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

# How Heralded works

> The loop behind the Command Center: ask what your buyers ask, measure the answers, act, and check it worked.

Heralded comes in two forms. A [Snapshot](/concepts/snapshot) is a one-off report on one brand. The **Command Center** is the subscription that tracks your brands every day, and these guides describe it. Both read answers the same way, so the score, ranges and share of voice mean the same in each. [Change](/methodology/change) and [verifying actions](/methodology/verifying-actions) need the Command Center's weekly reports.

The Command Center runs a four-step loop.

## 1. Ask the engines what your buyers ask

You track [prompts](/concepts/prompts), the questions your buyers type, such as "best invoicing software for freelancers". Every day Heralded asks each prompt of the AI [engines](/concepts/engines) due that day and keeps the answers: four every day, Perplexity and Microsoft Copilot twice a week. One day a week the [weekly report](/concepts/daily-updates-and-weekly-reports) takes the daily update's place: it asks every engine, reads everything in full and sets your figures.

## 2. Measure how the answers treat you

A language model reads each answer for whether it names you, how strongly and how it describes you. Those readings become [Mentioned, Recommended](/methodology/mentioned-and-recommended) and [the Heralded Score](/methodology/heralded-score), set against your [competitors](/concepts/competitors) and traced to the [sources](/concepts/sources) the engines cite. [How Heralded measures](/methodology/how-we-measure) covers the method.

## 3. Act on what moves the answers

Heralded turns what it finds into [actions](/concepts/actions) and drafts what it can. In [Autonomy](/concepts/autonomy) you decide how much it does without asking. With a [connection](/concepts/connections) to your site, it can apply some fixes itself.

## 4. Check whether it worked

When work aimed at your prompts or your site is done, Heralded watches whether it moved the answers, net of what the market did anyway. [Verifying actions](/methodology/verifying-actions) explains the tests.

## The pages

The rail starts with your brand menu and account avatar, followed by your trial status when you are on a trial. **Home**, **Actions**, **Prompts** and **Content** sit above **More**. Open **More** for the other pages. It opens automatically when you visit one of them.

| Page | The question it answers |
| - | - |
| **Home** | Where do I stand, what changed, and what is waiting for me? |
| **Actions** | What work is open, who holds it, and did finished work pay off? |
| **Prompts** | What does each engine answer to each prompt, and where am I missing? |
| **Content** | Which of my pages answer buyers' topics, and what should I write or improve? |
| **Engines** | How does my [score compare across engines](/concepts/engines#the-engines-page)? |
| **Competitors** | Who do the engines name instead of me, and who is gaining? |
| **How AI describes you** | How does AI [describe my brand](/methodology/how-ai-describes-you)? |
| **Sources** | Which third-party sites do the engines cite? |
| **Website** | [Can AI read my site](/concepts/site-readability)? |

**Content Studio** lives under **Actions**. Its **Content** link returns you to the [topic table](/concepts/content#topics).

**Chats** sits below the pages. The Herald, Writer, Publicist and Engineer conversations are pinned there, with your unread message counts. Select one to open its [conversation](/concepts/actions#team-conversations).

**New chat** starts a private chat with the Herald. Your recent chats sit below the pinned conversations. **All chats** includes older and archived chats. [Conversation access](/concepts/actions#team-conversations) explains sharing and who can post.

On a phone, open the navigation menu beside the brand to see the rail as a drawer. You can also swipe right on the page. The left 24 pixels stay available for your browser's Back gesture, and sideways-scrolling content keeps its own swipe. The drawer follows your finger and opens when you drag past a third of its width. Drag left, tap outside, or use Back to close it. Conversations fill the screen.

Your brand menu holds **Brand settings**, prompt usage and content credits, and your [workspace settings](/concepts/workspaces-and-brands#settings). The account avatar beside it opens your account settings. Settings pages have their own menu and back link.

Tables use numbered pages. When a page needs to load, the current rows stay in place, dimmed, with a loading message over them. Paging pauses until the new rows arrive. If loading fails, the old rows return with an error above the table, and you can try the page again.


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