HOW TO
ADVERTISE ON
CHATGPT

1. Confirm availability, category eligibility and current policies

OpenAI made self-service Ads Manager available across the 31 announced European markets on 31 August 2026. Before building anything, confirm that the advertiser-account country and business category remain eligible.

The second check is policy. Version 1.6, updated on 10 September 2026, continues to restrict categories and placement contexts and explicitly clarifies OpenAI’s right to reject ads that conflict with its advertising principles, business interests or competitive position. Eligibility is a live product constraint rather than a one-off launch detail.

Official sources: ChatGPT Ads expands across Europe · OpenAI Advertising Policies.

2. The advertiser creates the account; the agency is invited afterwards

OpenAI’s current setup flow requires an account owner from the business to create the advertiser account. It explicitly states that agencies creating advertiser accounts on a client’s behalf are not supported; an agency can be invited after the client account exists.

Setup covers onboarding, verification, account information, billing and payment before campaigns can deliver. Country, currency, time zone, legal business name, website and billing details deserve careful review because some account-level fields may not be self-service editable after creation.

Official source: Ads Manager Beta Account Setup.

3. Choose the objective and billing model before creating the campaign

The standard campaign workflow documents CPM for reach, CPC for traffic and oCPC for downstream post-click conversion optimisation. OpenAI also documents oCPM in beta for conversion-optimised campaigns: delivery remains focused on a selected conversion event while billing is impression-based.

oCPC is still billed on valid clicks and oCPM on impressions. For conversion-optimised campaigns, the objective, billing model and selected conversion event should not be treated as provisional settings: current documentation says the objective and billing model cannot be changed after campaign creation, and at least one supported standard conversion event is required.

Official sources: Create Campaigns for ChatGPT Ads · Conversion-optimized Campaigns.

4. Set budget, markets and controls without mistaking budget for pacing

The campaign holds its objective, budget, dates, countries and available controls. Ads Manager supports campaign-total and daily budgets. A campaign total is an aggregate spending limit; a daily budget is an average delivery target over seven days rather than a fixed ceiling for every calendar day.

OpenAI documents that spend can reach up to 2× the selected daily budget on a given day and up to 7× that budget over the applicable seven-day period, subject to proration when the budget changes. The technical minimum daily budget also varies by billing currency. Separately, we set a commercial pilot limit: how much downside are we prepared to accept before requiring an answer on delivery, cost, conversion and quality?

Depending on account and campaign availability, optional controls can also include platform targeting or Custom Audiences for inclusion/exclusion. Those are platform controls, not a substitute for a clear acquisition hypothesis.

Official sources: campaign structure and budgets · Daily Budgets.

5. Build ad groups around one coherent need or intent area

OpenAI recommends keeping each ad group focused on a common product category, theme or customer need. Where two use cases represent materially different products, messages or landing pages, separating them helps each group describe an interpretable intent.

The objective is not to maximise account complexity. Each ad group should answer a clear question: which need are we covering, which offer are we presenting and what would we expect to learn if that group receives traffic?

Official source: Create Ad Groups for ChatGPT Ads.

6. Write context hints as additional context, not keyword lists

Context hints explain what the product offers, who it helps or when it may be useful. OpenAI recommends natural phrases that add information beyond the headline, ad copy or landing page and connect genuine capabilities with real needs or situations.

They are not exact-match controls, audience-targeting rules or commands to appear for specific words, topics or audiences. They also do not guarantee delivery. Their role is to add context within a coherent ad group.

Official source: Context hints and ad-group structure.

7. Build creative and a landing page that continue the intent

The ad needs to make the advertiser, proposition and destination clear. We test messages that are meaningfully different enough to teach us something rather than cosmetic variants of the same claim. OpenAI lists signals such as landing page, headline and copy alongside conversation context and other available inputs.

After the click, the landing page needs to resolve the intent that made the ad relevant. That second layer is CODE’s conversion methodology: proposition, proof, friction, form and CTA should help the user decide without pretending OpenAI publishes a Google-style landing-page Quality Score.

Official source: Ads in ChatGPT: The Basics.

8. Install measurement before launching a conversion-optimised campaign

OpenAI supports conversion events through Pixel, Conversions API or both. The click reference oppref is appended to the landing-page URL and should be preserved through redirects and navigation; when the same conversion is sent browser-side and server-side, the same event ID enables deduplication.

For oCPC/oCPM, tracking needs to be ready before campaign creation and at least one supported standard event must be available. We complement platform measurement with UTMs and a first-party business source — ecommerce, analytics or CRM — because a difference between Ads Manager and external systems does not automatically mean one of them is broken.

Official sources: Conversion Measurement · Conversion-optimized Campaigns.

9. For retail, decide whether a product feed creates a real advantage

Retail advertisers can import catalogue data via CSV/TXT, a hosted HTTPS URL or SFTP and create dedicated product-feed campaigns. Feed items expire after two weeks, so a live catalogue usually needs automated refresh.

During the beta, uploading a product makes it eligible for ads rather than organic ChatGPT appearances. Ads Manager can filter products by ad group and use ads_metadata where custom grouping is required. For eligible new ad groups, Maximize results may be selected by default and automatically adjust bids towards clicks or conversions according to the campaign goal.

OpenAI says ads created from product feeds have been among the strongest-performing ads in its programme to date. We present that accurately as a vendor claim, not an independent benchmark or a ROAS promise.

Official source: Create Campaigns from Product Feeds.

10. Launch with QA and explicit scale, correction and stop criteria

Before media goes live, we confirm that account setup, billing, ad, destination, event and measurement are ready and define the question the pilot must answer. Then we set the commercial outcome that matters, the cost that would be defensible and the amount of signal needed for a sensible decision.

The first campaign is not there to prove that “ChatGPT Ads works”. It is there to establish whether the channel can work for this business, under which conditions, and whether the next decision should be to scale, correct a specific dependency or stop buying more evidence.