Paid traffic
Google Ads or ChatGPT Ads: where to invest in 2026
One channel has twenty-five years and detailed attribution. The other has been in Brazil for a week, costs less and rejects half the sectors. How to decide without betting wrong.
7 min readFluxo Ads
On 11 August 2026 OpenAI opened advertising in ChatGPT to Brazilian advertisers. That created the first genuinely new paid media question in years, and one that still has little data-based answer: is it worth moving budget out of Google to test ChatGPT?
The short answer is do not move it, add to it — and only if your sector is accepted. The rest of this article explains why.
The difference is the moment, not the format
On Google, the person typed the problem. They stopped what they were doing to search, and the ad appears right there. It is the highest intent that exists in paid media.
On ChatGPT, they are mid-thought. They asked something, they are reading the answer, and the ad appears in a labelled box below it. That is not purchase intent: it is interest in the topic. It is worth less per impression, which is why it costs less.
Direct comparison
| Criterion | Google Ads | ChatGPT Ads |
|---|---|---|
| Targeting | Typed keyword | Conversation context |
| Intent | High | Medium |
| Volume | Capped by search, but large | Lower, and Free and Go tiers only |
| Billing | CPC | CPC or CPM |
| Attribution | Detailed, through to conversion | Aggregated, for privacy |
| Accepted sectors | Almost all | Retail, e-commerce and travel |
| Predictability | Twenty-five years of history | Beta, rules change without notice |
The attribution row changes operations most. ChatGPT Ads returns aggregated data by privacy design, so the detailed conversion reporting anyone coming from Google is used to does not exist. Measurement has to be built on your site and in your CRM, or you will not know what the campaign produced.
Half the sectors cannot even decide
At the start of the beta, the permitted categories are retail, e-commerce and travel. Health, finance, politics and services considered sensitive are not accepted.
The cost window, and why it closes
In the beta, CPMs have been running below what Google charges for equivalent contexts. That is not generosity from OpenAI: it is what happens on every new platform, when there is a lot of inventory and few advertisers competing.
The pattern repeated on Facebook in 2012, Instagram in 2015 and TikTok in 2020: whoever entered early bought cheap and learned the platform before it got expensive. Whoever entered later paid mature auction prices and still had to learn from scratch.
The difference is that all three above became enormous channels, and ChatGPT Ads may still not. Entering early is a low-priced bet, not a certainty.
How to decide
- Is your sector accepted? If not, the decision is made: stay on Google and work on organic citation.
- Do you already run and measure Google Ads? If not, start there. Testing a new channel with no baseline teaches nothing.
- Is there test budget that may not come back? If every unit of spend has to return predictably this quarter, wait for the beta to end.
- All three yes? Then allocate 10% to 20% of budget to ChatGPT Ads for ninety days, with your own measurement built before launch.
What not to do
- Do not migrate the budget. Google remains the highest-intent channel, and nothing in ChatGPT Ads replaces that today.
- Do not judge it in thirty days. A platform in beta changes behaviour mid-test, and thirty days of aggregated data concludes nothing.
- Do not reuse creative from another channel. The ad sits below an answer, with the reader mid-thought. Feed creative interrupts; there the copy has to continue the conversation.
If you want us to run both channels, that is the paid traffic management service — and the first thing we do is check whether your sector can advertise on ChatGPT before proposing any test.