Five things in ChatGPT Ads Manager you cannot undo

Account country, currency, and timezone lock at confirmation. Negative keywords cap at 25. Landing pages must be reachable by OpenAI's crawlers. Identity verification can take days. Feed and creative fields are validated on upload, not after.

ChatGPT Ads Manager opened to self-serve advertisers in May 2026 and, if you have used Google Ads or Meta, the interface will feel familiar and simpler. That familiarity is the problem. A few decisions behave differently here than they do on platforms where almost everything is editable later.

None of these are difficult to get right. All of them are annoying to get wrong.

1. Country, currency, and timezone lock permanently

Setup runs in four steps: business details, account configuration, identity verification, and billing. At the configuration step you set country, currency, and timezone, and once you confirm, those are fixed. Changing them means opening a new account and abandoning whatever history the first one accumulated.

For a single advertiser in one market this is a five second decision. For anyone else it deserves a conversation first:

• Agencies setting up for a client. The account should reflect the client's advertiser identity, country, and billing, not the agency's. Get this wrong and you have built the client's permanent ad account under the wrong entity.
• Multi-market advertisers. Currency affects how every historical performance figure reads. Picking the wrong one makes your own reporting harder forever.
• Timezone. Daily budgets reset against it and every daily report is bucketed by it. Match it to how the business already reads its numbers, not to where the person doing the setup happens to be sitting.

2. Negative keywords cap at 25

If you come from search, this is the number that will surprise you. Twenty-five, total.

The cap makes sense once you remember there is no keyword bidding here. Targeting is topic, intent, and conversation context, so negatives are a blunt corrective rather than the primary control surface. But the practical effect is that you cannot port a mature search exclusion list, and you should not try.

Spend them on high-cost, clearly unqualified patterns: the free and DIY intent that will never convert for you, the competitor or job-seeker traffic that eats budget, the adjacent product category you keep getting pulled into. Then leave room, because you will learn more in week three than you knew at launch.

3. Your landing page has to be reachable by OpenAI

Landing page URLs must be valid, reachable, and not blocked to OpenAI's crawlers, and image URLs must be publicly accessible. This is where a technically perfect campaign fails for a reason that lives entirely outside the ad platform.

The usual culprits are a robots.txt rule blocking OpenAI user agents, a bot-protection or WAF layer challenging non-browser traffic, or images served from a storage bucket that is not actually public. Plenty of sites added AI crawler blocks during 2024 and 2025 as a content-protection measure and never revisited the decision. If that is you, the block and the ad program are now in direct conflict, and it is worth resolving deliberately rather than discovering it as a disapproval.

Check this before you build campaigns, not after an ad fails review.

4. Identity verification takes days, not minutes

Verification runs through a third-party identity provider and has been slow because of signup volume. It is not a step you can compress, and it sits between you and a live campaign.

The only real mitigation is sequencing. Start verification first, then do the work that does not depend on it while you wait: conversion tracking, feed preparation, creative, and question mapping. Teams that treat setup as linear lose a week staring at a pending status with nothing built.

5. Creative and feed fields validate on upload

Ads Manager checks required fields when you upload rather than flagging problems later, so incomplete creative and malformed feed rows fail at the door. For feed-driven campaigns this is the substantive one, because the carousel format pulls from your retailer feed and feed data quality carries more weight here than any targeting setting.

Get the feed clean first. A well-structured campaign pointed at a feed full of out-of-stock items and missing attributes is not a campaign, and no amount of bid management fixes it.

The order we run it in

1. Confirm category eligibility, before anything else, because for some advertisers this ends the project.
2. Decide country, currency, and timezone deliberately, in writing, with the client.
3. Start identity verification immediately, then work in parallel while it processes.
4. Audit the landing pages for crawler access and the images for public reachability.
5. Build and validate conversion tracking, and confirm it with a live click.
6. Clean the feed.
7. Write creative, then launch, then spend the 25 negatives as you learn.

Nothing here is clever. It is just the difference between launching once and launching twice.