ChatGPT Ads

Advertisers admitted to the Ads Manager beta: registration, identity verification, and an approved category are the gate. Larger budgets can also buy through OpenAI’s agency and technology partners. Access has been expanding since the May opening, but it is not universal, and some categories remain restricted. Eligibility is the first thing we check for every prospect, before any budget talk, and the check costs you nothing.

Enough to buy a decision, small enough that a no costs nothing you will miss. For most clients that means a slice of the existing paid budget rather than new money, run for a defined test window with a written question it must answer. We size the number with you in the proposal; nobody should guess it from a webpage.

Differently from keywords, and that is the craft. People describe situations to an assistant instead of typing two-word queries, so placement follows conversational intent. The controls OpenAI exposes are newer and narrower than what Google Ads veterans expect, and they change as the beta matures, which is why we map offers to buyer questions rather than building around any single control.

Two layers. The Ads Manager and OpenAI’s Conversions API report what the platform saw; our own tracking reports what your business got: conversion events in GA4, referral sessions from chatgpt.com, and calls or forms tied to the campaign, which is as close as attribution gets on a conversational surface. The channel gets judged on cost per tracked lead beside your other platforms, never on the platform’s own scorecard alone.

OpenAI gates or restricts several sensitive categories, and beta policies move, so any list printed here would age badly within months. What holds: regulated and sensitive offers face extra review or exclusion, the way they do on every major ad platform. We confirm your category against the current policy during the eligibility check.

Impressions can start the day a campaign clears review. Judgment takes longer: a fair test needs setup, a defined spend window, and one honest read of the data, which is a quarter for most accounts. On a channel this new there are no industry benchmarks to shortcut with, so the timeline to a confident answer is the honest deliverable.

You do. You register as the advertiser, ad spend goes to OpenAI directly, and our management fee stays flat and separate, the same structure as our Google and Meta engagements. If we ever part ways, the account, its history, and every lesson the test budget bought stay with you.

Yes. Copy gets written for the conversation it lands in: a relevant next step after an answer, matched to the intent of the prompt, checked against OpenAI’s ad policies before submission, and pointed at a landing page that continues the conversation. Variants run against each other on a schedule, and what a format this new teaches us gets folded back in monthly.

A defined test window first, usually one quarter, because a fair read needs setup, spend, and a full look at the data. After the test, month to month with thirty days notice, and the test ends with a written recommendation: scale, hold, or stop. Locking a long contract to an unproven channel serves the agency, not you.