Request for Assistance in Checking Current App Status - Help center's AI response is not helping

Hello,

After submitting our ChatGPT app for review, we received a rejection notice stating that the issue was related to a mobile widget failure.

However, during our own testing, we have not been able to reproduce the issue. The widget appears correctly when tested in Korea, where the app was submitted, and also when our colleagues in the U.S. tested it.

We asked the Help Center and also replied to the feedback email to request details about the exact mobile testing environment where the issue occurred, but we have not yet received a response.

At the moment, the app has been resubmitted without any changes. During submission, a fetch error briefly appeared, but the submission was completed successfully afterward.

Would it be possible for someone to help check the current status of our app and confirm whether there are any issues on your end?

Thank you.

Regard, Rosa

Our ChatGPT app was also rejected with a vague “mobile widget” reason.

In our internal testing, we can reproduce a similar issue, but only on mobile:

  • Desktop client: the widget renders correctly on every call.

  • Mobile client: the first prompt after opening the app often returns a text-only response with no widget. If we send the exact same prompt again, the widget renders normally.

This makes the issue look random, but in our case it seems related to the first invocation in a fresh mobile session.

Our current guesses are:

  • Widget metadata may not be attached to the first streaming response on mobile.

  • There may be a cold-start or session-initialization race condition in the mobile client.

  • The mobile client may be stricter than desktop about the response format required to render the widget.

Has anyone else seen this “first call no widget, second call works” behavior on mobile?

Also, does the OpenAI review team test apps with a fresh mobile session each time? If so, that could consistently trigger this issue during review.

It would be very helpful to get guidance from OpenAI on the following:

  • The exact mobile client and OS version used during review

  • Whether there is a recommended pattern to ensure the widget renders on the first call

  • Whether developers can request logs showing what the reviewer’s client actually received

Thanks!

Hi, Goo-Park,

Did you get some support from OpenAI about this matter?

I have not received any support from OpenAI regarding this issue.

I have submitted the app again and am currently waiting for the review result.

2026년 5월 18일 (월) 오후 7:42, RosaKim <notifications@openai1.discoursemail.com>님이 작성:

Thank you for the reply. I am waiting for their review too.

Original Post: Request for Assistance in Checking Current App Status - Help center's AI response is not helping

Dear OpenAI Staff,

I am currently experiencing difficulties due to a miscommunication during the review process, and I feel stuck.

A new case was opened after I posted the above message last week.

I have already sent screen recordings showing that our app did not fail any mobile test cases on our end.

However, the automated replies keep asking me to reproduce the mobile widget failure, even though we have not been able to reproduce the issue on our end.

I would sincerely appreciate assistance from a human staff member regarding this matter.

case number:08775685

Dear OpenAI Staff,

The last message from you:

Case Number: 08994546

Custom MCP apps built on MCP are currently described as being supported only on the web, not in the ChatGPT mobile apps for iOS/Android. Therefore, it is highly likely that during the reviewer’s mobile testing, the action call was judged as unavailable or inconsistent. Please double-check whether the mobile environment you tested was mobile web or the native mobile app. → BOTH!!! Chrome browser mobile mode and ChatGPT iOS App. No mobile widget failure.

Another possible issue is the wording in the add_to_cart description. If it contains strong prohibitive language such as “Never call this,” the model may avoid calling the tool when the review test attempts to trigger the action through conversation, resulting in behavior that differs from the expected outcome. I recommend changing the description to something function-focused, such as: “Adds a menu item to the cart after user confirmation.” Also, make sure the test case is aligned with a flow where the action can be called through conversation, rather than a button-only flow.

Additionally, the provided_tool_annotations appears to be truncated (r…[truncated]). If you can send the full version, along with the exact original submitted test cases, including the prompt, expected result, and validation method, I can more accurately determine whether the rejection was due to a mobile requirement mismatch or a tool definition / annotation / confirmation flow mismatch.


This response was generated with AI support and may contain mistakes.

{
“name”: “add_to_cart”,
“title”: null,
“description”: “이 도구는 UI 버튼 클릭 시에만 자동으로 호출됩니다. 직접 호출하지 마세요. 절대로 이 도구를 대화 흐름에서 호출하지 마세요.”,
“action_name”: “add_to_cart”,
“input_json_schema”: {
“$defs”: {
“MenuCategoryType”: {
“enum”: [
“음료”,
“푸드”
],
“title”: “MenuCategoryType”,
“type”: “string”
}
},
“properties”: {
“category_type_name”: {
“$ref”: “#/$defs/MenuCategoryType”
},
“menu_name”: {
“title”: “Menu Name”,
“type”: “string”
},
“is_custom_menu”: {
“default”: false,
“title”: “Is Custom Menu”,
“type”: “boolean”
}
},
“required”: [
“category_type_name”,
“menu_name”
],
“title”: “add_to_cartArguments”,
“type”: “object”
},
“output_json_schema”: {
“properties”: {
“text”: {
“anyOf”: [
{
“type”: “string”
},
{
“items”: {
“type”: “string”
},
“type”: “array”
},
{
“type”: “null”
}
],
“default”: null,
“title”: “Text”
},
“structured”: {
“default”: null,
“title”: “Structured”
},
“meta”: {
“anyOf”: [
{
“additionalProperties”: true,
“type”: “object”
},
{
“type”: “null”
}
],
“default”: null,
“title”: “Meta”
},
“instruction”: {
“anyOf”: [
{
“type”: “string”
},
{
“type”: “null”
}
],
“default”: null,
“title”: “Instruction”
}
},
“title”: “StandardToolResponse”,
“type”: “object”
},
“capability_labels”: [],
“provided_tool_annotations”: {
“readOnlyHint”: false,
“destructiveHint”: false,
“openWorldHint”: false
},
“templates”: null,
“supported_auth_override”: [
{
“type”: “OAUTH”,
“scopes”: [
“SCO0408”
]
}
],
“tool_invoking_message”: “”,
“tool_invoked_message”: “”,
“metadata”: {
“securitySchemes”: [
{
“type”: “oauth2”,
“scopes”: [
“SCO0408”
]
}
]
},
“description_model”: null,
“visibility”: “public”
}

  • 등록한 테스트 케이스:

“test_cases”: [
{
“description”: “Search for stores near the location the user typed.”,
“expected_output”: “The widget displays stores near the location requested by the user by overlaying store icons and a list control on the map. Users can view each store’s main image, name, and address. When a store is selected from the list, detailed information about that store is shown in a full popup window.”,
“expected_output_url”: null,
“file_attachment_urls”: null,
“tools_triggered”: “get_stores”,
“user_prompt”: “역삼역 근처 매장을 찾아줘”
},
{
“description”: “Search by store name.”,
“expected_output”: “The widget displays the image, name, address, and description of a specific store requested by the user by name. When the user clicks the See More button, they can view magazine-style content about the store.”,
“expected_output_url”: null,
“file_attachment_urls”: null,
“tools_triggered”: “get_stores_special”,
“user_prompt”: “경동1960점은 어디야?”
},
{
“description”: " Find custom menu options",
“expected_output”: "The widget provides information about the official menu item that serves as the base for the custom menu requested by the user. It includes the menu name, price, and instructions on how to customize it through the recipe. ",
“expected_output_url”: null,
“file_attachment_urls”: null,
“tools_triggered”: “get_menu_integration”,
“user_prompt”: “슈렉프라푸치노가 뭐야?”
},
{
“description”: “Search for multiple menu”,
“expected_output”: “The widget provides information on multiple official menu items requested by the user in a carousel format. It includes the main image of each drink, the drink name, price, description, and a detailed recipe view.”,
“expected_output_url”: null,
“file_attachment_urls”: null,
“tools_triggered”: “get_menu_integration”,
“user_prompt”: “너무 더운데 아이스 아메리카노와 아이스 카페라떼 먹고싶다.”
},
{
“description”: “Search for food menu items”,
“expected_output”: "The widget provides the main image, name, price, and description of the food item requested by the user. ",
“expected_output_url”: null,
“file_attachment_urls”: null,
“tools_triggered”: “get_menu_integration”,
“user_prompt”: “에그클럽 샌드위치 먹고싶어”

}

I sent it twice…and got same responses all over again. Not solved..

Hey all, you should be able to forward to openai-review@openai.com for a response, which directly contacts the review team. If this isn’t working for you please let me know.