Updated October 22, 2026. This guide separates what is confirmed from what is our own reading.
News: on Wednesday, October 21, 2026, OpenAI turned on ChatGPT Ads in Argentina, just one day ago, so there is no local results history yet. To help you decide we use data from Mexico and Brazil, where the platform has run since August 11, 2026, and from the United States, where it started in February.
Short answer
For most Argentine companies in an allowed category, a small, measured test is worth doing, but not moving the big budget. ChatGPT Ads works more like Google Ads than like Instagram, but with less targeting control, less reporting data and a channel that is still very young.
And one clarification that causes a lot of confusion: paying for an ad does not make ChatGPT recommend you inside its answers. They are two separate things.
ChatGPT Ads vs Google Ads: the differences that matter
| Google Ads (search) | ChatGPT Ads |
|---|
| How the ad is triggered | By keywords you choose | By the conversation context and the context hints you write |
| Targeting | Keywords, location, audiences, devices and more | Country (and more detail in the US), custom audiences with a minimum of 25,000 matched users, and context hints |
| What is shown | Ads above or below the results | One ad below the answer, labeled as sponsored |
| Who sees it | Anyone who searches | Only users on the Free and Go plans, aged 18 or older |
| What you know about what triggered the ad | Search terms | Nothing: you do not see conversations or topics |
| Auction | By relevance and bid | Second-price auction weighted by relevance |
| Measurement | Mature, with many integrations | Pixel and Conversions API, in beta |
| Maturity | More than two decades | Launched in February 2026 in the US, in Mexico and Brazil since 8/11/2026 and in Argentina since 10/21/2026 |
Sources: policies, OpenAI help center and the MarTech guide of August 11, 2026 (US account data).
The underlying difference is this: on Google you buy searches. On ChatGPT you compete for decision moments inside a conversation. Someone tells the AI what they need and, if your ad is relevant, it appears below the answer.
What the first numbers from Mexico, Brazil and the United States show
There are no official benchmarks: OpenAI has not published reference CPM, CPC, CPA or ROAS. What exists are cases from agencies and advertisers, with few accounts and different periods. They help set expectations, not project results.
| Data point | ChatGPT Ads | Google Search reference |
|---|
| CTR in Brazil (8/13 to 9/22/2026, agency Marktech) | 1.33% | 8.98% (July median) |
| CPC in Brazil | R$ 2.40 | R$ 2.52 |
| CPM in Brazil | R$ 32.05 | R$ 248.51 |
| Conversion in a US test (about 1,500 users, StubGroup) | 0.21% | 3.71% |
More data, from different sources:
- Brazil, evolution: between the first and fourth weeks the conversion rate fell 88% and the cost per conversion multiplied by more than 30, while the agency was increasing the budget. In the sixth week conversion was still 78% below the first. It is a single case and the explanation is the agency's interpretation.
- Mexico, CPM: buyers consulted by trade media mention around US$ 60, a value that is not official.
- Mexico, relevance: a study of more than 50,000 commercial queries, cited by AdTech México, found that 14.35% of the ads analyzed appeared in conversations without enough topical relation. Buyers consulted by Adweek also point to little control over audiences and exclusions, and little visibility into the conversations where they ran. OpenAI is testing contextual exclusion tools with a small group of advertisers.
- United States, high-ticket e-commerce (Common Thread Collective): US$ 9,620 invested, CPC of US$ 4.41, CTR of 0.94% and an estimated return of 3.3 to 6.8 times depending on attribution.
- United States, B2B: CPC of US$ 9.29, CPM of US$ 64.34 and CTR of 0.7%. Among 336 paid clicks, the company identified 146 organizations and 5 matched its ideal customer.
- Channel scale: OpenAI reported more than 100,000 advertisers in a little over eight months. ChatGPT Ads' annualized revenue run rate passed US$ 1 billion in under 200 days, according to AdTech México, although it is an extrapolation and not revenue billed over a year.
How we read it: results vary enormously by category, ticket size and landing page. The channel has volume and attention, but a click arrives with less intent than a Google search. It works best when the product is compared and decided through a conversation, and when your site converts well. There is still not enough evidence to say it is cheaper or more profitable than your other channels.
Which categories can advertise on ChatGPT?
According to OpenAI's policies (updated September 10, 2026), ads are currently concentrated in consumer categories:
Categories with the most direct path
- Household and consumer goods
- Local services
- Travel and experiences
- Digital products
- Education and training
Prohibited or heavily restricted categories
- Financial services outside the US: generally prohibited.
- Regulated health outside the US: generally prohibited.
- Legal services outside the US: prohibited.
- Alcohol, tobacco and vaping, real-money gambling, adult content and dating, recreational drugs, politics and weapons.
- A specific job opening or a specific property for sale or rent.
That leaves out, for now, Argentine banks, fintechs that offer financial products, clinics, law firms and wineries. OpenAI reviews the ad and the landing page, and can reject an advertiser for any reason tied to its advertising principles. If you are in a doubtful category, read the policy before investing in the campaign.
If I pay, does ChatGPT recommend me in its answers?
No. OpenAI states that the ad system is separate from the model that answers: advertisers cannot pay to alter, rank or influence what ChatGPT says. And the ad is clearly distinguished from the answer.
So there are two distinct paths, which can coexist:
- Paid advertising: you appear as sponsored content when the system considers your ad relevant. You can turn it on and off.
- Organic AI visibility (GEO): getting AI assistants to understand who you are, what you offer and to mention you when appropriate. It is content, reputation and digital presence work that is built over time.
It is the same logic as Google Ads and SEO: paying for one does not give you the other. And according to MarTech, OpenAI also clarifies that advertising does not change whether your brand is mentioned organically, nor the other way around.
Who should try it?
Our reading, based on how the platform works:
It makes sense to test if:
- Your category is allowed.
- You sell something people research and compare before deciding (technology, travel, courses, local services, consumer products).
- You already receive traffic from ChatGPT. MarTech points to this as a good sign for testing.
- You have reliable tracking and a landing page per topic.
- You can tolerate a new channel with little data.
- You sell products and can upload a catalog for product ads.
Better to wait if:
- Your category is restricted.
- You need predictable volume now, this month.
- You do not measure sales precisely.
- Your audience is very young or rarely uses ChatGPT's free plans. Remember that Plus, Pro, Business, Enterprise and Education do not see ads.
- You have not yet sorted out the basics: your website, your tracking and your own customer base. In that case, investing there first pays off more.
How to run a 30-day test without burning budget
- Set your benchmark: your current cost per customer on Google or Meta.
- One campaign, Clicks objective and a daily budget near the minimum (US$ 25 per day on dollar accounts).
- Two ad groups with the same ad: one with keyword-style hints and one with a conversational description. In a StubGroup test, the keyword one delivered more, but it is an isolated case and worth verifying.
- Measurement from day one: pixel, Conversions API and UTMs in Google Analytics 4.
- Weekly review of spend, clicks and conversions, without changing everything at once. Expect volatility at the start: in Brazil one account's weekly cost per click went from R$ 1.03 to R$ 4.09 in four weeks. That is why a test of several weeks, with a small budget and without raising it suddenly, is advisable.
- Decision at the end: if the cost per customer gets close to your benchmark, scale gradually. If not, adjust the landing page and the offer before raising the budget.
A warning backed by data: the agency StubGroup compared about 1,500 users who reached the same site from Google Ads and from ChatGPT Ads and found 3.71% conversion on Google and 0.21% on ChatGPT, with less time on site. It is a single test and does not represent the channel, but it is one more reason to measure with your own numbers instead of taking the result for granted.
What ChatGPT Ads does not solve
No new channel fixes a weak foundation. If your site does not convert, if you have no customer base of your own or if your offer is unclear, a more or less expensive click does not change the problem. That is why, before adding another paid channel, it pays to have the underlying business in order: a website that sells, reliable measurement and an owned base you can talk to again.
At We make it Lab we work on paid media, measurement and automation for small and mid-sized businesses. If you want to know whether ChatGPT Ads belongs in your mix or whether it is better to prioritize something else today, write to us and we will look at it with your data.
Sources
- OpenAI ad policies, updated September 10, 2026.
- OpenAI help center: creating campaigns.
- MarTech, "A 7-step guide to running ads on ChatGPT", August 11, 2026 (US data, StubGroup test).
- OpenAI, expansion to 26 countries, October 21, 2026, via CNBC.
- Forbes Argentina, October 16, 2026.
- Marktech, "ChatGPT Ads in Brazil: CTR, CPC and lessons from six weeks", September 22, 2026 (preliminary data from one agency).
- AdTech México, August 28, 2026, and AdTech México, August 31, 2026.
- Search Engine Journal, September 8, 2026 (US cases).
- El Cronista, on plans and user age.
