ChatGPT Ads just grew up — barely. What a performance marketer should actually do this week

What it is

OpenAI's October 5 update to ChatGPT Ads bundles three things: a new visual ad unit, a wider measurement partner list, and a brand-suitability pilot. The company is introducing cost-per-click bidding alongside existing CPM options, and the new visual format appears against a generated image in mobile ChatGPT. The buying happens through Ads Manager Beta, OpenAI's self-serve console.

The next-work-session test

You're a paid-media lead planning next month's test budget. Today, "ChatGPT" is a line item you can't defend to finance because you can't attribute it. This update changes exactly that one thing: you can now wire ChatGPT campaigns into the same measurement stack you already use. The announcement widened ChatGPT Ads measurement with conversion-data integrations, ten attribution partners, geo-lift and brand-lift tests, and brand-safety pilots with DoubleVerify and Integral Ad Science. Concretely: if your team runs Triple Whale, Northbeam, or AppsFlyer, ChatGPT spend can finally sit next to Meta and Google in the same dashboard — OpenAI listed outside companies including Hightouch, Tealium, LiveRamp, AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, Tenjin, Fospha, Measured, INCRMNTAL, Haus.

The visual format is more speculative. OpenAI said it will test the new visual ad format in ChatGPT during image generation, starting later this month in the US with an initial group of advertisers, so most teams won't touch it this quarter.

Pricing

Pricing is unverified in any useful sense. OpenAI has not published rate cards or minimums for ChatGPT Ads. What's public: the company is introducing cost-per-click bidding alongside existing CPM options, and buying is auction-based inside Ads Manager Beta. OpenAI has not provided pricing details in the supplied announcement, so advertisers will need to wait for campaign-level information as the beta develops. You'll see your actual costs only after you're approved into the beta and fund an account. Assume parity with other emerging search/chat auctions until proven otherwise.

What we'd actually use it for

A small, defensible incrementality test. Pick one SKU or offer where ChatGPT users plausibly show research intent — B2B software evaluations, travel comparisons, consumer electronics — and run a geo-lift study using one of the named partners. That gives you a number for the "is ChatGPT doing anything?" question your CMO will ask within six months. It is not yet a channel for scaled always-on performance budget.

For most teams, the honest use is reconnaissance: get an account, run a $5–10K test, learn the interface, and have a point of view before competitors do.

Limits

• US-only, approval-gated. Approved U.S. advertisers can sign up at ads.openai.com as the beta expands. If your billing entity is elsewhere, you're still waiting.
• Narrow category eligibility. Eligible categories include household and consumer goods, local services, travel and experiences, and digital products and education. Dating, alcohol, tobacco, healthcare, financial and legal services, gambling, and political content are excluded. That rules out big chunks of B2B fintech, pharma, and regulated verticals.
• Audience is limited. Ads only appear to logged-in adults on the Free and lower tiers — paying subscribers don't see them, which skews the audience away from exactly the high-income professionals many B2B marketers want.
• Visual format is not generally available. It's a test with a hand-picked advertiser group, mobile-first, triggered during image generation.
• Transparency is still thin. OpenAI could close the measurement gap and establish ChatGPT as a more mature ad channel — the operative word is could. There's no public auction insights, no competitive benchmarking, and inventory volume isn't disclosed.
• Attribution to conversations is deliberately walled off. OpenAI announces the format test alongside expanded measurement partners for ChatGPT Ads and brand-suitability pilots that, it says, work without accessing private user conversations. Good for user trust, limiting for marketers who want prompt-level signal.

Try it if

• You run paid media for a US consumer brand in an eligible category and have budget to spare for a learning test.
• You already use one of the named attribution partners (Northbeam, Triple Whale, AppsFlyer, Measured, Haus) and can plug ChatGPT in without new vendor work.
• Your CMO is asking "what's our AI-channel strategy" and you need a concrete answer by Q1.
• You want to be first-mover on a visual format before CPMs rationalize.

Skip it if

• Your billing entity isn't in a supported country, or your category is on the exclusion list.
• You don't have a measurement partner in place — running ChatGPT Ads on last-click alone will produce numbers you can't trust.
• Your ICP is enterprise B2B — ChatGPT Plus and Pro users don't see ads, and that's where many of your buyers are.
• You're under pressure to hit this quarter's pipeline number. This is a learning budget, not a performance channel yet.