On October 5, 2026, OpenAI published a post titled "Building advertising for the way people use AI", confirming that ChatGPT is testing a new visual ad format and expanding how advertisers measure results. TechCrunch and Digiday both covered it the same day, and the short version is this: ChatGPT Ads just went from a single text format to something that starts to look like a real ad platform, with the measurement partners to match.

If you run paid search for a living, you've probably had a client ask "should we be doing ChatGPT ads." Up to now the honest answer was "there's barely anything to buy." That's starting to change. Here's exactly what OpenAI announced, what's confirmed versus reported, and whether it's worth your time yet.

What ChatGPT Ads already was, before this update

ChatGPT Ads isn't brand new. OpenAI first laid out the program in a September 16, 2026 post called "Reimagining advertising with AI," then expanded availability to Southeast Asia and Taiwan on September 23. The original format, per TechCrunch's reporting, was a single placement: advertiser name, logo, headline, description, landing page link, and a static image, shown below ChatGPT's own answer. One ad slot per conversation, not the stack of placements you get on a Google results page.

OpenAI has been expanding ChatGPT's commerce footprint on more than one front this year, not just ads; ChatGPT merchant setup covers the separate product-listing side of that build-out.

That's the context for why the October 5 update matters. OpenAI wasn't launching ChatGPT Ads. It was trying to fix the two complaints advertisers had already filed: there's only one format, and it's hard to prove it worked.

What's new as of October 5

Two things shipped together.

A visual ad format, in testing. Starting later in October 2026, in the US only, a test group of advertisers can show image ads that appear after ChatGPT generates an image for a user, not blended into the generated image itself. OpenAI's own language, as quoted by TechCrunch and Digiday, is that the ads will be "clearly labeled" and "won't influence the answers ChatGPT provides." The use case is straightforward: someone asks ChatGPT to generate a picture of, say, a mid-century living room, and a furniture brand's product shot shows up next to the result. OpenAI hasn't said when this goes beyond the initial test group or beyond the US.

A much bigger measurement bench. This is the part that actually matters more for anyone running budget through the platform. OpenAI added:

Measurement type

Partners named

Click/app attribution

AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, Tenjin

Full-funnel measurement

Fospha, Measured, INCRMNTAL

Geo-based incrementality testing

Haus, Measured, WorkMagic

Brand suitability pilots

DoubleVerify, Integral Ad Science

That's a lot of attribution and measurement vendors, which tells you OpenAI is explicitly chasing the gap between "an impression happened" and "that impression drove a sale." For the brand suitability pilots specifically, OpenAI has been clear about one limit: DoubleVerify and IAS won't get access to actual user conversations, only aggregated placement data. That's a privacy line OpenAI is drawing on purpose, because full conversation access would be the thing that kills advertiser trust in the product fastest.

Where ads appear and what they look like

As of this update, there are effectively two ChatGPT ad surfaces:

  • Text/link ads below a chat answer: advertiser name, logo, headline, short description, static image, link to a landing page. This is the original, already-live format from the September launch, available in the US, Southeast Asia, and Taiwan.
  • Visual/image ads next to AI-generated images: in testing only, US only, limited advertiser group, no confirmed broader rollout date.

Both carry the same stated rule: ads are labeled as ads and don't change what ChatGPT says in its actual response. That separation, ad content never touching the model's answer, is the thing OpenAI keeps repeating across both the September and October posts, probably because it's the single most obvious trust risk in putting ads inside a conversational product.

Targeting and privacy: what OpenAI has actually said

OpenAI has not published a detailed targeting rulebook the way Google Ads has a help center full of audience-signal documentation. What's confirmed, from OpenAI's own posts as reported by TechCrunch and Digiday:

  • Ads are shown based on conversation context, not a documented third-party audience/data-broker targeting stack.
  • OpenAI states advertising doesn't influence model answers, a separation it draws explicitly, for the obvious reason that an LLM quietly favoring sponsored products inside its reasoning would be a bigger story than any ad format.
  • Evaluation partners (DoubleVerify, IAS) are given placement and brand-safety data, not user conversation content.

What's unconfirmed, because OpenAI hasn't published it yet: exact targeting signals (topic, intent, account-level data, geography beyond country), frequency capping, whether free-tier and Plus/Pro users see different ad loads, and whether EU rollout is even planned given GDPR's stricter rules on profiling. If a client asks you for OpenAI's data processing addendum or a targeting spec sheet before they'll sign off on a test, that documentation doesn't appear to exist publicly yet.

Who can buy, and what it costs

This is the least mature part of the whole program. Neither OpenAI's posts nor TechCrunch's or Digiday's reporting specify:

  • Whether buying is self-serve through a dashboard, or requires a direct sales relationship with OpenAI.
  • Any CPM, CPC, or minimum spend figures.
  • A formal application or waitlist process for the visual ad test.

What is confirmed: the text ad format has been live since September in the US plus Southeast Asia and Taiwan, and the image ad format is a closed test with "an initial group of advertisers" OpenAI hasn't named. If your agency hasn't been contacted directly by OpenAI, you're not currently eligible for the visual test, full stop. For the existing text format, reach out through OpenAI's advertising page rather than assuming it's a self-serve Google Ads-style signup; nothing published says it is.

Why OpenAI is moving this fast

Worth noticing: this is OpenAI's third advertising announcement in under a month. The original format launched September 16, the Southeast Asia and Taiwan expansion followed September 23, and now the measurement overhaul and visual format test on October 5. That cadence isn't normal for a company easing into a new product line. It reads like a company under real pressure to diversify revenue beyond subscriptions and API usage, and willing to ship ad infrastructure in public, in pieces, rather than wait for a single polished launch.

The inventory math backs that up. TechCrunch's reporting cites Zeno Group's Shamsul Chowdhury pointing out that ChatGPT handles roughly 2.5 billion prompts a day worldwide against Google's roughly 15 billion searches, with one ad slot per ChatGPT conversation compared to Google's several placements per results page. OpenAI can't out-supply Google on raw ad inventory any time soon. What it can do is make each impression worth more by proving it converts, hence the sudden parade of attribution and incrementality partners rather than more ad formats. That's a deliberate choice: fix measurement credibility before scaling the format itself.

How this compares to Google AI Max and ads in AI Mode

It's a fair comparison to make, because Google is doing something structurally similar on its own turf. AI Max for Search is Google's automation layer on top of regular Search campaigns: broad-match and keywordless targeting, automated asset generation, and the ability to show in AI Overviews and AI Mode results alongside normal search ads. Google has also been testing a "Direct Offer" asset that puts targeted promotions directly inside AI Mode conversations.

The practical difference for a media buyer: Google Ads already has the infrastructure, the reporting, the bidding strategies, and the sheer inventory volume. One industry source, Zeno Group's Shamsul Chowdhury, put the scale gap bluntly to TechCrunch: roughly 2.5 billion prompts happen on ChatGPT daily worldwide, versus roughly 15 billion searches on Google. ChatGPT also has one ad slot per conversation right now; Google Search has multiple ad positions plus Shopping, Performance Max, and AI Overview placements. If you're deciding where to put test budget first, Google's AI-surfaced ad inventory is the more proven buy today, by volume and by reporting maturity. ChatGPT Ads is the one to watch, not yet the one to lead with.

Should you test it? A practical checklist

If you're a Google Ads practitioner wondering whether to bring this to a client, here's how to think about it honestly.

Test it now if: you have a client with meaningful US brand awareness budget already sitting idle, a product that photographs well (since the new format is tied to image generation), and tolerance for a channel with no benchmarks, no self-serve guarantee, and no published CPMs. Treat it as earned-media experimentation, not a performance channel you can forecast against.

Hold off if: your client's whole ask is direct-response ROAS, you don't have spare test budget outside your core Google/Meta allocation, or you need auditable targeting documentation for a regulated industry (finance, healthcare, anything with compliance review). The targeting transparency gap alone will stall most compliance teams.

Either way, do this before you spend a dollar: ask OpenAI's ad team directly for current eligibility, format specs, and whether self-serve exists yet, because public reporting hasn't nailed any of that down. Don't build a client proposal on secondhand screenshots of the ad unit.

A few operational things to line up if a client does want to move: pick which attribution partner you'd actually connect (if you already run Triple Whale or Northbeam for Meta and Google, that's the lowest-lift integration), decide what "success" looks like for a channel with zero historical benchmarks (a brand-lift read, not a CPA target, for the first cycle), and set client expectations in writing that this is a test with a defined budget cap and a defined end date. The worst version of this is a client who treats a speculative test the same way they treat a proven Google Search campaign, then is disappointed when it doesn't behave like one.

On the Google side, the closest equivalent is AI-driven matching in Search campaigns; our guide to Google AI Max for Search covers how it works and how to control it.

What's confirmed vs. what's reported

To be precise about the sourcing here: the existence of the visual ad test, the October timing, the US-only scope, the "clearly labeled" and "won't influence answers" language, and the full measurement-partner list all come from OpenAI's own posts, as directly quoted by TechCrunch and Digiday on October 5 and 6. The scale comparison (2.5 billion vs. 15 billion queries) is an industry analyst's estimate relayed by TechCrunch, not an OpenAI-published figure. Pricing, self-serve status, exact targeting signals, and an EU timeline are all unconfirmed as of this writing; treat any number you see elsewhere for those as a guess until OpenAI publishes something itself.

For the bigger picture on how AI platforms are reshaping where your traffic and ad dollars go, see Google keeps expanding AI Overviews: here's what it means for organic traffic and generative engine optimization for beginners. If you'd rather earn visibility inside ChatGPT answers organically instead of paying for a placement, how to get cited by ChatGPT and Google AI Overviews is the place to start.

Is ChatGPT Ads self-serve like Google Ads?

Not confirmed either way. OpenAI hasn't published a self-serve dashboard or signup flow publicly; the visual ad test is limited to an unnamed initial advertiser group, which points toward a managed or invite-based process for now, at least for the new format.

Where exactly do ChatGPT ads show up?

Two places as of October 2026: a labeled text ad below a chat response (live since September, in the US, Southeast Asia, and Taiwan), and a labeled image ad shown after ChatGPT generates an image for the user (in testing, US only).

Does OpenAI use my conversations to target ads?

OpenAI says ads don't influence ChatGPT's answers and that its brand-safety evaluation partners, DoubleVerify and Integral Ad Science, don't get access to actual user conversations. Beyond that general statement, OpenAI hasn't published a detailed targeting-signal breakdown.

How much do ChatGPT ads cost?

Unconfirmed. No CPM, CPC, or minimum-spend figures have been published by OpenAI or reported by TechCrunch or Digiday as of this writing.

Should I move budget from Google Ads to ChatGPT Ads?

Not yet, for most advertisers. The query volume, ad inventory, targeting documentation, and measurement track record all favor Google for now. ChatGPT Ads is reasonable to test with discretionary budget if you have it, not to replace a channel that's already performing.