ChatGPT Ads Manager Setup Guide: How To Launch Your First Campaign

Set up ChatGPT Ads Manager measurement-first. Choose the conversion event and your spending limit before you create the account, then build one campaign, fire a test conversion and only then submit for review. Skip that order and you pay for clicks you can never attribute.

This applies to the self-serve Ads Manager Beta at ads.openai.com when you run conversion-optimised campaigns, where OpenAI requires conversion tracking through the pixel, the Conversions API or both.

OpenAI changes screens and eligibility during the beta, and its pages publish no budget floor or recommended bids. Check each step against the current OpenAI quickstart before you spend.

What You Need Before You Open ChatGPT Ads Manager

ChatGPT Ads Manager is OpenAI's self-serve ads tool, where you set up the campaign, budget and goal, then bulk upload your ad details or create them in the tool. The current version is a Beta, and the steps in the OpenAI Ads Manager quickstart start with creating or signing in to your account at ads.openai.com.

Know who you are paying to reach before you go further. Ads show to logged-in adult users on the Free and Go tiers, and Plus, Pro, Business, Enterprise and Education users see no ads at all. If your buyers sit on the paid tiers, your ad will not reach them in the chat. Whether your own account gets into the beta is something only the sign-in screen can tell you.

This guide assumes you have already decided to test ChatGPT ads. It does not argue the case for or against. If ChatGPT is new to you as a channel, our guide to AI search engines shows where it sits among the others.

The order here is not the order of the screens. It is the order that makes your first campaign testable.

1. Conversion event: Pick it and get tracking in place.
2. Budget: Work it out from your own numbers.
3. Account: Create it, verify it and add your team.
4. Campaign: Build the campaign, ad group and ad.
5. Bidding: Choose the billing option and the Bid Cap.
6. Pre-launch test: Fire one real conversion and confirm it shows.
7. Week one: Reconcile what Ads Manager reports with your own analytics.

Set The Conversion Event Before You Build Anything

Pick one standard conversion event, and have tracking in place for it, before the campaign exists. Conversion-optimised campaigns require conversion tracking through the Conversions API, the JavaScript pixel or both, and each campaign takes one standard event. Custom events are not supported.

You cannot fix a wrong choice later. The event and the billing option cannot be changed in place, so changing either means cloning the campaign. That is why this step comes first.

1. Match the event name to a data source: Hardal's measurement guide says the two standard events are order_created for a purchase and lead_created for a lead. The campaign's event setting has to match the exact event name and the data source that sends it. Pick order_created if your goal is a sale, and lead_created if your goal is a form fill or demo request. An event that reaches OpenAI is not automatically an attributed conversion, so do not read "events are arriving" as "the campaign is working".
2. Choose the attribution window now: Hardal's measurement guide lists click-through attribution windows of 7, 14 or 30 days, with an optional view-through window of 0 or 1 days. Match the click window to your real sales cycle. A lead form that converts the same day needs less than a B2B product with a long evaluation. Leave view-through at 0 unless you have a reason to count people who saw the ad and came back on their own.
3. Check deduplication if you send both: You do not need both the pixel and the Conversions API, but you may want both. When both send an event, they must carry the same event name and a stable event ID so OpenAI can deduplicate. Generate the ID once per action and send the same value from both sides. Turn on automatic advanced matching too, because OpenAI says it improves matching.

If you can only do one thing this week, install the pixel and fire the event from a test page. A conversion you can see arriving is worth more than a polished campaign you cannot measure.

Set A Budget From Your Own Numbers

The OpenAI pages we reviewed publish no budget floor and no recommended bid, so this guide does not quote one. Figures from third-party guides conflict, and none names a source you can check. Run the numbers on your own business instead.

Start from what a conversion is worth to you and what you already pay for the same result elsewhere. If you run paid search, you know your cost per lead or cost per order. That is your baseline. Then set a pass threshold before launch as a multiple of that baseline. A new channel with no history should be allowed to cost more than your proven one, but only up to a ceiling you write down now. Deciding after the money is spent is how a weak test gets rationalised.

Then size the test so it can produce enough conversions to read. A budget that buys three conversions tells you nothing, because one lucky or unlucky click swings the result.

The formula: Test budget equals the target cost per conversion times the conversions needed to read the result.

Here is a worked example. Every number is invented for illustration. Use your own.

• Paid search baseline: $80 per lead
• Pass threshold: 1.5 times the baseline, so $120 per lead
• Conversions needed to read the result: 20 leads

Multiply $120 by 20 and the test budget is $2,400. If that number is more than you can spend, do not shrink the budget and hope. Lower the number of conversions you need, accept a longer test, or pick a cheaper event to optimise for. A smaller budget with the same expectations just produces a result you cannot trust.

Keep the threshold in a note. The pre-launch checklist and the week-one review both point back to it.

Create The Account, Verify It And Add Your Team

Follow the order OpenAI gives in the Ads Manager quickstart. The screens may change during the beta, so keep the live page open while you work.

1. Sign in or create your account: Go to ads.openai.com and create the Ads Manager Beta account, or sign in if you have one.
2. Add your logo: Upload it in Settings.
3. Set billing: Add the payment details the account asks for.
4. Complete verification: Work through whatever verification the account requires.
5. Invite your team: Add the people who will build and monitor campaigns.
6. Create campaigns: Covered in the next section.
7. Submit for review: Covered two sections on.
8. Monitor serving: Watch the status of each campaign once it is submitted.

The quickstart lists these steps but the pages we reviewed do not publish the exact verification steps. Do not leave verification to the day you want to launch. A billing or verification action left undone can stop your ads from serving, which the status section below covers.

Invite your team before you build, not after. It is much easier to give a colleague access up front than to chase an admin later, and it means the person who owns the analytics can check your conversion setup. Whether to add an agency as a user or set up separate client accounts is a question the reviewed OpenAI pages do not answer, so ask OpenAI or check the live help pages before you decide.

Build The Campaign, Ad Group And Ad

The quickstart splits the build into three levels. Know what each one holds before you start clicking.

OpenAI defines the three levels this way, and the build follows the same order.

1. Create the campaign: Choose the conversion-optimised objective you prepared for earlier. This is where your standard event, your budget and your dates go. Set the budget from your worked example, and set a daily cap from the first day.
2. Set targeting: Keep the first campaign narrow enough that you can tell what produced the result. One audience, one event, one landing page.
3. Add an ad group if you want one: Ad groups hold optional context hints. They are optional, so skip them if you have nothing specific to say about where the ad should appear.
4. Create the ad: Add the title, description, image and landing page. You can bulk upload ad details or create them in the tool, whichever fits the volume you have.
5. Tag the landing page URL with UTM parameters: This is the step people skip, and it is the one that lets you match Ads Manager to your own analytics in week one. Use a consistent source, medium and campaign name.

Ad copy and creative strategy are out of scope here. Write the best landing page and the clearest offer you can, and come back to the creative once you know the measurement works.

Choose The Bidding And Billing Option

For a first campaign, choose the billing option that matches how you want to learn. Ads Manager offers Impressions or Clicks as the billing option, and conversion-optimised campaigns use a Bid Cap, which is the maximum bid per conversion.

Be careful, because both the billing option and the conversion event are locked once the campaign exists. Changing either means cloning the campaign. Decide with the cost of that clone in mind. If you are unsure, make the choice you can live with for the full test, because a half-way switch muddies your data anyway.

OpenAI's page gives no recommended bid, and third-party starting prices conflict with each other, so none are given here. Derive the number yourself.

• Derive the maximum bid from your own economics: Your pass threshold from the budget section is the ceiling. If you decided a lead is worth paying $120 for, the Bid Cap should never sit above that.
• Start below it: Beginning under the ceiling leaves room to raise the bid once you see the campaign serving. Overbidding early burns budget before you know anything.
• Set daily caps from day one: A campaign-level daily cap keeps one bad afternoon from eating the whole test budget.

If the campaign is not serving at a bid below your ceiling, raise it in small steps and note each change. You want to know which change caused which result.

Run The Pre-Launch Checklist

Work through this list before you submit. Every item should be a thing you can see on screen, not a thing you believe is true.

• Confirm the conversion event is the one you chose, and that the event name in the campaign matches the name your pixel or API sends.
• Check that tracking is installed on the page where the conversion happens, through the pixel, the Conversions API or both.
• Verify that the event ID is identical on both sides if you send both.
• Open Settings and confirm the account is verified and billing is set.
• Review the campaign budget and Bid Cap against the pass threshold you wrote down in the budget section.
• Confirm the daily cap is set at campaign level.
• Check that the landing page URL carries your UTM parameters.
• Run one real action from an ad landing URL, then look at the recent events monitor and the Conversions columns in Ads Manager.

That last item is the test. Run one real action from an ad landing URL and check both places before you scale spend. A pass means the event appears in the recent events monitor under the right name, and the Conversions column moves for the campaign you expect. If the event shows up but the column does not, remember that an event reaching OpenAI is not the same as an attributed conversion. Fix the match before you submit, not after you have paid for traffic.

Submit For Review And Read The Not Serving Status

Once the checklist passes, submit the campaign for review, as the quickstart describes. Then monitor serving.

A campaign can show a Not serving status. Hover over it and Ads Manager gives the reason. OpenAI names four causes.

• Policy rejection: The ad was rejected under OpenAI policy.
• Account review: The account needs a review before it can serve.
• Business verification: Verification of the business is still required.
• Billing action: Something on the billing side needs your attention.

Read the hover reason first and fix that one thing. Three of the four are account problems, not campaign problems, so rebuilding the ad will not help. This is also why you complete verification and billing early rather than on launch day.

What To Check In Week One

Ads Manager gives you aggregate performance information such as views, clicks and conversions. That is all the reporting the reviewed pages describe. It is enough for a first read, but it is a platform view. You need your own analytics to tell you what those clicks did after they arrived.

Reconcile the two with three checks. The UTM parameters you added to the landing page are what make this possible.

• Clicks against sessions: Compare the clicks Ads Manager reports with the sessions your analytics attributes to your UTM campaign name. A large gap points to a tracking or redirect problem on the landing page.
• Conversions column against your own conversions: Compare the Conversions column with the leads or orders your own system recorded for the same UTM campaign. Expect them to differ a little, and investigate if they differ a lot. The attribution window you chose decides what the platform counts.
• Event arrivals against attributed conversions: Compare the events in the recent events monitor with the Conversions column. If events arrive but conversions do not, the event name or data source in the campaign is probably mismatched.

Then judge the week against the pass threshold you set in the budget section, and nothing else. Do not invent a new bar because the first numbers look bad or good. If you want more than clicks and conversions, our breakdown of AI search visibility metrics covers the measures that apply to the organic side of ChatGPT.

Remember what needs a clone to change. If the data tells you the event or the billing option was wrong, you are building a second campaign, not editing the first. Treat the first as a paid lesson and carry the tracking setup straight across.

Paid ChatGPT Ads Versus Being Cited In ChatGPT Answers

You now have two routes into ChatGPT, and they work differently. Paid ads are sponsored listings within the chat, always labelled as sponsored and visually separated from the organic answer. Being cited means the answer itself names or recommends your brand, which you earn rather than buy.

Organic citation is the work of Generative Engine Optimization and Answer Engine Optimization. Start with a clear picture of AI search visibility, and see how ChatGPT answers mention you with a set of ChatGPT answer monitoring tools. The practical side is getting cited in AI answers. If the terms are tangled, read how AISO differs from SEO, AEO and GEO, and for the wider shift see what changed in SEO after AI search arrived.

Keep the two budgets and the two reports separate. The pages reviewed for this guide say nothing about paid placement changing organic results, so do not assume it does.