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

# Save conversation answers to GHL custom fields

> Create an extraction, assign it to an assistant, and verify the contact field after a test.

Automatic extraction reads a conversation and writes selected answers to mapped GoHighLevel contact fields. Configure it in Assistable. You need a connected GHL account and the target contact custom field. Direct accounts do not offer this GHL field-mapping workflow.

## Create and assign an extraction

1. Open the assistant and choose **Map Custom Fields** to open **Automatic Extractions**.
2. Select **New extraction**. Enter a unique **Name** using letters, numbers, underscores, or hyphens, up to 64 characters. For example, `preferred_service`.
3. In **Description**, define the value to capture. For example: “The service the customer explicitly requested or confirmed. Use their latest correction. Leave this empty if they never gave a service.”
4. Under **Mapping**, select the intended GHL **Custom field**. Check its displayed type; use a field appropriate for the answer.
5. Select **Create extraction**. Then choose **Add to assistant** on the extraction. Creating an item in the library does not assign it.
6. Give the assistant a prompt that actually asks for the information when needed.

## Extraction fields

| Field | Accepted value and behavior |
| - | - |
| Name | Required and unique within the account; 1–64 letters, digits, underscores, or hyphens. No spaces. |
| Description | Up to 800 characters. It can be empty, but a precise description tells the extractor what to capture and what to omit. |
| Custom field | Required. Choose from the connected GHL account's fields; the list shows the field name, field key, and data type. Available names and types come from that account. |

An extraction maps to one custom field. Create separate extractions for different destinations, and assign each to the assistant. The destination's type and allowed values still govern whether GHL accepts a write.

## Verify the stored value

Use a test contact and provide a distinctive answer. Finish the call, or allow the chat reply and background processing to complete, then open the contact in GHL and inspect the target field. A reply can arrive before the custom field write completes.

Repeat with a correction and with the answer omitted. The extractor is instructed to use the latest user-provided value and omit unknown values. An omitted value does not mean the existing contact field will be cleared. For summary fields, explicitly describe the summary to write rather than presenting it as a verbatim customer answer.

## Fix missing or wrong values

* **Field list fails to load:** repair GHL authorization and reopen the mapping form.
* **Extraction exists but nothing changes:** check **Add to assistant**, the selected contact, and whether the conversation actually contains the answer.
* **Wrong field changes:** inspect the mapping and duplicate GHL field names. Use distinct field names so the destination is unambiguous.
* **GHL rejects the value:** check the field type and allowed values. An invalid field ID or value needs correction; repeating the conversation alone will not fix it.
* **A value is guessed or too broad:** tighten the description and retest with exact, ambiguous, corrected, and missing answers.

For names and contact identity, use [Update contact details](/v3/guides/workflows/correct-contact-name).


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