ChatGPT is testing visual ads during image generation, plus new measurement partners

ANNOUNCEMENT

OpenAI is adding a new kind of ad to ChatGPT. It is a picture-based ad, and it will first show up while ChatGPT is making an image for you. OpenAI announced it on October 5, 2026, together with a long list of new measurement partners for advertisers.¹

Testing starts later in October 2026, in the US, with a small first group of advertisers. Below is what OpenAI actually confirmed, what the measurement news means in plain terms, and how to stop these ad clicks from getting mixed into your organic ChatGPT numbers.

What OpenAI announced on October 5

CONFIRMED, by OpenAI directly:

• A new visual ad format. Ads will use images showing "product inspiration, product usage, or the experiences they make possible."¹
• It starts in image generation. OpenAI says: "Initially, we'll test this new ad format during image generation in ChatGPT."¹
• Testing timing and place. "Testing will begin later this month in the US with an initial group of advertisers."¹ OpenAI did not name those advertisers.
• New measurement partners. These cover data platforms, attribution, and incrementality (more on what those words mean below).¹
• Brand safety pilots with DoubleVerify (DV) and Integral Ad Science (IAS).¹

OpenAI also repeated a promise it has made since ads began:

"Ads will be clearly labeled, and remain separate from the image being created. As with our other ad formats, advertising does not influence the answers ChatGPT provides."

Source: OpenAI, "Building advertising for the way people use AI," October 5, 2026.¹

Early in the post, OpenAI gave its own reach figure: "ChatGPT reaches 1.2 billion people each week."¹

How the visual ad format works

Making an image in ChatGPT takes a little time. The new ad uses that waiting time. While the image loads, an ad with its own picture can appear.

CONFIRMED, by OpenAI directly: the ad is labeled, and it stays separate from the image you asked for.¹ So the brand's product is not placed inside your generated picture. The ad is its own item.

REPORTED, from OpenAI's example image: Search Engine Roundtable described the sample screenshot OpenAI shared. In it, "the image is being generated above, it says 62% done, and there is an ad below that is already loaded and waiting for you to click on."² This shows one example layout. OpenAI's text does not promise that every ad will sit in exactly that spot.

Not stated by OpenAI: which ChatGPT plans will see the visual ads. Earlier ads were limited to the Free and Go plans for logged-in adults.³⁴ The October 5 post does not say whether the new format follows the same rule.

Ripplix covered OpenAI's previous new format, Sponsored Agents, in a separate piece: OpenAI's Sponsored Agents: what ChatGPT's new ad format actually does. This article sticks to the visual format and the measurement tools.

How this fits into the ChatGPT ads rollout

The visual format is the latest step in a rollout that began in January.

The new measurement partners, in one table

Measurement is how an advertiser checks whether an ad worked. OpenAI says its approach is "focused on supporting both in-house and partner-led measurement solutions."¹ In other words, brands can use OpenAI's own tools or the outside tools they already pay for.

A few terms first, in plain words:

• Conversion: the action an advertiser wants, such as a purchase or a sign-up.
• Attribution: deciding which ad gets credit for a conversion.
• Incrementality: checking whether the ad caused extra sales that would not have happened anyway.

All rows: OpenAI, October 5, 2026.¹

A geo-based experiment compares places. Ads run in some regions and not in others. Then you compare sales between the two groups. That is a common way to test whether ads caused extra sales.

The advertiser results, and how to read them

OpenAI shared three early results. Each one was reported by a measurement partner. These are results published by OpenAI and its partners about their own platform, not independent studies. That is worth knowing before you compare them with your own numbers.

Source: OpenAI, October 5, 2026.¹

These three numbers measure different things. So they should not be read as one story.

• WeightWatchers is about cost. "Cost per acquisition" means how much the brand paid, on average, for each new customer. The number depends on how credit was assigned. It compares ChatGPT Ads with a "blended paid-search benchmark," which mixes several search ad sources.
• Dose is an incrementality result. It says the ads caused a measurable lift. Of the extra purchases, 67% came from customers who were new to the brand.
• Portland Leather is about who clicked. 93% of visitors from ChatGPT Ads were new. It does not say anything about sales or cost.

OpenAI's post gives no time period, budget, or sample size for any of the three.¹ So treat them as early examples, not benchmarks.

Brand safety: Negative Phrases and the DV and IAS pilots

Brand safety means making sure an ad does not appear next to content a brand would not want to be seen with. OpenAI announced two things here.

First, a control for advertisers. OpenAI says "Negative Phrases are now available to qualifying advertisers," for "narrow placement requirements specific to a brand's own policies."¹ OpenAI also says its "placement guardrails assess whether a conversation is appropriate for advertising, helping keep ads out of emotionally vulnerable, sensitive, or otherwise unsuitable contexts."¹

Second, outside checks. OpenAI is "developing brand suitability evaluation pilots" with DV and IAS. OpenAI says this lets them check its standards "in a controlled testing environment, without accessing private user conversations."¹

Both company leaders were quoted in OpenAI's post. DoubleVerify CEO Mark Zagorski said:

"ChatGPT is creating a fundamentally different advertising environment, where context is defined by the conversation itself rather than a page or video."

Source: OpenAI, October 5, 2026.¹

That is a fair summary of the problem. On a website, a brand safety tool can check the page. In ChatGPT, the "page" is a private chat. That is why these pilots use a test setup instead of reading real chats.

Why this matters for AI visibility reporting

Many brands now track how much traffic ChatGPT sends them. They often treat this as a sign of "AI visibility," meaning how often AI tools mention or recommend them. Paid ads make that number harder to read.

Here is the problem in simple steps:

1. For organic (unpaid) links, OpenAI says "ChatGPT automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs."⁸ A UTM parameter is a small tag at the end of a link. Analytics tools read it to tell where a visitor came from.
2. Ad clicks carry their own tag. OpenAI's help center says its click reference, "called oppref, is appended to the end of the landing page URL."⁹
3. If your reports group all visits from chatgpt.com together, paid clicks and earned visits end up in the same bucket.
4. So a rise in "ChatGPT traffic" could come from your ad budget, not from AI answers recommending you more often.

A new ad format can add a burst of clicks. Without a split, that burst looks like organic growth.

Checklist: keep ChatGPT ad clicks separate from organic referrals

Use this list before your first ChatGPT ad goes live.

• Add your own UTM tags to every ad URL. OpenAI Ads Manager lets you add "custom landing page query parameters using your own text or supported macros," including {campaign_id}, {ad_group_id}, {ad_id}, and {ad_account_id}.¹⁰ Use a paid medium such as utm_medium=cpc so analytics files these visits as paid.
• Build a "ChatGPT paid" segment based on oppref. Any visit whose landing URL contains oppref came from an ad click.⁹ This works even if a UTM tag goes missing.
• Build a "ChatGPT organic" segment. Include visits from chatgpt.com or with utm_source=chatgpt.com, and exclude anything with oppref or your paid UTM tags.
• Keep oppref through redirects. OpenAI tells advertisers to "preserve the OpenAI click reference (oppref) through redirects and landing-page navigation."⁹ If a redirect drops it, an ad click can look organic.
• Report the two numbers side by side. Show organic ChatGPT referrals and paid ChatGPT clicks as separate lines. Never add them into one "AI visibility" figure.
• Note the start date of any campaign. Mark it on your traffic charts. Then a jump on that date is easy to explain.
• Expect totals to differ between tools. OpenAI says Ads Manager, "your analytics provider, and other ad platforms may report different conversion totals."⁹ Pick one source of truth for each report and say which one.
• Check the reporting window. Ads Manager offers "a 7-, 14-, or 30-day click-through reporting window" and an optional 1-day view-through window.¹⁰ A view-through conversion counts someone who saw the ad but did not click. Your analytics tool will not see those, so the totals will not match.
• Keep measuring organic AI answers directly. Referral traffic only shows clicks. Whether ChatGPT mentions your brand in its answers is a separate question, and ads do not change it, according to OpenAI.¹

Open questions

Some things are still unclear from what OpenAI has published so far.

• Which plans will see visual ads? The October 5 post does not say. Earlier ads ran only on Free and Go.⁴
• Who is in the first test group? OpenAI said "an initial group of advertisers" but named none.¹
• When will it expand beyond the US or beyond image generation? OpenAI says it will "test this new ad format during image generation" first.¹ No later dates were given.
• How will results be shared from the DV and IAS pilots? OpenAI calls them pilots that it is still developing.¹ It has not said when or how findings will be published.
• Will the advertiser results hold up at scale? The three examples have no stated time period or sample size.¹

See what ChatGPT says about your brand when no one is paying

Ads in ChatGPT are about to get more visible. That makes it more important to know where you stand in the unpaid answers. Ripplix tracks how AI tools like ChatGPT mention, describe, and cite your brand, separately from any ad spend. Get your free AI visibility report.