Conversation
Name the problem, comparison, or decision where your offer can help.
Source basis: OpenAI's current Ads in ChatGPT documentation · Updated July 28, 2026
How to run ChatGPT Ads: from account setup to the first campaign review
Use this field guide to confirm access, choose the right objective, write context hints, build the ad, connect measurement, launch cleanly, and decide what happens after the first week.
7countries currently listed for Ads Manager availability
3objectives: Reach, Clicks, and Conversions
$3 to $5OpenAI's current recommended starting CPC bid in USD
24hminimum learning window OpenAI recommends before judging early reporting
00
The direct answer
Build the decision path before you open the campaign builder.
The campaign should connect one buyer situation to one useful offer, one honest landing page, one measurable action, and one clear decision rule.
ChatGPT Ads are not a keyword list placed beside a search result. The ad can appear when OpenAI determines it may be useful in the context of a conversation. That changes the planning job. You still need a clear offer and a page that converts, but you also need to describe the conversation where the offer belongs.
Use this order: confirm eligibility, define the business result, choose the objective, separate the campaign into focused ad groups, write broad context hints, build matched ads and landing pages, connect measurement, set budget boundaries, launch, and reconcile the first results.
Do not start with “What ad should we write?” Start with “Which conversation should end with this next step?”
Next step
Choose the page and action that fit that moment without forcing a leap.
Proof
Define the event and business record that will show whether the test worked.
01
Account and eligibility
Confirm that the business can advertise before you plan the test.
Current access is limited. Country availability does not override verification, policy, billing, or landing page requirements.
OpenAI currently lists Ads Manager availability in Australia, Canada, Japan, Korea, New Zealand, the United Kingdom, and the United States. One account owner creates the account, enters business details, completes Persona verification, waits for platform approval, adds the account logo and information, configures billing, and then invites teammates.
Choose the country, currency, and time zone carefully. OpenAI says those settings cannot be changed after account creation. Save the account owner, verification status, billing owner, and selected time zone in the launch record.
| Gate | What to confirm | Evidence to keep |
|---|---|---|
| Country | The advertiser is in a currently supported Ads Manager country. | Business address and selected account country |
| Business | The legal business details match the information used for Persona verification. | Owner and verification status |
| Category | The offer and claims comply with the current OpenAI ad policies. | Policy review date and any approval notes |
| Destination | The landing page is public, accurate, mobile-ready, and consistent with the ad. | Final URL and a tested conversion path |
| Billing | The payment method, spend owner, currency, and time zone are correct. | Billing owner and approved spend boundary |
02
Objective
Pay for the action you are actually ready to evaluate.
OpenAI currently offers Reach with CPM, Clicks with CPC, and Conversions with optimized CPC.
Use when exposure is the result
Fit for a broad awareness goal when the business can judge qualified reach without pretending that impressions are sales.
- Primary platform result: impressions
- Business check: useful audience exposure
- Risk: buying visibility with no next-step owner
Use when the page is the next test
Fit when the landing page, offer, and analytics are ready and qualified site visits are the immediate learning goal.
- Primary platform result: clicks
- Business check: visit quality and next actions
- Risk: optimizing curiosity instead of buyers
Use when the event is already real
Fit when a supported standard conversion event is working and the business can reconcile it with a final record.
- Primary platform result: conversions
- Business check: lead, booking, or sale quality
- Risk: teaching the campaign from a weak event
A Conversions campaign requires a new campaign, working conversion tracking, and a supported standard event. OpenAI says custom events cannot be used to optimize an oCPC campaign. The campaign uses a bid cap and bills for valid clicks, not for conversions.
If the conversion event is not stable yet, use Clicks to test the conversation-to-page match and repair measurement first. The deeper ChatGPT Ads measurement dossier shows how Ads Manager, the Pixel or API, and the final business record should be reconciled.
03
Campaign structure
Keep one business question inside each ad group.
A campaign sets the objective, conversion event, country, budget, and dates. Each ad group should hold one product, audience, or use case with its own context hints and ads.
Use the narrowest structure that still produces a useful answer.
Campaign: the business result, objective, country, budget, dates, and conversion event if required.
Ad group: one offer, use case, or buyer situation with related context hints.
Ad: one specific promise, one aligned image, and one relevant destination.
Do not mix three services, five audiences, and unrelated pages in the same ad group. If performance moves, you will not know what caused it.
The useful constraint
One ad group should answer one question: “Can this offer earn the next step in this kind of conversation?”
If two products need different proof, different pages, different context hints, or different success rules, they need different ad groups.
04
Context hints
Describe the conversation, not a pile of exact-match keywords.
OpenAI calls context hints broad thematic signals. They help communicate where the ad could be useful, but they do not guarantee placement.
Weak: product fragments
“CRM, software, lead tool, sales app, follow-up, automation.” This names a category without explaining the buyer's situation.
Stronger: buyer situation
“A small service business comparing ways to respond faster to website leads and stop quote requests from going cold.”
Weak: every possible customer
“Businesses, owners, managers, teams, startups, ecommerce, local services.” Broad reach erases the decision context.
Stronger: problem plus next step
“An HVAC company owner deciding whether missed-call text-back can recover after-hours estimate requests.”
Write a context hint in three parts: who is in the conversation, what they are trying to decide, and why your offer is a useful next step. Use a separate ad group when the product, customer, use case, page, or success definition changes.
Do not claim that a hint targets a specific private conversation or exact phrase. OpenAI says advertisers do not receive individual user conversations or search queries.
05
Ad and landing page
Make the ad useful enough to click and the page honest enough to trust.
OpenAI's current guidance calls for relevant, accurate ads with simple aligned images and a destination that matches the promise.
Start with the next step, not the company biography. A strong ad title identifies the useful result or offer. The copy explains why it belongs in the current decision. The image should support recognition without imitating ChatGPT, presenting a fake interface, or making a claim the page cannot prove.
OpenAI currently recommends a title of 16 to 24 characters, with a 50-character maximum, and copy of 32 to 48 characters, with a 100-character maximum. Images can be PNG or JPG, square, up to 1200 by 1200 pixels, and publicly accessible.
| Element | Write or build this | Reject this |
|---|---|---|
| Title | A specific next step or result the page can support | Vague superlatives, fake urgency, or an unsupported outcome |
| Copy | Why this offer helps in the current decision | A generic company description or a list of services |
| Image | A clean image aligned with the product or use case | Fake ChatGPT interface, hard-to-read text, or unrelated stock art |
| Page | The same promise, proof, price context, and next action | A home page that makes the buyer restart the search |
| URL | Static UTM parameters that survive redirects | Untested dynamic macros or a destination that changes after review |
If the page is the weak point, use the local website structure guide for ChatGPT Ads and AI search before buying traffic. It maps the service, location, proof, citation, and conversion pages that local businesses need.
06
Budget and bids
Set a learning boundary, not a hope budget.
The platform minimum tells you what can launch. Your business boundary tells you what the test is allowed to cost.
OpenAI currently recommends a starting CPC bid of $3 to $5 USD. That is platform guidance, not a promise about traffic, lead quality, or sales. The auction is described as relevance-weighted and second-price, so eligibility and relevance matter alongside the bid.
Current country minimums are $25 USD in the United States, $25 CAD in Canada, $25 AUD in Australia, $25 NZD in New Zealand, £15 in the United Kingdom, ¥2,500 in Japan, and ₩25,000 in Korea.
Since July 27, 2026, OpenAI describes daily budget as an average across seven days. Spend may reach up to twice the average daily budget on one day, but no more than seven times that budget across seven days. Use a written total test ceiling and check actual pacing.
07
Conversion measurement
Connect the click to a business record you can inspect.
Use the OpenAI Pixel, the Conversions API, or both. Then reconcile attributed conversions with the record that matters to the business.
Track three records, not one dashboard.
Ads Manager: impressions, clicks, spend, CTR, average CPC, average CPM, and attributed conversions.
Pixel or API: the exact event, timestamp, value, and event ID sent to OpenAI.
Business record: the lead, call, booking, order, revenue, qualification, or sale in your own system.
OpenAI appends an oppref parameter to the ad click. Preserve it in a first-party cookie and send it with the Conversions API event when using server-side measurement. If the same conversion is sent from the Pixel and API, use the same event ID so the records can be deduplicated.
Only collect and send data you are allowed to use. OpenAI also supports advanced matching and modeled measurement, which means differences between Ads Manager, analytics, and the CRM are expected. The job is not to force every total to match. The job is to explain the differences before changing spend.
A conversion event is a signal. The sale, qualified lead, completed booking, or retained customer is the business result.
08
Interactive launch check
Twelve checks before the campaign can spend.
Progress is saved only in this browser. No campaign or account information is sent to Stan Consulting.
First-campaign readiness
0 of 12 readyBuild the first-campaign brief
Fill this out before Ads Manager. The brief stays in this browser until you clear it.
Your campaign brief will appear here.
09
Launch and first review
Give reporting time to settle, then review the whole buyer path.
OpenAI says reporting can be delayed by up to seven hours and recommends allowing at least 24 hours before judging early results.
| When | What to inspect | What not to do |
|---|---|---|
| Before launch | Approval status, final URLs, mobile page, event firing, spend boundary, and campaign dates | Assume an approved ad means the conversion path works |
| First 24 hours | Serving status, spend pacing, clicks, destination errors, event receipt, and obvious policy problems | Rewrite the campaign from a few delayed rows |
| Days 2 to 7 | Conversation-to-page fit, CTR, CPC, event match, lead or order quality, and budget pacing | Optimize to the cheapest click without checking the buyer |
| First decision | Ads Manager, Pixel or API, analytics, CRM or store, and the written stop rule | Call an attributed conversion a profitable customer without proof |
Change one major variable at a time. If the ad receives impressions but few clicks, inspect relevance and creative. If clicks arrive but the page produces no qualified action, inspect message match, page proof, mobile friction, and the offer. If the business record contains good outcomes but Ads Manager does not, inspect the event, oppref, deduplication, consent, and reporting delay.
For customer-list inclusion, exclusions, or bid multipliers, use the separate ChatGPT Ads custom audiences and bid adjustments dossier. It covers current size thresholds, identifier rules, and the difference between owning a customer list and having a usable audience strategy.
10
Primary sources
Check the platform source before you spend.
Availability, minimums, policies, and product behavior can change. These are the OpenAI pages used for this guide.
Current supported countries.
Business details, verification, approval, billing, and team access.
Objectives, campaign fields, and country budget minimums.
Ad-group separation and context hints.
Guided and bulk build, ad limits, creative dimensions, and reporting delay.
Auction, bid guidance, reporting, and static UTMs.
Pixel, Conversions API, oppref, deduplication, and advanced matching.
Supported events, bid caps, and click billing.
Average daily budget and seven-day pacing.
Current restricted content, claims, creative, and landing-page requirements.
11
FAQ
Questions to settle before the first campaign.
Can any business run ChatGPT Ads?
No. Access depends on the current country rollout, account approval, business verification, ad policy, and landing page eligibility. Confirm all five before planning spend.
Which objective should a first campaign use?
Use Reach when exposure is the result, Clicks when qualified landing page visits are the result, and Conversions only when a supported conversion event is already working and can be reconciled with the business record.
Are context hints the same as search keywords?
No. They are broad thematic signals. Describe the buyer situation and decision where the ad can help. Do not treat them as an exact-match list.
How much should a first test cost?
There is no universal number. Start with the current platform minimum for the country, set a bid boundary you can defend, and write the total learning cost the business will accept. OpenAI's $3 to $5 starting CPC recommendation is guidance, not a result guarantee.
Should we use the Pixel, the Conversions API, or both?
Use the method your team can implement and test correctly. Both can improve coverage, but they require the same event ID for deduplication. Always reconcile platform conversions with the business record.
Can a customer list improve a campaign?
It can support inclusion, exclusion, or bid adjustments if it meets OpenAI's current audience rules. A list alone is not a strategy. The campaign still needs a clear objective, offer, context, page, and review rule.
Have us run it
Turn the guide into a campaign your business can measure.
We can build the campaign structure, landing path, measurement plan, launch record, and first review around the result your business needs.