We will explain how to integrate ListeningMind data with ChatGPT using the Custom GPT feature provided by ChatGPT.
To create a custom GPT, you must subscribe to one of ChatGPT's paid plans.(*Plus, Pro, Team, Enterprise, or Edu plans)
In the left sidebar of ChatGPT, click ① GPTs, then click ② Create.
When you enter a prompt describing your desired custom GPT in the Create tab's input field, details such as the profile image and other input values will be generated automatically.
In the Configure tab, you can review the automatically generated settings and make any necessary edits.
① Name / ② Description
These fields are used to introduce your custom GPT with a name and a short description.
③ Instructions 💡
This section provides detailed guidance on how the custom GPT should respond to user questions — including tone, style, and any specific rules it must follow.
Below is an example of detailed instructions designed to help the GPT accurately analyze ListeningMind intent data and improve the overall quality of its responses.
If the user does not specify a country, select the country based on the language of the query when making API requests.
When referencing the source of an API used in a response, clearly indicate the endpoint or path used in the request.
Read and analyze the user's question thoroughly. Retrieve the broadest possible range of relevant data, perform cross-analysis, and extract deep insights. In every response, present the supporting evidence and data used to draw conclusions.
If the user requests analysis of a specific keyword, display information such as trend data over time, search pathways, and searcher demographics.
If the user provides a product or brand and requests analysis of consumer usage, purchase contexts, or category entry points, use the Cluster Finder to identify related keywords. From these, extract relevant situations, reasons, or emotions that might prompt users to recall the product or brand. If few related keywords are found, suggest investigating additional high-volume keywords with similar meanings, or alternative competing brands or products. The default limit for analysis is set to 800.
For queries using non-Korean keywords (e.g., in English), always convert the keyword to lowercase when making ListeningMind-related API requests.
If an error occurs while retrieving data, attempt to re-fetch the data by adjusting request parameters or variables as needed.
When using the Cluster Finder API, utilize the community (group) information provided by the Cluster Finder for analysis.
Before making a request to the google_ads API, be sure to verify the available list of locations.
Before making a request to the google_trend API, also ensure the list of locations is checked.④ Conversation Starters
You can set example questions that will be shown by default in the prompt area of your custom GPT.
⑤ Knowledge
You can upload various files (PDF, TXT, etc.) to provide your GPT with the necessary background knowledge.
⑥ Recommended Model
Specifies the default AI model your GPT will use.
⑦ Capabilities
Choose additional features your GPT can access and use.
⑧ Actions
This setting allows your GPT to perform specific tasks by integrating with external services or APIs.This is where ListeningMind data can be connected to your custom GPT.
In Actions, click Create new action.
In the Add actions window, click the settings icon for ① Authentication.
① Authentication Type: Select API Key.
② API Key: Copy and paste the API key issued from [Manage API Keys].
③ Auth Type: Select Custom.
④ Custom Header Name: Enter LM-API-Key.
⑤ Click Save to complete the authentication settings.
① In the Schema section, click Import from URL.
② In the input field, paste the schema URL displayed in [Manage API Keys]
③ Click Import, then check that the Schema and Available operations are displayed correctly as shown in the image below.
In the Preview pane on the right, you can test your custom GPT in advance. For example, if you ask, “Why is Labubu so popular?”, a request like the one below will appear, allowing the GPT to call the data API. Select Allow.
If the request below is not displayed, it means that the previous configuration was not completed correctly.
Custom GPT wants to talk to listeningmind-mcp-api.ascentlab.io
[Allow] / [Always allow] / [Decline]
Click Create at the top to generate your custom GPT.
Access to the GPT integrated with ListeningMind data must be restricted to Private (Only me) or My Team (Workspace) members only.
Public sharing, such as via the GPT Store or Anyone with the link, is strictly prohibited.
This falls under the “No sharing or transferring to third parties” clause agreed upon when the API key was issued.
If large-scale API requests or inappropriate usage is detected due to public access, the issued API key may be deactivated.
In GPTs > My GPTs, you can access the custom GPT you created.
You can also pin your GPT to the sidebar for easier access.
If needed, you can enhance the Instructions by editing your GPT.
Try asking questions like the examples below:
Analyze the search volume trends for “cordless vacuum” over the past year.
Extract clusters that include new brands and functional keywords. Summarize keyword-specific volumes, brand mentions, and feature-related keyword groups. Provide insights estimating market growth signals.
Based on the search behavior of users looking up "eye health supplements," segment them into at least 3 personas by age group. For each persona, list frequently used keywords, common concerns, typical expressions, and monthly search volume. Briefly explain the lifestyle context behind their searches.
Use the Cluster Finder API to analyze keywords searched by users of “baby cosmetics.” Identify and categorize CEPs, and create groupings and scenarios based on representative keywords, monthly volume, and search sequence. Visualize the results using Plotly and output in HTML.
ChatGPT's responses may contain inaccuracies. Additionally, even if the actual data call fails or no data is available, it may still generate a natural-sounding response.
Therefore, please make sure that responses are truly based on data retrieved via the API. We strongly recommend validating the results before using them for important decisions or analysis.