7 Things to Check Before OpenAI Starts Rewriting Your ChatGPT Ads

On October 20, 2026, OpenAI emailed ChatGPT advertisers to say that from the end of the month it will enable "creative enhancements" on new and existing self-serve campaigns. Once the setting is on, OpenAI can rewrite your headlines and descriptions, resize your images, send clicks to a different page on your site and translate your copy into other languages. The variations can run without you reviewing or approving each one.

The email gives Friday, October 30 as the start date. OpenAI's help page says Thursday, October 29. Either way, the setting is already toggled on in Ads Manager, and the only way to keep your ads as you wrote them is to turn it off before then.

Default-on creative automation is familiar from Meta's Advantage+ and Google's AI Max. ChatGPT is a different surface. Your ad does not sit in a feed. It appears directly under an answer that ChatGPT has just written for the user, and the rewrite will be built from that conversation. Here are seven things to check before the switch flips.

Key Takeaways

1. Decide your default before October 29, not after

The setting has two layers. The account-level toggle in Manage settings is the default for every campaign that has no setting of its own, and a campaign-level setting takes priority over it. That means you do not have to choose between all on and all off. You can set the account default to off and switch it on for the campaigns where you are willing to test, or leave the default on and protect specific campaigns.

What you should not do is leave the decision until the first November report. By then the enhanced impressions are already blended into your numbers, and there is no way to separate them afterwards.

Action: Open Ads Manager, go to Manage settings and record the current state of the toggle for the account and for every campaign. Decide the default now. If you are unsure, off at the account level with a deliberate opt-in per campaign is the reversible choice.

2. Audit every page the rewrite can draw from

The rewrite is sourced from your landing page as well as your ad. Whatever is on that page is now candidate ad copy: last quarter's promotion, a price that changed, a feature you no longer sell, a testimonial you are no longer licensed to use. On the same surface, destination-URL optimization means a click can be sent to another page on your site that OpenAI judges relevant to the conversation. OpenAI has not published how that page is chosen.

So the question is not only whether your landing page is accurate. It is whether every page a user could be sent to is one you would have chosen: old campaign pages still live, collection pages with out-of-date stock, regional pages with the wrong currency, documentation written for a different audience.

Action: List the destination URL of every active ad, then the pages near it that could plausibly be selected for the same intent. Check price, promotion wording, availability and claims against what the ad promises. Retire or update the ones that fail. Treat the rest of the site as creative input, because from October 29 it is.

3. Ring-fence regulated and approved copy

If your ads carry wording that legal, compliance or a partner signed off, an automated rewrite can run without that wording. OpenAI has not said whether enhanced versions go through the same policy review as the original, or how liability is allocated when a generated headline makes a claim you did not. In financial services, healthcare, insurance or anything with a mandatory disclaimer, that is not a risk to take on a default.

The same applies to positioning you have worked to establish. If your brand deliberately avoids a category label, a rewrite tuned to match the user's conversation may put it straight back in.

Action: Move any ad with approved, contractual or regulated wording into its own campaign and set creative enhancements to off at the campaign level. Keep that separation permanently so a future account-level change cannot reach it.

4. Tighten your context hints

ChatGPT Ads match on conversation themes, not keywords, and context hints are how you tell OpenAI which situations your ad belongs in. With creative enhancements on, those same hints shape the conversations the rewrite is built for. Broad hints give the rewrite room to drift: an enterprise security ad group with a hint about "software problems" can produce a headline that reads like a consumer help desk, because that is what the matched conversation was about.

Specific hints do the opposite. They narrow the conversations your ad enters, which narrows the material the rewrite has to work with, which keeps the generated copy closer to the offer you actually sell.

Action: Review every ad group's context hints. Replace single-word or generic hints with specific buyer situations. If you cannot describe the conversation you want the ad in, do not let the rewrite guess it for you.

5. Check translation against what your site can support

ChatGPT Ads are now sold in 89 countries after the 26-country expansion OpenAI announced on October 21, and creative enhancements include translating ad text into other languages. Translating the ad does not translate the landing page. A German headline that lands on an English-only checkout, or a translated promotion that your support team cannot honour in that language, is a worse outcome than no ad at all.

This is only a benefit where you already have localized pages, localized pricing and someone who can handle the lead. Everywhere else, it is a leak.

Action: For each campaign, confirm whether the targeted markets have localized destinations. Where they do not, either narrow the targeting or set enhancements off for that campaign.

6. Freeze a baseline and fix your tracking before the switch

Because Ads Manager will not separate enhanced impressions from original ones, the only way to judge the feature is a before-and-after comparison you set up yourself. That needs a clean baseline captured before October 29 and tracking that survives a changed destination.

The second point is the one most accounts miss. If your UTM parameters are typed onto a specific landing page URL and OpenAI sends the click to a different page, the parameters do not travel with it. Attribution breaks silently, and the swapped page looks like direct or organic traffic in your analytics.

Action: Export campaign, ad group and ad reporting through October 28 and record impressions, clicks, spend, click-through rate, cost per click, and attributed conversions per campaign. Apply UTM tracking at the campaign or account level so it attaches to whichever page is served, and make sure your conversion measurement is firing on every page a click could land on, not only the designated landing page.

7. Keep a control, and judge it in weeks, not days

OpenAI says the feature is designed to improve performance and has published no performance data. The honest position is that nobody outside OpenAI knows yet whether it helps, and it will differ by category. The way to find out in your account is the same way you would test any other default: one campaign on, a matched campaign off, the same budget and the same context hints.

OpenAI's own help page warns that performance changes over time do not by themselves prove the feature's effect. Seasonality, auction density and the 26 new markets will all move your numbers in the same weeks. Give the comparison enough time to separate those from the setting.

Action: Pair campaigns as test and control before the switch. After it, check your analytics for which landing pages are receiving ChatGPT ad traffic, and look at conversion quality rather than conversion count. If a page you did not choose is receiving clicks, fix the page or turn enhancements off for that campaign.

Why this matters more in ChatGPT than on Meta or Google

A ChatGPT ad is read in the context of the answer above it. If ChatGPT has just recommended three brands and yours is not one of them, your sponsored card is arguing with the assistant. If yours is one of them, the card confirms what the user was already told. Creative enhancements make that relationship tighter, because the rewrite is now built from the same conversation.

Whitebox's monitoring of the OpenAI Ads surface, covering 153,514 sponsored placements across 7,171 questions and 3,352 advertisers, found that only 16.4% of the brands ChatGPT names in its answers have ever advertised on the surface beneath, and that 61.1% of brand-name questions do not surface that brand's own ad. 84.2% of repeat asks of the same question return a different advertiser. The surface is still open and still volatile, which is exactly when you want to know what your ads say and where they send people.

The same goal, a different kind of automation

OpenAI's premise is right: an ad inside a conversation should fit that conversation. The problem with creative enhancements is not the ambition, it is the mechanism. The rewrite is built inside the platform that sells you the impression, you cannot see the conversation it was built from, you cannot see the version that ran, and the only lever you are given is a single on-off switch.

Whitebox was built to do the same job from the other side of the table. It sits on top of OpenAI Ads Manager and automates the work of matching ads to conversations, but every step is visible, every change is yours to approve, and every result is measured against a control. Three things make that different from leaving the toggle on.

You see the conversation before you write for it

Ads Manager shows you a chat card and aggregate numbers. Whitebox shows you the prompts. It maps each campaign, ad group and ad to the real ChatGPT conversations it matched, scores how well the copy fits each one, and shows which context hints are pulling the ad into conversations it does not belong in. That is the input OpenAI's rewrite uses and does not show you. Whitebox generates context hints and creative variants from those conversations too, but as drafts you read before they run, not as versions that go live on their own.

Automation you can audit

Whitebox automates bids, budgets and creative changes through rules with your thresholds, and every change it proposes sits in a review queue with the reasoning attached until someone approves or reverts it. Variations are run as structured A/B tests with a control, which is exactly the comparison Ads Manager cannot give you for enhanced creative. The platform also flags the gaps that make check 6 fail, untracked destination URLs and missing conversion integrations, and applies UTM tracking across the account in one step so attribution survives whatever page a click lands on.

The answer above the ad, in the same view

This is the part no ad platform offers, because no ad platform monitors its own organic answers for you. Whitebox is a Generative Engine Optimization platform first. It tracks whether ChatGPT, Gemini, Perplexity and Google AI Overviews name and recommend your brand, which sources they rely on to do it, and how that changes week to week. For an advertiser, that means you know whether your sponsored card is confirming what the assistant just said or arguing with it, and you can fix the organic side with content rather than paying to contradict it. The 153,514-placement market report above exists because Whitebox watches the questions, the answers and the ads together. That is the view you need before you let any system, OpenAI's or ours, rewrite what your brand says inside a conversation.

Frequently Asked Questions

What are creative enhancements in ChatGPT Ads?

A setting in OpenAI Ads Manager that lets OpenAI generate variations of your ad: new or updated headlines and descriptions, resized images, a different destination URL on your site, and translations into other languages. Variations draw on the conversation context or user interests, your existing ad and your landing page, and can run without you reviewing or approving each version.

When does OpenAI turn creative enhancements on by default?

OpenAI's help page gives Thursday, October 29, 2026. The email sent to advertisers on October 20 gives Friday, October 30, and refers to self-serve campaigns. Opting out before October 29 covers both dates.

How do I opt out of creative enhancements?

In OpenAI Ads Manager, open Manage settings and turn the creative enhancements toggle off. The toggle sets an account-level default, and campaign-level settings override it. Existing account and campaign opt-outs are preserved, and you can change the setting at any time.

Will Ads Manager show which rewritten version ran?

No. Reporting stays at the campaign, ad group and ad level, with no split between enhanced versions and the original. OpenAI also says you cannot preview every variation, so your own web analytics is the only place a swapped landing page or a translated click will show up.

See which conversations your ChatGPT ads are matching, and what the answer above them says, before October 29.

Book Now