Use a datafeed as product feed for the ChatGPT product search

Much line Google Shopping, OpenAI offers sellers possessing an own online shop the option to provide product data for ChatGPT product search. This chapter explains which steps need to be carried out beforehand and which mandatory fields need to be included in order to create a product feed that complies with the OpenAI product specifications. Furthermore, this chapters shows a simple example of how a datafeed export file can be used as the basis for a product feed file.

Preconditions

  1. Register as merchant: Sellers must register on the "Merchant Portal" before they can submit a product feed. To do this, they must enter their basic company details on the "Merchant Portal"; once the approval process has been successfully completed, they will be activated as a merchant.
  2. Preparing the product feed: In this step, an upload file is created in .csv or .jsonl format and compressed into a .gzip folder. Please note that the file must contain all mandatory fields*.
  3. Submit/upload the product feed: In this step, the upload file is transfered to OpenAI via SFTP, via file upload or via a hosted URL**.
  4. Ingestion and indexing at OpenAI: In this step, the upload file is processed by OpenAI, the data is validated and the products are indexed so that they can be shown in the ChatGPT product search***.

Note
*According to OpenAI, the mandatory fields guarantee the correct display of prices and availabilities, whereby recommended attributes such as rich media or product reviews improve the ranking, the relevance and the confidence of the users. Thus, it is recommended to provide as much optional data as possible.

Note
**Updates should be upload frequently so that prices and stock numbers remain current. At this point in time, OpenAI accepts one upload per day.

Note
***Make sure that the robots.txt file does not block OpenAI.

Mandatory fields

The following table describes all the mandatory fields which need to be filled to meet the minimum requirements for a product feed.

Field name Meaning Mandatory?

OpenAI Flags
is_eligible_search The product is available in the ChatGPT search. Yes (The default setting is true.)
is_eligible_checkout The product can be activated for the checkout. Yes (The default setting outside the US is false, as the checkout is yet to activated there.)

Basisproduktdaten
item_id Unique product ID (e.g. SKU) Yes
title Product title Yes
description Product description in plain text format (no HTML) Yes
url URL of the product detail page Yes

Product information
brand Brand name Yes

Media
image_url URL of the main image Yes

Prices and promotions
price Price and currency (e.g. 79.99 USD) Yes

Availability
availability Stock status (in_stock, out_of_stock, preorder, backorder, unknown) Yes
availability_date Date of availability for the type preorder Yes (for the type preorder

Variant information
group_id The "group_id" value should reflect how the product (product_id) is displayed on the retailer’s website (the canonical product page or the parent listing shown to the customer). If variant lines (e.g. by color or size) are submitted, the same "group_id" value must be specified for each variant. Individual variant SKUs should not be submitted without a group ID. Yes
listing_has_variations Indicates whether the product has variants Yes

Seller information
seller_name Name of the seller Yes
seller_url URL of the shops Yes
seller_privacy_policy URL of the privacy policy Yes
seller_tos URL of the terms and conditions Yes

Return information
return_policy URL of the return policy Yes

GEO information
target_countries Target country the product is sold to (currently only US) Yes
store_country Country from where the product is sold (currently only US) Yes

Use case

Using a simplified examle, this chapter explains how create a datafeed export in the .csv format. In this example, the attributes of the exported item data from the iPIM system are mapped to the appropriate mandatory fields for product feeds.

Create a datafeed and make the basic settings

In this step, the datafeed is created and the basic settings are made:

  1. Access the iPIM Export module and go to the “Datafeed Exports” section.
  2. In the “Datafeed Exports” card, click on the plus button.Chevron rightA new datasets with empty input fields is created. Only the “Datafeed ID” fields contains a datafeed ID that was created by the system.
  1. Enter a datafeed identifier in the “Datafeed ID” field .
  2. Select the option “iPIM Standard datafeed” from the “Type” menu .
  3. Select the desired taxonomy from the “Version” menu . Here you select the taxonomy that contains the items to be exported
  4. Select the desired context from the “Context” menu . Here you select the context level on which the item data to be exported is stored.
  5. Select the desired data locale from the “Locale” menu .
  1. Select the option “Item” from the “Export area” menu .
  2. Select the option “Valid products” from the “Validity range” menu .
  1. Enter a file name as well as a file extension in the “File name” field When entering the file name, add the placeholder {0} to the file name so that a time stamp for the point in time of the export is included in the file name of the export file..
  2. Enter a date format in the “File name date format” field (e. g. yyyy-MM-dd).
  3. Aus dem „Transfer profile“-Menü a transfer profile auswählen. Here you select a transfer profile for which the transfer type “Download” is specified.
  4. Aus dem „Export format“-Menü die Option „CSV“ auswählen. Chevron rightAdditional input fields and menus are shown.
  5. Select an encoding from the “Encoding” menu .
  6. Enter a separator in the “Separator” field .
  1. Click on the SAVE button in the lower right corner to save your input.Arrow to the right A success message is shown. Following this, you can proceed to the next step.

Create attribute mappings

In this step, the attribute mappings for a basic selection of item data are created. The mappings are created in a way the column headers in the export file correspond to the respective mandatory fields for the product feed file.

Note
This section merely shows a selection of mappings to illustrate the principle.
  1. Go to the “Attribute mappings” tab.
  1. Create the first attribute mapping. This is the mapping for the item number:
    1. Select the option “Standard-Infos” from the “Mapping type” menu .
    2. Click on the Hinzufügen button to create an empty mapping.
     
  1. Configure the mapping:
    1. Select the option “Item number” from the “iPIM value” menu .
    2. Select the option “Constant” from the “1st Target column configurations” menu .
    3. Enter the value “item_id” in the “1st Target column” field .
     
  2. Click on the SAVE button in the lower right corner to save your input.Arrow to the right A success message is shown.
  1. Create all the remaining attribute mappings:

Create a filter

In this step, a category filter for the category “Toaster” is created so that only the data of items from this category is exported.

  1. Go to the “Filter” tab.
  2. Scroll down to the “Category filter” card.
  1. Click on the CREATE to call up the corresponding dialog box. The dialog box shows the category directory of the selected taxonomy.
  1. Choose the desired category from the category directory.Optional: Enter a search term in the search field to filter the list of categories by a search term.
  2. Tick the category to select it.
  3. Click on the SPEICHERN button to das Erstellen des Kategoriefilters abzuschließen.
    The dialog box closes and a success message is shown. The newly created category filter is shown in the table.
  1. In the table row of the category filter, doube-click in the cell of the “Whitelist” column.
  2. From the menu, select the option YES to put the category on the whitelist so that items from this category are selected.
  3. Click on the SAVE button in the lower right corner to save your input.Arrow to the right A success message is shown.

Execute the datafeed export

In this step, the datafeed export is carried out.

  1. In the “Datafeed exports” card, choose the datafeed from the list of datafeeds. Optional: Enter a search term in the search field to filter the list of datafeeds by a search term.
  2. Move the mouse cursor over the datafeed.Chevron rightThe Action menubutton is shown.
  3. Click on the Action menubutton to show the action menu.
  1. In the action menu, click on the option “Dateiexport Export” to trigger the export.Chevron rightA process message is shown. Once the export is complete, the export file can be downloaded via the “Tasks” menu which is hidden behind the Tasksbutton in the header.