a measured zero. |

## What can't this tell you?

- Source share is citations to a domain divided by all citations in the window. It is not traffic, not authority, and not influence over the answer text — a model can cite a page it disagreed with.
- The list reflects the prompts you track, not your category. A prompt set that skews to one topic produces a source mix that skews with it.

### What did it show on our own workspace?

The models answer our category's questions from the engines' own documentation first, ahead of any rival's marketing page. That shapes what the crew writes: pages that explain how the engines behave, sourced to the engines themselves, rather than another feature comparison.

### Why split the list by engine before concluding anything?

Answer surfaces do not share a reference set, and a category-wide list is the average of several different worlds. Run the call once per engine with `filter`. A domain that dominates one surface and is absent from another is a real finding about that surface; the blended list shows it as mid-table.

### Which two readings change what you do next?

Your own position. If you are on the list, the work is incremental: more of what already gets cited, on more of your prompts. If you are absent across 90 days and a wide prompt set, the work is different in kind: you are trying to enter a reference set, not win share of one.

How concentrated the list is. One dominant reference domain rewards placement: getting onto it, or being what it cites, is worth more than any number of your own pages. A flat tail rewards breadth, because there is no gatekeeper to get past.

The paired question is which of your own pages already earn citations: [the page audit](/mcp/use-cases/audit-which-pages-get-cited). If a rival sits on the reference domains and you do not, [check whether they are gaining](/mcp/use-cases/check-whether-a-competitor-is-gaining).

## FAQ

### What exactly is source share?

The share of all citations collected in the window that point at one domain. It is computed over the answers Attensira collected for your tracked prompts, so it describes the reference set those answers drew on.

### Why use a ninety-day window here when other recipes use seven or thirty?

Because this is a distribution rather than a rate. Citations spread over many domains, so a short window gives a handful of citations per domain and a ranking that reorders itself every few days without anything having changed.

### My own domain is absent from the list. Does that mean my pages cannot be cited?

No. It means you were not cited in the answers collected in that window. Absence from a sample is weaker evidence than a measured zero, and the first thing to check is whether AI crawlers can fetch the pages at all.

### Does a high source share mean that domain influenced the answer?

It means the domain was cited often. Citation and influence are not the same thing — a model can name your brand in its prose while citing a competitor's blog underneath, and it can cite a source to disagree with it.

### What does the position figure next to a source mean?

It is a mean index within a citation list per source domain — how early that domain tends to appear among the citations on an answer. It is not a rank against competitors and cannot be read as one.

## Sources

Each entry below is one claim on this page. `retrieved` is the date the exporter or a human last saw the source say this, and nothing more: it is not a publication date and not a promise of re-verification. `cite` is the anchor on the canonical page where this claim is rendered.

### source-share-definition

Attensira's MCP recipes state that source share is citations to a domain over all citations in the window, and that it is not traffic, authority, or influence on the answer text.

- source: https://docs.attensira.com/mcp/recipes (docs.attensira.com)
- retrieved: 2026-09-04
- cite: https://attensira.com/mcp/use-cases/see-which-sources-models-cite#claim-source-share-definition

> Source share is citations to a domain over all citations in the window — it is not traffic, authority, or influence on the answer text.