ChatGPT ads cost on average about $16 per thousand impressions and $2.50 per click across the campaigns we manage. OpenAI publishes no CPM rate and recommends starting click bids at $3 to $5. The minimum daily budget is 25 USD in the US. At those averages, a $3,000 month buys roughly 190,000 impressions and 1,200 clicks.
Published price vs what campaigns pay on average
OpenAI documents how ChatGPT ads are bought and leaves the price to the auction. Its guidance is the ceiling most advertisers plan around, and the averages across the campaigns we manage, converted to US dollars, come in under it on every line.
| Metric | OpenAI | Average across our campaigns |
|---|---|---|
| CPM | No published rate (launch quotes ran $25 to $60) | $16 |
| CPC | $3 to $5 recommended starting bid | $2.50 |
| CTR | No published benchmark | 0.7% |
Two things pull the price down. The auction is second-price and weighted by relevance, so a well-matched ad pays less than its maximum bid. And we write context hints that describe the buyer's situation, which raises relevance on every placement. Averages move with the market and the topic, so treat them as a planning figure and the first weeks of your own campaign as the real quote.
What you actually pay
There are two costs, and it helps to keep them separate. The first is media: the money that goes to OpenAI when your ads serve. The second is management: what it costs to build and run the campaign well, whether that is your own time or an agency fee. This guide is about the media cost, because it is the one most people get wrong before they start.
Media is billed on a postpay basis. You are charged after the ads deliver, when unpaid spend reaches your account's payment threshold, and any remaining balance is charged at the end of the month. The charge lands in your account currency, so a UK account is billed in pounds even though OpenAI gives its bid guidance in dollars.
CPM or CPC, and which to pick
You choose the pricing model at the campaign level. CPM charges per thousand impressions and suits awareness once your relevance is high. CPC charges per click and ties your spend directly to people who engage, which is the safer default while you are still learning what converts. Since 21 August 2026 there is also a Maximize results bid strategy that sets and adjusts bids automatically toward your selected campaign goal. Since September 2026, Conversion Optimization is generally available as well: the platform bids toward a conversion event you configure, and OpenAI's first benchmark for it is a median 44% lower cost per result than click bidding, measured within the same advertiser across a sample of 27. Automated bidding is only as good as the conversion history it can learn from, so we keep new accounts on manual bids until the pixel has at least five conversions on the event we want to buy. When to make that switch, and what the six published case studies actually showed, is in the Conversion Optimization guide.
| Objective | You pay for |
|---|---|
| Views (CPM) | Every thousand impressions |
| Clicks (CPC) | Each valid click to your site |
| Conversions | A valid click (oCPC) or a thousand impressions (oCPM), never a conversion |
Switching an existing CPM campaign to CPC is a one-click clone in Ads Manager since September 2026, from the menu on the campaign name, labelled Clone as CPC. The objective on a live campaign cannot be edited in place, which is why it is a clone.
The lever most people miss is that this is a relevance-weighted auction, not a pure bidding war. The closer your ad, your context hints and your landing page match the user's real situation, the less each placement costs. That is why our context-hints method is a cost lever, not just a creative one: better relevance lowers your effective CPC.
What $3k, $5k and $10k a month buy
Rough math makes the budget conversation concrete. At the $2.50 average cost per click across the campaigns we manage, every $1,000 of media buys about 400 clicks. At the $16 average CPM, the same $1,000 serves about 62,000 impressions.
| Monthly media | Clicks at a $2.50 average | Impressions at a $16 average | What it is good for |
|---|---|---|---|
| $3,000 | ~1,200 | ~190,000 | One tight market, a few intent clusters, first signal |
| $5,000 | ~2,000 | ~310,000 | Broader intent coverage, faster learning |
| $10,000 | ~4,000 | ~625,000 | Multiple markets or products, real optimization room |
Signal, not spend, is the thing to watch in the first weeks. Expect the numbers to move in steps rather than a smooth line: flat while the system learns which ads and situations work, then a jump once it concentrates budget on the winners. Judging a campaign in week one is like judging a race at the first corner.
The $100 that is not a charge
When you add a card, the platform places a temporary authorization hold of about $100 to verify it. This is the single most common thing new advertisers misread as a charge. It is not. It never settles into a payment, which is why your outstanding balance correctly shows zero while the hold is pending.
- +It may show as the equivalent amount in your card's currency.
- +It usually releases within a few days, and OpenAI says it can stay visible for eight days or longer depending on your bank.
- +If it is still showing after that, contact your bank first, then OpenAI billing.
Daily budget or campaign-total budget
A campaign carries one of two budget types. A daily budget is an average over a seven-day period: a single day can spend up to twice it, and the seven days together never more than seven times it. A campaign-total budget is the stricter control, a hard ceiling on what the campaign spends across its full run, and since September 2026 it paces itself so the spend spreads evenly across the flight. OpenAI recommends starting on a daily budget while you learn the delivery, and the choice is only partly reversible: a campaign-total budget can be changed to a daily one, and that campaign can never go back.
Campaign-total pacing has a period, and it is not always the one you assume. With an end date, the platform paces to that date, up to a maximum of 365 days. With no end date it uses a default 60-day period, so an open-ended campaign is being paced across two months without anyone choosing that. Setting an end date is how you make pacing match the flight you actually planned. Neither period is a promise: pacing does not guarantee the budget lasts to the end date, nor that it is spent in full, because spend still depends on the delivery the auction offers.
Two rules about changing a campaign-total budget are worth knowing before you touch one. A revised total has to cover what the campaign has already spent, and a total set equal to the amount already spent stops delivery. That is the trap: the campaign goes on reading active while it can never serve again. Raising bids does not rescue a campaign that pacing is holding back either, because a bid-strategy change does not override total-budget pacing, and low bids can still leave the budget unspent.
A plan that targets other countries needs one more line in the budget. OpenAI says some new self-serve accounts can at first advertise only in their home country, and other countries unlock once verification is approved and the account has spent enough on campaigns at home. OpenAI does not publish that amount, so a first campaign in the home market is the cost of reaching the rest.
The published minimums, and where our floor sits
OpenAI publishes a minimum daily budget per billing currency, and the table grew from ten currencies in August 2026 to 30 by late September. The euro row is 15 EUR a day. The full list: 25 USD, 15 GBP, 15 EUR, 20 CHF, 25 CAD, 25 AUD, 25 NZD, 40 BRL, 150 MXN, 2,500 JPY, 25,000 KRW, 725 INR, 350 CZK, 110 DKK, 5,500 HUF, 160 NOK, 65 PLN, 80 RON, 175 SEK, 70 QAR, 65 AED, 50 SAR, 60 ILS, 4.00 SGD, 5.50 MYR, 33.00 THB, 50.00 PHP, 60.00 TWD, 14,000.00 IDR and 19,000 VND. Turkey is on the advertiser list with no lira row yet.
The cheapest market in the world changed hands on 23 September 2026. India held that place all year at 725 INR, roughly 7.60 USD a day, and the seven Southeast Asian and Taiwanese rows that arrived that week all came in under it. Vietnam is the lowest at 19,000 VND, about 0.72 USD a day, followed by Indonesia, the Philippines, Thailand, Malaysia, Taiwan and Singapore. India is still the cheapest of the large markets, and a euro minimum works out at about 450 EUR a month.
The daily figure is a delivery target, and the mechanics are worth knowing before a budget looks broken. The budget week runs Sunday to Saturday in the account's timezone. Spend on any one day will not exceed twice the daily budget, spend across the week will not exceed seven times it, and a day that underspends can be made up later in the same week. A campaign that uses its weekly allowance early can therefore stop delivering until the next week while still showing as active, which reads as a fault and is not one.
Changing the number mid-flight follows two rules that decide what you can be billed. Raise a daily budget partway through a day and that day's ceiling is twice the highest amount that was active at any point in it, so moving 100 to 110 in the afternoon allows up to 220 that day rather than 200. Change it partway through a week and the seven-day limit is recalculated from the moment of the change through to the next Sunday at midnight, which means a mid-week raise does not carry the whole week at the new rate.
Our floor is the platform's: 15 EUR a day, about 450 EUR of media a month. It is where a campaign starts being worth running. That budget buys enough clicks to tell a real winner from noise, to cover more than one intent cluster, and to give the system something to optimize against before the month is over. If you are exploring below that, the honest move is to get launch-ready first: the complete guide covers everything you need, and when the budget is there, we build and run the campaign end to end.
Why a daily budget will not go lower
On a campaign that uses a daily budget with fixed bidding, the daily budget has to be at least the highest effective per-event bid across the campaign's non-archived ad groups, audience bid multipliers included. One ad group with a high bid cap therefore sets a floor under the whole campaign's budget. Lower the offending bid caps first, then save the smaller budget. The rule does not touch automated bidding or campaign-total budgets.
The wrong version of this travels, so it is worth being exact. OpenAI's own example is a 250.50 effective ad-group bid blocking any daily budget below 250.50, and its documentation states that this is not a universal $250 platform minimum. The real minimum is the currency row above. Anything higher that refuses to save is a bid cap in one of your ad groups.
