Yes, if your business is registered in a country OpenAI has opened and your category is allowed. The self-serve Ads Manager is in beta, bills a credit card, and OpenAI sets the minimum daily budget at $25. A sensible first test is $700 to $1,400 over four weeks on one Clicks campaign, with enquiry tracking working before the first click.
Three checks come before any budget. Ads show only to adults on the Free and Go plans, so paying subscribers never see them. Financial services is a restricted category that OpenAI reviews advertiser by advertiser. And since 30 October 2026, creative enhancements (automatic rewriting of your ad copy and destination) have been on by default for self-serve campaigns, which matters if your wording has to stay exact. The figures below come from OpenAI's ChatGPT Ads help centre and ad policies page as they read in early November 2026. The beta changes often, so Ads Manager itself is the final word.
Three gates to clear before you spend anything
A first test can fail for reasons no budget will fix, simply because one of these gates was shut and nobody looked. Check them in this order, because each one is cheaper to settle than the one after it.
| Gate | What to check | Where to look | If it fails |
|---|---|---|---|
| Country and business | Whether Ads Manager is open where your business is based, and whether a new account can only advertise at home until verification and a spend threshold are met | The Ads Manager Availability page in OpenAI's help centre and the country screen at sign-up | Sign up anyway and OpenAI emails you when your country opens. Register only the real business, in its real country |
| Category | Whether what you sell is allowed, restricted or banned | OpenAI's ad policies page, especially the finance, health and legal sections, and the eligibility table in its onboarding and policy troubleshooting guide | If restricted, get approval before building anything. If banned, stop here |
| Audience | Whether your customers use the Free or Go plans | Ask five recent customers which ChatGPT plan they use, if any | If most pay for ChatGPT or never use it, the test measures the wrong crowd |
On the first gate, OpenAI's availability page listed 89 countries as open for self-serve when I checked it in early November 2026, and it says the legal entity that advertises and is billed must be based in one of them. OpenAI's campaign help page adds that some new self-serve accounts can at first advertise only in their home country, until identity verification is approved and the account reaches a required spend from campaigns at home. Ticking other countries in the campaign does not override that. One early walkthrough described entering another country's details as a way into unsupported markets. Don't: OpenAI's advertiser policy says advertisers may not misrepresent their business location or their eligibility to advertise in a market, and verification exists to check the business behind the account.
On the third gate, the plan rule has a practical consequence for professional firms. If your clients are owners and managers who use a paid ChatGPT plan for work, they are the people who never see an ad. A nursery or a music school selling to parents may well meet more of its audience on the free tier than an accountancy practice selling to business owners, though only your own customer check can tell you.

The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowAccount set-up: what to have on the desk first
Set-up takes minutes to fill in and sometimes days to clear, so start it before you plan the campaign. Have these ready:
- The registered legal name of the business, which OpenAI uses for billing and business verification.
- A business address and an invoice email address.
- A credit card for billing. Adding it can place a temporary authorisation hold, separate from ad charges.
- Your logo, which shows beside every ad as the advertiser's favicon.
- A landing page you control, which can record an enquiry.
- The person with authority to complete the identity check.
OpenAI treats account verification, account approval, ad review and ad delivery as four separate checks, and finishing one does not mean the others are done. If verification is needed, a banner in Ads Manager asks for it, and ads may not serve until it clears. Billing must be set up before anything delivers, so a delay at any of these steps is a delay to the whole test.
One trap sits in the sign-up path that builds a first ad from your website address. OpenAI's set-up guide warns that ads created during sign-up can go live as soon as they pass review, once billing and the account are ready, and spend starts to accrue. Your first campaign may not even offer start and end dates until after it is published. If you are not ready to advertise, open Ads Manager and pause that campaign straight away, or skip the sign-up ad and build the campaign later.
OpenAI's set-up guide confirms that country or region, currency and time zone cannot be changed after the account is created. Choose the ones you will actually be billed and report in, and write them down so the person who later reads the reports knows which clock they are on.
How the money works: objectives, bids and daily budgets
Ads Manager is built on three levels: a campaign (objective, budget, dates, location), an ad group (bid and context hints) and the ads themselves. OpenAI's help pages list three objectives. Views (called Reach in places) buys per thousand impressions. Clicks buys per valid click, against a maximum bid you set for each ad group. Conversions optimises towards an event tracked with OpenAI's pixel or Conversions API, and is billed per click or per thousand impressions, never per conversion. For a first test, Clicks is the one to pick, because you pay only when someone arrives on your page.
Three money facts shape a small budget.
- Daily floor. OpenAI's help centre gives $25 as the standard minimum daily budget for dollar-billed accounts, with its own figure for each other billing currency. With fixed bidding, the daily budget must also be at least your highest ad-group bid, which a $3 to $5 bid never trips. Read the minimum Ads Manager shows you today.
- Weekly averaging. A daily budget is an average over a Sunday-to-Saturday week in your account's time zone. One day can bill up to twice the daily figure, but the week never bills more than seven times it, so judge by the week, not the day. A campaign that uses its weekly allowance early can stop delivering until the next week starts, even while it shows as active.
- Bid guidance. OpenAI recommends starting Clicks campaigns at a maximum bid of $3 to $5 a click, and Ads Manager may show bid-strength guidance when a bid looks likely to limit delivery. The auction is relevance-weighted and second-price, so the average cost per click can land below your maximum. The sums below use the maximum to stay on the safe side.
Put those together and a $25 day buys somewhere between five and eight clicks. That is why the smallest sensible test runs for four weeks: you need weeks, not days, to see a pattern at this volume. Start it on a Sunday so the 28 days line up with four budget weeks. OpenAI says the platform has no performance benchmarks yet across advertisers or industries, so the 4% enquiry rate in the table is an assumption, not a benchmark; replace it with your landing page's real figure as soon as you have one.
| Test size | Daily budget | 28-day total | Clicks at $5 to $3 | Enquiries at 4% |
|---|---|---|---|---|
| Minimum | $25 | $700 | 140 to 233 | 6 to 9 |
| Standard | $40 | $1,120 | 224 to 373 | 9 to 15 |
| Stretch | $50 | $1,400 | 280 to 467 | 11 to 19 |
Fund the test from money you can lose, ring-fenced from the channels that already bring customers. If you cut a working channel to pay for it, you cannot tell whether a dip in enquiries came from the cut or from the test.
A music school's four-week test, sum by sum
This worked example is illustrative. A small music school wants new students and decides to test the channel with the $700 minimum: $25 a day for 28 days, one Clicks campaign, a maximum bid of $4. Before spending anything, it works out what an enquiry is worth.
- A new student pays about $720 in the first year (24 lessons at $30).
- After teacher pay and room costs, 55% of that is gross profit: $396.
- Three in ten enquiries become paying students, so one enquiry is worth $396 x 0.30 = $118.80.
- The school wants to keep half of that as profit, so the most it should pay for an enquiry is about $59.40.
Now the channel's side. At $4 a click and a landing page where 4% of visitors send an enquiry, an enquiry costs $4 divided by 0.04, which is $100. That is above $59.40, so at 4% the channel would lose money on every enquiry. At 8% the cost is $50 and the test passes. Divide the maximum bid by the most you can pay for an enquiry ($4 / $59.40) and you get the rate the landing page has to reach: about 6.7%.
That number turns a vague worry into a job. The school's first four weeks are partly a test of ChatGPT and partly a test of whether its trial-lesson page can reach a 6.7% enquiry rate. If the clicks arrive but the enquiries do not, the fix is on the page (a clearer trial offer, a shorter form, a phone number that is visible on mobile), not in the ad. Running the same sum with your own customer value and your own close rate takes five minutes and tells you whether the test is worth starting at all.
One campaign, three ad groups
Keep the first campaign small enough to read. One campaign with the Clicks objective, a location set to the area you actually serve (OpenAI supports regions, cities and postal codes where available), and three ad groups, each tied to a single theme. There are no keywords to buy. Instead, each ad group carries context hints: plain phrases describing what you offer, who it helps and when it is useful. OpenAI's own example is that "cushioned everyday running shoes for beginners training for their first 5K" tells its systems more than "running shoes". It also says hints inform relevance but are not exact-match keywords or targeting rules and do not guarantee delivery, so expect some drift and read the results with that in mind.
| Ad group | Context hint (plain words) | Sample headline | Sample description |
|---|---|---|---|
| Adult beginners | Adults asking how to start learning piano or guitar from nothing | Piano lessons for adult beginners | Trial lesson, small friendly studio, evenings and Saturdays |
| Children's first instrument | Parents choosing a first instrument for a child aged six to nine | Choosing your child's first instrument | Try three instruments in one trial session |
| Exam preparation | Students and parents preparing for a graded music exam | Graded exam coaching | Weekly lessons built around the exam syllabus |
All of the text above is illustrative. Ad fields are short, and the form in Ads Manager shows the live character limits, so work to those rather than to figures quoted elsewhere. Write the headline as the answer to the question in the context hint, and make the description carry one concrete offer. OpenAI's ad guidance also asks for several distinct headline and description variations per offering, each with a different angle, rather than one line repeated. Ads appear below a ChatGPT response, clearly labelled, with your brand name and logo, so the person has usually just read an answer and is deciding what to do next.
One technical check catches people out. OpenAI reviews every landing page before an ad can run, and its help page says pages must not block its crawlers (OAI-AdsBot and OAI-SearchBot). If your website's firewall or bot settings turn those away, the ad cannot be reviewed and will not run, whatever its wording. Ask whoever manages the site to confirm before you submit.
Give each ad group its own landing page or a clearly separate section of one page, so a parent asking about a seven-year-old does not arrive on a page about adult piano. If you want to see how the other big platforms structure the same idea, How Meta and Google Use AI to Run Your Ads: A Plain-English Guide is a good primer for comparison.
Creative enhancements: decide before the default decides for you
This is the setting most first-time buyers will miss. OpenAI's help page says creative enhancements are enabled by default for new and existing self-serve campaigns from 30 October 2026. When they are on, OpenAI may change four things, and the changed ads can run without you reviewing or approving each version:
- Rewrite or replace the headline and description to fit the conversation.
- Translate the ad text into another language.
- Resize the image to fit the format.
- Optimise the destination, sending the click to a different page on your website.
OpenAI says there is no preview of every variation, and its reporting runs at campaign, ad-group and ad level, so you will not see which wording won a given click. You opt out under Manage settings in Ads Manager. There is an account-level default, a campaign-level setting overrides it, and opt-outs you saved earlier are kept. The help page describes one setting rather than a switch for each of the four changes, so check on your own screen whether you can keep resizing while refusing rewrites before you assume either way.
| Your situation | Setting for the first test | Why |
|---|---|---|
| Prices, guarantees, qualifications or regulated wording in the ad or on the page | Off | You cannot approve text you never see, and a rewritten claim is still your claim |
| Landing pages tied to specific UTM tags or offers | Off | A swapped destination page may not carry your tags or offer, and the help page does not say how tracking parameters are handled |
| You want to learn what your own wording does | Off | Otherwise you cannot tell a copy effect from a rewrite effect |
| Week five onward, with the first result in hand | Try on in one campaign | A fair comparison needs a baseline run with it off |
OpenAI's help page itself says that changes over time do not, on their own, establish the effect of creative enhancements. That is a fair description of why you should hold them off until you have a baseline.
Tracking before the first click
Three layers of tracking, in order of effort. First, put UTM tags on every destination URL, which OpenAI recommends, so your own analytics can see where visitors came from. An illustrative tag looks like this:
https://example.com/lessons?utm_source=chatgpt&utm_medium=cpc&utm_campaign=first-test&utm_content=adult-piano
Second, make the enquiry itself countable: a thank-you page that only appears after the form is sent, or a form-submission event in your analytics. Third, set up the pixel or the Conversions API as Ads Manager's measurement page directs, so conversions show beside clicks in OpenAI's reports. If the third layer is out of reach this month, the first two are enough for a four-week test.
Two reporting quirks will save you a worried phone call. OpenAI notes that its ad-review crawler may visit your landing pages with the tracking tags attached, so your analytics may log a few paid-looking visits that are not billable clicks. And spend figures in Ads Manager can lag 7 to 8 hours behind impressions and clicks, which refresh about every 15 minutes, so a spend of zero on a busy morning does not mean nothing has been charged.
Add one more cheap layer: put "ChatGPT" in the options of your "how did you hear about us?" question. Some people who see an ad will search for you later or phone instead of clicking, and that answer is the only way to count them. Set Up Conversion Tracking Before You Let AI Spend Your Ad Budget walks through the basics for any ad platform.
Checkpoints that tell you when to stop, fix or continue
Set the rules before the money moves, because a half-spent budget makes everyone optimistic. These checkpoints use the $25-a-day test and are my own rules of thumb, not OpenAI guidance.
| Checkpoint | Spend so far | Question | If the answer is no |
|---|---|---|---|
| End of week 1 | About $175 | Are the ads delivering, with clicks arriving at a cost near your bid? | Check the ad shows as Serving, the bid-strength guidance is not flagging your bid, and the context hint is not too narrow |
| End of week 2 | About $350 | Have any enquiries arrived, and are visitors staying on the page? | Fix the landing page before spending more; the ad is doing its job if clicks arrive |
| End of week 4 | About $700 | Is cost per enquiry below your ceiling (the $59.40 in the example)? | Stop, or run one more month with one specific page change, not a bigger budget |
Two things not to do with a small account. Do not compare click-through rates with Google search ads, because a conversation placement is a different audience in a different mood. And do not split the budget across many campaigns: at $25 a day, three ad groups already thin the data, and each extra campaign makes it harder to read.
When you have a cost per enquiry, put it next to the channels you already use. If you also run Google Ads, Can AI Manage Your Google Ads Without an Agency? is the comparison baseline. For a shop with a product feed rather than a service business, ChatGPT Ads Cost for a Small Store: CPC, CPM and a Test Budget covers the product-ad side.
If you sell finance, legal or health services
The category gate bites hardest for the professional firms this series covers. OpenAI's ad policies page says that during the test period ads are mainly limited to consumer verticals: lifestyle and household goods, local services, travel and experiences, and digital products or education. Financial services, health and legal services are restricted, with approvals reviewed manually, case by case, and proof of licence may be required. The policy bans ads for credit repair, debt settlement, debt help and alternative investments such as bullion, and says financial and legal services ads are generally prohibited outside the one market where they are currently open.
The eligibility table in OpenAI's onboarding and policy troubleshooting guide goes product by product. Financial software is allowed everywhere. Insurance, deposit accounts and auto loans are allowed in one market and restricted to select approved advertisers elsewhere. Mortgages, credit cards and investing are restricted to select approved advertisers wherever they are allowed at all. Financial planning, personal loans and payments are restricted in one market and not allowed anywhere else. Debt relief, payday loans and earned wage access are not allowed anywhere. Legal services are allowed only in one market, and only for advertisers licensed to practise law there. Above the table OpenAI adds that it is working with a small group of advertisers and is not accepting applications for approval in some categories at all, so read the policy on the day you apply rather than relying on any summary, including this one. The policy's changelog also says medical, legal and financial advice conversations are no longer blocked from ads by default, while sensitive contexts such as conversations about mental and personal health stay off limits.
What this means in practice:
- Approval first, build second. Do not pay for creative work or tracking set-up until you know your category is allowed, and ask OpenAI through Ads Manager support if the policy does not say which bucket you fall into.
- Do not assume your job title decides it. The eligibility table lists financial software as allowed and financial planning as restricted, but it does not mention bookkeeping or tax preparation. Ask OpenAI which row your firm falls under before you plan.
- Your own rules still apply. What OpenAI allows is a floor, not permission under your profession's advertising rules. Can You Use AI to Advertise a Professional Practice Within the Rules? covers that side.
- Think about the conversation, not just the keyword. A care agency selling to families should read the sensitive-context rules before budgeting, because the conversations where families ask about care may be exactly the ones where ads are kept out.
For firms in the sectors that have their own tutorials, How to Advertise on ChatGPT: Approval, Budget and Reporting for Agencies covers the buying side for those who buy on behalf of others, and the further reads below cover law firms and estate agents.
A launch-day checklist you can tick through
- My country and business category are allowed, and I have approval if my category needs it.
- I asked five customers which ChatGPT plan they use, and enough of them use the free or Go plans.
- The account is verified, billing works, and country, currency and time zone are the ones I will report in.
- I know the minimum daily budget Ads Manager shows today and the bid it suggests.
- I worked out the most I can pay for one enquiry, and the landing-page rate that makes the test pass.
- Creative enhancement is set the way I decided, at account and campaign level.
- Every destination URL has UTM tags, and an enquiry produces a thank-you page or an analytics event.
- My enquiry form asks how people heard about us and offers ChatGPT as a choice.
- The stop rule is written down with the three checkpoint dates in the calendar.
Treat the first month as buying information, not customers. If the arithmetic says the channel can work for your business, you will know which page, which offer and which budget to try next. If it says it cannot, you will have spent a few hundred dollars finding out, which is cheaper than finding out after a quarter of drift.
Questions before you fund a ChatGPT ads test
Is there a minimum budget for ChatGPT ads?
Yes. OpenAI's help centre sets the standard minimum daily budget at $25 for dollar-billed accounts, with separate minimums for other billing currencies. A daily budget is a weekly average: one day can bill up to twice the figure, but a Sunday-to-Saturday week never more than seven times it. Read the minimum Ads Manager shows before you size your test.
Who actually sees ChatGPT ads?
Adults using the Free and Go plans. OpenAI's help page says Plus, Pro, Business, Enterprise and Edu accounts do not see ads, nor do accounts identified as under 18, and Free users can switch to an ads-free mode with lower limits. If most of your customers pay for ChatGPT or do not use it at all, a test will reach fewer of the right people than the impression numbers suggest.
Can I target ChatGPT ads by keyword?
Not in the way Google Ads works. There are no exact-match keywords. You describe what you offer and when it helps in plain words, called context hints, at ad group level, then set locations and platforms. OpenAI says hints inform relevance but do not guarantee delivery. Custom audiences exist, but including one needs at least 25,000 matched users.
Do I need an agency to buy ChatGPT ads?
No. Self-serve sign-up opened after OpenAI's pilot ended in April 2026, and you can create your own account. OpenAI's set-up guide says one person creates the advertiser account and invites others as Admin, Member or Viewer, so an agency or freelancer can be invited in while you keep ownership of the account and billing.
Further reads
- ChatGPT Ads for Law Firms: Can You Advertise, and Is It Allowed? — How the same questions play out for a firm in a restricted category.
- ChatGPT Ads for Realtors: Access, Cost and Fair-Housing Rules — Access, cost and housing-ad rules for another regulated sector.
- AI for Creating Advertisements: One Ad Brief, Five Formats — Write one brief and adapt it into the short formats ads need.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI's ad policies page and its ChatGPT Ads help centre (Ads in ChatGPT, The Basics, account set-up, availability, creating campaigns, daily budgets, ad groups, creating ads, creative enhancements, onboarding and policy troubleshooting, FAQ); checked 4 November 2026.