> ## Documentation Index
> Fetch the complete documentation index at: https://docs.attensira.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Brand family

> The other domains, products and sub-brands you own count as you once you confirm them. Agents propose them with evidence, and prefer publishing on your own domain.

## What counts as you?

Your workspace's domain counts as you, and so does **every subdomain of it**. An answer that cites `blog.example.com` has always counted as citing `example.com`. Nothing needs setting up for that.

The **brand family** is everything else you own: a second domain, a product with its own name, a sub-brand. Once you confirm one, an answer that names it, or cites a page on its domain, counts as you. That applies to every headline number: share of voice, citation rate and position. A confirmed property is never listed as a third-party source, a deciding source, an outreach target or a new competitor.

| Kind | What it is | How answers are matched |
| - | - | - |
| `subdomain` | A host under your own domain, such as `blog.example.com` | Already counted as you. Confirming one only names it, so the [split](#how-much-of-you-each-part-is) can show how much of "you" it carries |
| `domain` | Another domain you own, such as `hello.com` | Citations of the domain and of any subdomain under it (`blog.hello.com`). Not the other way round: owning `shop.hello.com` does not make the rest of `hello.com` yours |
| `product` | A product name, such as "Acme Analytics", with an optional domain of its own | The name in the answer text, matched the way your brand name is, and citations of its domain when it has one |
| `sub_brand` | A sub-brand name, with an optional domain | As for a product |

### What a property cannot be

* **Your own domain, or anything under it, as a `domain`.** It already counts as you. A host under your domain is added as a `subdomain`.
* **A tracked competitor's domain.** One domain cannot be both you and a rival, so it is refused with `property_is_competitor`. If it really is yours, remove it from [Competitors](/tracking/competitors) first.
* **A name shorter than three characters, a placeholder, or a common word.** "Analytics", "Cloud" or "Studio" would make every answer about the subject an answer that named you. Name a product the way the answers would, such as "Acme Analytics".
* **More than 50 properties** in one workspace, counting proposals and rejections.

A product name that is also an everyday phrase can still over-count, as a brand name can. A confirmed name is matched the way your brand name is, so confirm only names the answers use to mean you.

## Only confirmed properties count

A property is **proposed**, **confirmed** or **rejected**. Only a confirmed one changes a number. A proposed one is a question an agent is asking you, and a rejected one is kept so the question is not asked again.

Every confirmed property records the person who confirmed it. Nothing is confirmed automatically, and an agent cannot confirm one: an agent may do the finding, but deciding what your numbers mean is yours.

You manage the list in **Settings → Workspace → Brand family**:

* Proposals sit at the top, marked **Proposed by an agent**, with the evidence and a confirm or reject control for each.
* Confirmed properties are listed below them. With none, the group shows your domain and its subdomains, which already count.
* **Add a property** takes a domain, or a name with its kind.

Proposals are also announced in the [Inbox](/agent/inbox) as one row, "Are these yours?", listing each with its evidence. You confirm or reject them here, in Settings. Settings has no control to remove a confirmed property.

## Confirming a property changes history

An answer that named "Acme Analytics" last month named you, once you confirm that Acme Analytics is yours. So confirming a property re-reads your stored answers against the new family. It is the same re-extraction a competitor change goes through. A queued job does it within minutes, and the nightly reconcile catches anything left over. Rejecting a proposal, or adding a new one, changes nothing that was measured, so it re-reads nothing.

Until an answer has been re-read, it is counted as stored in the rolled-up numbers and left out of the per-property split. The split says how many answers are waiting, and why.

## Agents propose properties

Once a week, alongside the [markets proposals](/tracking/markets#agents-propose-markets), an agent looks for what else is yours:

* **Subdomains of your domain** seen in citations over the last 90 days, or in the crawl of your site. They already count; proposing one is asking you to name it.
* **Cited domains whose name carries your brand**, such as `examplehq.com` for `example.com`.
* **Domains your home page links to** in its structured data (`sameAs`) or its footer. Profiles on platforms such as LinkedIn, GitHub or G2 are left out.
* **Products** your Knowledge profile lists under what you sell, when they read as names rather than descriptions.

A domain candidate is also asked a "same organisation?" question before it is proposed. That check can only stop a proposal; it can never confirm one. A tracked competitor's domain is never proposed, and neither is anything you have already confirmed, rejected or listed. One pass proposes at most ten properties.

## How much of you each part is

Your headline numbers already include every confirmed property. The **Brand family** panel on **Sources** shows how much each part carries. It appears once a workspace has at least one confirmed property.

* **Rolled up**: how often answers named your brand or any property, and how often they cited a page of any of them, over every answer read, with the most-cited pages.
* **Per part**: your main brand first, then each confirmed property. Each has a **hit rate**, the answers in which it counted as you, named or cited. It is one rate because the stored answer records that the part counted, not which of the two it was. Each also has a **cited rate**, the answers citing a page on its domain (unmeasured for a part with no domain of its own), and its `n` and most-cited pages.

**The rows overlap.** One answer can name your main brand and cite your blog, so it counts on both rows. The per-part rows do not add up to the rolled-up line, and are not meant to.

`query_evidence` splits share of voice by `property`, and narrows it with `property_ids`. Only share of voice can be split this way, for the same reason as the hit rate.

## Where agents publish

The rule: if your main brand is `example.com`, then `blog.example.com` is much better than `blog.hello.com`. Agents rank every place a page could go in this order, and say why:

1. **A path on your domain**, such as `example.com/blog/`. It inherits everything the domain has earned, and every citation of it counts as you.
2. **A subdomain of your domain**, such as `blog.example.com`. It carries your brand and counts as you, but search engines treat a subdomain as a partly separate site, so it starts behind a path.
3. **Another domain you have confirmed**, such as `hello.com`. It counts as you, but it builds its authority alone and splits your brand across two sites.
4. **A new domain. Agents never propose one.** It starts with no authority and splits your brand, and starting a new site is your decision.

Agents have this ranking twice: in the knowledge brief they start from, and in a `brand_family` tool they ask before choosing where a draft, a new page or a tool will live. Given the place a draft is meant for, the tool also checks it: anywhere but a path on your domain is flagged unless the draft states why the better option does not fit, and a domain you have not confirmed is flagged whatever reason it gives. A proposed property is listed to the agent as proposed, so it is never treated as yours before you confirm it.

## Over MCP and the API

[`get_brand_family`](/mcp/tools#get_brand_family) returns every property with its status and evidence, then the window rolled up and split by part. It is MCP-only for now. Confirming, rejecting and adding properties is done by a person, signed in to the dashboard. An API key or an assistant connection cannot change the list.

Client reports carry a **Brand family** section once the workspace has a confirmed property. See [Client reports](/agencies/client-reports).

## Related

* [Markets](/tracking/markets): reading your prompts in more than one country.
* [Citations and sources](/measure/citations-and-sources): what counts as a citation of you.
* [Competitors](/tracking/competitors): the brands you are measured against.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.