Hi everyone — we’ve shipped a batch of improvements to make app submission and review smoother for developers and workspace admins. We really appreciate all the feedback you’ve given!
Clearer submission errors: submitters should now see more user-facing validation errors directly in the submission flow. For example, token-limit errors on tool metadata are now surfaced in the submission flow.
Submission wizard fix for MCP tool validation: we fixed an issue where the wizard could hold onto stale validation errors after an MCP URL or tool form changed.
Required tool annotation explanations: the submission form now requires tool annotation explanations before submission, which should prevent avoidable review rejections caused by missing scope justification.
Better resubmission behavior: if an app is moved back to draft and resubmitted, it now correctly creates a fresh reviewable item instead of getting stuck in limbo.
Less repeated review for open iframing: once an app has already been approved for open iframing, future version submissions that include frame_domains no longer get sent back through the sensitive review path just for that reason.
Improved version ordering: app versions with the same semver now sort correctly, which should reduce confusion when comparing published and rejected versions.
More granular submission permissions: organizations can now grant app-submission access through fine-grained role permissions instead of requiring broader org-owner access.
Overall, these updates are aimed at reducing avoidable submission failures, cutting down repeat review work, and making the submission experience more predictable for both developers and admins. As we continue building out Apps in ChatGPT and Codex, please keep flagging to us what’s frustrating or delightful.
We received a rejection, and I’d appreciate some clarification on the specific reasons behind it. Could you please indicate which tests did not pass and provide details on the failures?
I verified the test cases both before submission and again afterward, and they appear to be working correctly on my end. Any additional insight would help me understand the issue and address it more effectively.
Agree with those last two posts 100%. Obviously the reviewer has this info - to not provide it and make us guess is part of what makes the current process maddening.
Thank you for your help Casey - it’s hard to explain how frustrating this is for us developers. I’ve been working on Apps since December and really want to see this succeed.
Bolding this a bit to make sure it’s clear guidance. While we look into getting more rejection information into the emails, the official guidance is that when you receive a rejection, you can reply to the email to get more information about why the rejection occurred. Hopefully this helps you get apps out faster!
An error occured while trying to deliver the attached email.
You may wish to report the following information to the owner of trustandsafety @5-2wbwtbbgk0mumkv30qsh7hmdf8ms6ggckanust9eeor04lip9e.5e-54eijeaq.usa700.case.salesforce.com.
trustandsafety @5-2wbwtbbgk0mumkv30qsh7hmdf8ms6ggckanust9eeor04lip9e.5e-54eijeaq.usa700.case.salesforce.com is not a valid address.
Same issue here. I’ve submitted several times and keep receiving the same rejection message:
“One or more of your test cases did not produce correct results. Please re-run all submitted test cases and align tool behavior/output with the documented expected outcomes. Ensure the same test cases pass consistently on both ChatGPT web and mobile.”
I’ve thoroughly re-tested everything on my side, and all test cases return the expected results. Interestingly, I’m not seeing any log entries that would indicate the MCP server is actually being tested during the review process, which makes it difficult to pinpoint what’s going wrong.
This is definitely an improvement point for the app submission process.
@casey-chow It would be great if you could take a look at my case: [C-Kyv3IshBNFSv].
We have a fix implemented and rolling out. For transparency, there were some upstream changes to the moderation process that we had to catch up with. Going forward, any rejection email will have a webform instead. It goes to the same shared inbox as the emails originally did, so there shouldn’t be any drop in response rate.
The developer name you entered does not match your verified individual or business name. Please update the developer name so it matches the verified name on your account. If you would like to publish under a business name but have not yet completed business verification, please complete business verification in Platform Settings
One or more of your tool’s annotations do not appear to match the tool’s behavior. Please confirm annotations are explicitly set to true or false (not null) for every tool. Include a clear justification for why the hint is set that way based on the tool’s actual behavior.
It’s so difficult to understand, I have 2 rejects with these same conditions and I can’t understand which tool have the problem, and which name suppose to be, i try with my personal name and the Business Name, but i have the same reply in these two emails, the same condition in these emails.
Worse than this, now I can’t reply the emails because the system reject the replies!
Message blocked
Your message to openai-review@openai.com has been blocked. See technical details below for more information.
Having the same issue with the email system rejecting my replies. A member of the OpenAI Review Team responded to my email asking for clarification and now my responses are being bounced.
