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

# Review conversations

> Find, filter, tag, and export the conversations your chatbot has with visitors.

**Chat History** holds every conversation your chatbot has with visitors. Use it to check answers, find conversations that need a person, and export conversations for your records.

## Before you start

* **Plan:** any plan. See [Plans and limits](/docs/reference/plans-and-limits.md).
* **Role:** any team role can read, filter, tag, and export conversations. To delete conversations, you need the Manager role or higher. See [Team members](/docs/guides/account/team-members.md).

## Start from the Dashboard

The **Dashboard** is the first page of your chatbot. It shows **Total Messages**, **Positive Feedback**, and **Negative Feedback**.

* Select **View Chat History** to open all conversations.
* Select **View Feedback** under **Positive Feedback** or **Negative Feedback** to open **Chat History** filtered by that feedback.

The Dashboard also shows your content totals, training status, a chat preview, and your install code. For trends over time, see [Analytics](/docs/guides/monitor/analytics.md).

## Choose a view

In your chatbot's sidebar, select **Chat History**. The views at the top of the list split conversations by state:

| View | Shows |
| - | - |
| **Attention** | Conversations that are not resolved and have an escalation, an unread visitor message, or negative feedback. |
| **Open** | Conversations that are not resolved. This is the default view. |
| **Closed** | Resolved conversations. |
| **All** | Every conversation. |

A resolved conversation leaves **Attention**, whatever its other flags. To resolve conversations, see [Reply to visitors and take over conversations](/docs/guides/human-support/reply-and-take-over.md#mark-as-resolved).

## Search and filter

**1. Search**

In **Search conversations**, type words from the conversation. The list shows matching conversations with the matching text. Search looks at all conversations. The date range and filters, including **Source**, do not apply to search results.

**2. Set a date range**

Open the date menu and select **Today**, **Last 7 days**, **Last 30 days**, **This month**, or **Last month**. To choose your own dates, select **Custom**, then set **From** and **To**. The range uses each conversation's last activity.

**3. Add filters**

Open **Filter** and select the filters you want.

| Filter | Options |
| - | - |
| **Source** | Where the conversation started: **Website**, **API**, **WhatsApp**, **Messenger**, **Slack**, **Google Chat**, **Crisp**, **Freshdesk**, **Zoho SalesIQ**, **Zendesk**, or **Email**. You can select one source at a time. |
| Tags | Your conversation tags. |
| **Read/Unread** | Read, Unread |
| **User Escalations** | Escalated, Not Escalated |
| **Importance** | Important, Not Important |
| **User Information** | Anonymous, Not Anonymous |
| **User Reactions** | Positive, Negative |
| **GPT Model** | The model that answered. |

To find answers that visitors did not like, set **User Reactions** to **Negative**. To fix a wrong answer, see [Fix a wrong answer](/docs/guides/answers/fix-a-wrong-answer.md).

### Conversation sources

Each conversation keeps the source where it started:

* **Website** is the chat widget on your website and the [hosted chat page](/docs/guides/install/control-where-it-appears.md#share-a-link-to-the-chat).
* **API** is a conversation that started through API v2 or API v0. This includes conversations from the CLI, the SDKs, and the MCP server.
* **Email** is a conversation that started with an email to your chatbot's [email address](/docs/integrations/email.md).
* The other sources are your [integrations](/docs/integrations/overview.md).

In the list, a conversation from an integration or the API shows the source icon on the corner of the visitor's picture: the integration logo, an envelope for **Email**, or a code icon for **API**. Website conversations have no icon.

Some older conversations show **Website** because SiteGPT did not save their source. For example, conversations that started through the API before this change (October 9, 2026, 19:22 UTC) show **Website**. Conversations from the retired Intercom integration can show **Intercom** (`intercom`).

To open **Chat History** with one source selected, add `?source=` and the source name to the page address, for example `?source=whatsapp`. The names are `website`, `api`, `whatsapp`, `messenger`, `slack`, `google_chat`, `crisp`, `freshdesk`, `zoho_salesiq`, `zendesk`, and `email`. For conversations from the retired Intercom integration, the name is `intercom`. Save the address as a bookmark to check one source every day.

In API v2, each conversation has a `source` field with these names, and `GET /chatbots/{chatbotId}/conversations` accepts a `source` filter. See [API v2](/docs/api-reference/v2/getting-started.md).

## Read a conversation

Select a conversation to open it. The conversation shows:

* Each message from the visitor, the AI, and your team, with the sources the AI used.
* Links in visitor messages. Select a link to open it in a new tab.
* Photos that a visitor sends on WhatsApp. Select a photo to open it full size in a new tab. Other files show as a link with the file name and size. If the photo is no longer available, a link shows instead. When a photo or file is the newest message and has no answer yet, the list shows its caption, or its type and file name, for example "Image: photo.jpeg".
* **Forwarded** above a message that a visitor forwarded on WhatsApp. The visitor did not write these words, and the AI knows that.
* A marker with a link when the visitor moved to a different page during the conversation. This shows the visitor's path through your site.
* The source at the top, for example **Website** or **WhatsApp**. It does not show on small screens.
* In an email conversation: a bar with the subject and the addresses, the drafts that wait for your team, and a line under each reply that says what happened to the email. See [Email](/docs/integrations/email.md#work-with-drafts).
* The **Escalated** and **Resolved** badges.

Select **View Details** to open **Conversation Details**. It shows the title, **Total Messages**, **Positive** and **Negative** feedback counts, **Started On**, **Source**, **Escalated**, the **Page URL**, and the visitor's **Name**, **Email**, and **Phone**.

To reply to the visitor or take over from the AI, see [Reply to visitors and take over conversations](/docs/guides/human-support/reply-and-take-over.md).

## Mark, tag, and rename conversations

In an open conversation, use the icons and the menu at the top:

* **Mark as Important** flags the conversation. **Remove Mark** clears it. Use the **Importance** filter to find flagged conversations.
* **Add Tags** adds tags. Type a new name to create a tag. **Modify Tags** changes the tags on a conversation.
* **Edit title** renames the conversation.
* **Mark as unread** puts the conversation back in the unread state.
* **Copy Link** copies a link to the conversation.
* **Open as User** opens the conversation as the visitor sees it.
* **Lock conversation** stops the visitor from switching a conversation in human mode to the AI. It shows only when your team holds the conversation without a lock. See [Lock a conversation that your team holds](/docs/guides/human-support/reply-and-take-over.md#lock-a-conversation-that-your-team-holds).

When a visitor starts a conversation by selecting a conversation starter, SiteGPT tags the conversation with the starter's button label in capital letters.

To delete a tag from every conversation, open **Filter** and select the delete icon next to the tag. You must confirm.

To act on many conversations at the same time, select their check boxes in the list. Then select **Mark as Resolved**, **Mark as Unresolved**, **Mark as Important**, **Remove Mark**, or **Delete**.

## Export conversations to CSV

**1. Set the view and filters**

Choose the view, date range, and filters that you want, including the source. The export uses them. The search text is not used.

**2. Export**

Select **Export Conversations**.

The browser downloads `chat_history_export.csv`. It has one row per message, with these columns:

`messageId`, `messageType`, `threadId`, `pageUrl`, `question`, `questionTimestamp`, `answer`, `answerTimestamp`, `reaction`, `mode`, `gptModel`, `escalated`, `important`, `resolved`, `anonymous`, `userEmail`, `tags`

The **Open** and **Closed** views limit the export to open or resolved conversations. The **Attention** and **All** views export conversations of both states.

`threadId` is the conversation ID. To export one conversation only, open it and select **Export Chat History**.

## Delete conversations

* **Delete Conversation** in the conversation menu deletes one conversation.
* **Delete** in the bulk bar deletes the selected conversations.
* **Delete Visitor Data** in the conversation menu deletes every conversation from this visitor, their lead record, and the visitor identity. It shows only when the visitor gave their details. You need the Manager role or higher.

Each option asks you to confirm. You cannot undo a delete. For the difference between deleting a lead and deleting visitor data, see [Manage your leads](/docs/guides/leads/manage-leads.md#delete-a-lead-or-all-of-a-visitors-data).

## Check that it works

1. Select the **Attention** view. Every conversation in it is open and has an escalation, an unread message, or negative feedback.
2. Select the **Open** view, set a date range and a filter, then select **Export Conversations**. The CSV file contains only messages from the conversations in the filtered list.


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