Calculator and notebook with hand-drawn charts on a light desk illuminated by natural light
ChatGPT Ads

How much does it cost to advertise on ChatGPT? Budget, CPC, CPM and an initial test

How much does it cost to advertise on ChatGPT? How CPM, CPC and conversion optimisation work, daily budgets and how to set up an initial test.

Digital Creation

Team

Quick answer: there is no rate card: the cost of ChatGPT Ads is determined by an auction. You can pay per impression (CPM) or per valid click (CPC), including when optimising for conversions. OpenAI suggests an initial maximum CPC bid of between 3 and 5 USD, but this is a starting point, not a guaranteed cost. Your test budget should be calculated based on your objectives.

Anyone requesting an internal budget needs a figure. For ChatGPT Ads, as of October 2026, the honest answer is not an average price but a method: understand what you pay for, what determines the price and how much you need to spend to obtain useful data. If you need an overview of formats, availability and requirements, start with our guide to how ChatGPT Ads work in Italy. All variable data in this article was verified on 5 October 2026.

The honest answer: the price is determined by an auction

What is and isn’t public

OpenAI publishes the buying models, the type of auction, how budgets and billing work, and a starting bid recommendation for clicks. It does not publish average costs by country or sector, cost-per-acquisition benchmarks or inventory volumes for Italy.

Estimates of CPC and CPM attributed to advertiser tests circulate online. They may be useful as a talking point, but not as a basis for a budget: they cannot be verified, often come from different markets and relate to earlier stages of the programme. We do not use them in this article.

The most reliable benchmark is the one you build yourself. After the first few weeks, you will have CPC, CPM, CTR and landing page conversion rate data for this channel: record them along with the period, offer and settings used, so you can compare subsequent tests on a like-for-like basis rather than against figures gathered elsewhere.

Buying price and cost per result

Three figures are often confused:

  • maximum bid: how much you are willing to pay per thousand impressions or per click;
  • price paid: how much you actually pay, as determined by the auction;
  • cost per result: your spend divided by the results obtained, such as leads or sales.

Only the third tells you whether a campaign is sustainable, and no setting guarantees this. Even in conversion campaigns, the Bid Cap is a bid used to compete in the auction: OpenAI’s documentation makes clear that it is not the price charged for a conversion, nor a guaranteed cost per acquisition.

Available buying models

Views: pay per impression (CPM)

With the Views objective, you buy impressions and set a maximum CPM at ad group level. You pay for the impressions served, regardless of clicks. This model is suitable when you want to raise awareness of a new offer and have already decided how to measure its impact, for example through brand searches or direct visits in the following weeks.

Clicks: pay per valid click (CPC)

With the Clicks objective, you pay for each valid click and set a maximum CPC. For this model, OpenAI suggests starting with a maximum bid of between 3 and 5 USD per click. Ads Manager may also indicate whether your bid is competitive or risks limiting delivery. It is the simplest starting point for an initial test, because the cost is tied to a user action.

Conversions: oCPC and oCPM

With the Conversions objective, the system optimises delivery towards a standard conversion event. After choosing the objective, eligible advertisers choose how to pay:

  • per click (oCPC): you pay for valid clicks, and optimisation takes into account conversions that follow a click;
  • per impression (oCPM): you pay for impressions, and optimisation may also take into account conversions that happen after an ad is viewed, when measurement allows.

In neither case do you pay per conversion. Active tracking and a standard event with sufficient signal are required; for this objective, the documentation states that there is currently no recommended bid.

A note on the documentation: the updated Help guide describes both click-based and impression-based billing for the Conversions objective, whereas the API page on bids and budgets, as of the verification date, shows only click-based billing for conversion campaigns. For setup in Ads Manager, the options visible in the interface and the Help guide are authoritative; anyone managing campaigns via API should check the documentation at the time of launch.

How the auction works and what affects the price

Second-price auction weighted by relevance

OpenAI selects eligible ads through a second-price auction weighted by relevance, with the stated aim of balancing value for advertisers and users. In general terms, in a second-price auction the winner does not necessarily pay their maximum bid, but an amount linked to the competition. Weighting by relevance means the highest bid does not automatically win.

OpenAI does not publish the exact formula. The practical consequence is clear nonetheless: improving ad relevance is a cost lever, not just a quality one.

Factors you can control

  • Stated relevance: context hints written as real-world use cases, not lists of words.
  • Creative consistency: headline, copy, image and landing page all describing the same offer.
  • Targeting breadth: very narrow geographies, platforms and audiences reduce inventory; the documentation suggests considering their combined effect when delivery is low.
  • Bid level: a bid flagged as too low is an indication, not a block, but it may limit delivery.
  • Ad variety: more variations with different angles give the system more opportunities to find relevance.

Budget: how actual spend behaves

Daily and total budgets

The budget is set at campaign level, in one of two forms:

  • daily: the average spend per day over a seven-day period, not a hard limit. On a single day, spend may reach twice the budget, but over seven days it will not exceed seven times the daily amount;
  • campaign total: the maximum for the entire campaign, distributed over time according to the dates.

You can switch from a total budget to a daily budget, but cannot switch back on the same campaign. Daily minimums depend on the account currency. After a pause, ads may remain visible for up to 24 hours, and those costs remain billable.

Post-payment billing and payment threshold

Self-service accounts are billed after delivery. Spend accumulates and the card is charged when it reaches a payment threshold assigned to the account, or at the end of the month. The threshold starts low and may increase with payment history: OpenAI illustrates this with a dollar-denominated example that rises in steps from 25 to 200 USD, while clarifying that this is not guaranteed for every account.

Three implications for anyone monitoring costs:

  • the threshold is not a budget or spending cap;
  • applicable taxes may mean the amount charged exceeds the threshold;
  • spend may appear in reports with a delay of 7–8 hours, so a zero does not mean there are no costs.

A calculation model for the initial test

The formula

A test is intended to produce enough data to make a decision. Calculate the budget backwards from the number of results you want to observe:

Test budget = conversions to observe ÷ landing page conversion rate × estimated CPC

  • Conversions to observe: how many you need to assess the channel without relying on chance. The rarer and more expensive they are, the longer the test will take.
  • Landing page conversion rate: use the page’s historical figure from other channels, reduced cautiously because the audience is new.
  • Estimated CPC: initially, this is an assumption to replace with actual data after the first week.

The number of conversions to observe is an internal decision threshold, not universal statistical proof. In long B2B sales cycles, where the contract comes through weeks later, it is worth adding qualified micro-conversions, such as a demo request or a technical datasheet download, to get signals within a useful timeframe.

If you buy impressions, you can calculate an equivalent cost per click: CPM ÷ (1.000 × CTR). With a CTR of 1%, written as 0,01, a CPM of 20 corresponds to 2 per click.

Add costs that do not appear in Ads Manager to the media budget: creative, landing page adaptation, tracking setup and analysis time.

Hypothetical example

The figures below are illustrative and do not represent market benchmarks.

A local services business wants to observe 30 quote requests. Historically, its landing page converts at 5% on other channels; it cautiously assumes 4%. It therefore needs 750 clicks. At a hypothetical CPC of 2,50 euros, the media budget is 1.875 euros. Adding a 20% margin for the learning phase brings the total to approximately 2.250 euros, spread over four weeks: around 80 euros per day in average budget.

After seven days, the business replaces its assumptions with actual data. If the effective CPC is higher or the landing page converts less, it recalculates: the formula is there precisely to help you understand in advance when the test becomes too expensive relative to the value of a customer.

Impression-based alternative, also hypothetical. If the same business chose the Views objective with a CPM of 25 euros and a CTR of 1%, each click would cost an average of 2,50 euros, as in the click-based scenario. With a CTR of 0,5%, the equivalent cost per click would rise to 5 euros, and the budget for 750 clicks to 3.750 euros. For an initial traffic-focused test, the Clicks objective makes it easier to understand the cost per visit; the choice should still be assessed against the objective, tracking and available inventory.

Framework: which buying model to choose

Answer the questions in order:

  1. Do you have a standard conversion event that is tracked, tested and frequent enough? If not, start with Clicks and measure conversions using UTM and Pixel. If yes, go to question 2.
  2. Do your customers tend to return later without clicking the ad straight away? If yes, consider Conversions with impression-based billing, if available for your account. If not, use Conversions with click-based billing.
  3. Is your main objective to raise awareness of a new offer, without a conversion KPI? Then choose Views, but only if you have already defined how you will assess its impact.

The objective, billing method and event cannot be changed on an existing campaign: plan tests in sequence rather than making corrections on the fly.

Stop criteria and pre-launch checklist

Set the criteria before launch in relation to your target cost per result. The thresholds below are proposed cautious rules, not recommendations from OpenAI.

  • Technical stop: no impressions after 48 hours. Check the status, reasons for non-delivery and bid level before increasing the budget.
  • Adjustment: spend equal to twice your target cost per result with no conversions. Check tracking, the landing page and context hints before changing bids.
  • Continue: cost per result within your target, with a volume you consider significant. Increase the budget gradually, giving the system time to adapt.
  • Close: test budget exhausted, cost per result consistently above target and no improvement in recent weeks. Close the test and document what you learned.

When reviewing results, always separate three levels: the cost of traffic (CPC or CPM), the ad’s ability to generate clicks (CTR) and the landing page’s ability to convert. If the cost per result is high, these three figures show where to intervene: bids and relevance, creative, or landing page.

Pre-launch cost checklist:

  • maximum sustainable value of a lead or sale calculated;
  • test budget calculated using the formula and approved;
  • non-media costs estimated;
  • budget type chosen deliberately, bearing in mind that a daily budget is a weekly average;
  • card and payment threshold understood by the person handling administration;
  • review date and stop criteria documented.

Frequently asked questions

Is there a minimum budget for advertising on ChatGPT?

Daily minimums exist and depend on the account currency; Ads Manager flags when an amount is too low. As of 5 October 2026, we have not found an official minimum amount in euros published in an open source.

Do I pay even if nobody clicks?

It depends on the model. With Views and conversion campaigns billed per impression, you pay for impressions served. With Clicks and conversion campaigns billed per click, you pay only for valid clicks.

Does the 3–5 USD per-click bid recommendation also apply in Italy?

It is the general recommendation in OpenAI’s documentation, expressed in dollars, for click-based campaigns. It is not an average cost for Italy: in an account in euros, use the bid-strength guidance shown in Ads Manager.

With the Conversions objective, do I pay per conversion?

No. The system optimises towards the selected event, but you pay for valid clicks or impressions, depending on the option selected. Cost per conversion is an outcome.

Does spending on ChatGPT Ads improve organic visibility in responses?

No. Ads are separate from responses and, according to OpenAI, advertising does not influence what ChatGPT says. Organic presence depends on content, sources and reputation, and is a separate area of work, like the AI Search Visibility & ChatGPT Ads service.

If you want to size an initial test using your company’s actual figures, talk to Digital Creation about your strategy.

Sources and updates

Updated: 5 October 2026

Did you enjoy this article?

Regione Lombardia

Attività avviata con il contributo di Regione Lombardia – Bando Nuova Impresa 2025