ChatGPT will rewrite your ads from 30 October. Unless you switch it off.

WHAT OPENAI IS SWITCHING ON

OpenAI emailed its advertisers this week. The email says that starting Friday 30 October 2026, it will "enable creative enhancements for your new and existing self-serve campaigns." Search Engine Roundtable's Barry Schwartz, who advertises there, published the email and a screenshot of the setting, already toggled on in his account.

Creative enhancements means OpenAI uses your ad and your landing page to make new versions of your ad. Four things can change:

1. Headlines and descriptions: new or updated text that links your offer to the conversation or the user's interests.
2. Language: your ad text can be translated into another language.
3. Images: resized to fit the ad format.
4. Destination URL: OpenAI can send people to a different page on your website that it thinks fits better.

The key line: "When enabled, these ads can run automatically without requiring you to review or approve each version." Versions can be prepared in advance or created in real time, based on the chat the person is having.

WHY "LESS WORK" IS THE WRONG TEST

OpenAI sells this as better performance with less work, because you no longer have to write many versions yourself. That is true. It is also the wrong test.

An ad is a promise in your name. When a person reads your ad inside a chat, they do not know a machine wrote that version. If it overpromises, uses words your legal team banned, or sends a buyer to a blog post instead of the product page, the damage is yours.

Paid search buyers reacted fast. Melissa Mackey said her team had already paused its ChatGPT ads test because of poor lead quality, and was glad it had. Anthony Higman's reading was blunter: when an ad platform frames a change as less work for you, it usually means more of your ads showing in places you do not want.

We would not go that far. But nobody should switch on a writer they cannot read.

THE MATHS OF A REWRITE THAT LOOKS GOOD

Here is a made-up example to show the trap.

Say your ChatGPT ads bring 200 leads a month, and 10% of leads buy. That is 20 sales.

After 30 October, say 30% of your leads come from rewritten versions. The rewrites match the chat better, so more people click. But the promise is looser, so those leads buy at 5%.

• 140 original leads x 10% = 14 sales.
• 60 rewritten leads x 5% = 3 sales.
• Total: 17 sales instead of 20. That is 15% fewer sales on the same budget, while your click rate looks better than ever.

• The formula to watch:
• Sales per 100 leads = sales ÷ leads x 100.
• If clicks go up and sales per 100 leads go down, the rewrite is costing you money.

WHAT YOU CONTROL

You can opt out before 30 October or change the setting at any time in Ads Manager. OpenAI says it keeps your existing account and campaign opt-outs, and that campaign settings take priority over the account default. So you can switch it off for the account and allow it on one test campaign.

THE PUNKD TAKE

We like automation. We do not like automation we cannot see. Every ad platform wants more control over your creative, because it can then sell more of your budget into more places. That is fine when you can measure what each version sold. Today you cannot read each version before it runs. So the honest setting is off by default, on by test, and judged on sales. We typically find 18-30% of ad spend doing nothing at all. A writer you cannot read is a good way to grow that number.

STEAL THIS

1. Before 30 October, log into ChatGPT Ads Manager and set creative enhancements for each campaign on purpose.
2. Switch it off at account level and on for one test campaign, with a twin campaign left off.
3. Run both for four weeks with the same budget and audience.
4. Compare sales per 100 leads, not click rate.
5. Check the landing pages your clicks actually reached every week, and block any page you never want as a destination.

Would you let a platform run ads in your name that you never get to read?