Updated 25 October 2026.
ChatGPT Ads and Google Ads do not sell the same thing. Google sells a keyword typed by someone who already knows what they are after; ChatGPT sells a slot beside a conversation in which the need is described before the product category is named. Compare them on CPC alone and you are judging them on the one line where they look alike.
What actually changes between the two?
The targeting unit: Google reads a query, ChatGPT reads a situation, and the other seven criteria follow from that.
| Criterion | Google Ads (Search) | ChatGPT Ads |
|---|---|---|
| Targeting unit | Keyword, audience, placement | Conversation, context, the situation the user describes |
| User signal | Short query | Full need, constraints, level of maturity |
| Timing | Active search | Exploration, comparison, decision |
| Format | Headline, description, extensions | “Sponsored” card separate from the answer: 24-character title, 48-character text, square image |
| Auction | Keyword and ad rank | Second price, weighted by relevance |
| Landing page | Built around a query | Built around a decision, continuing the conversation |
| Measurement | Click, conversion, ROAS, CAC, visible search terms | Six native columns: impressions, clicks, spend, CTR, average CPC, average CPM. No search terms |
| Risk | High CPC, dense competition | Opaque matching, unstable volume, strict ad policies |
On Google you buy the word “crm”. On ChatGPT you write a context hint along the lines of “sales team that wants to replace an overpriced prospecting tool without losing its sequences”. The hint guides matching without guaranteeing it, and the ad never enters the generation of the answer: it sits beside it, labelled.
The “Measurement” row is the most expensive one to ignore. Google shows you the queries behind each click, whereas ChatGPT does not show you the conversation, so you learn through the creative and the landing page and never through a search terms report.
The four cases where ChatGPT Ads loses
On any query where the user already knows exactly what they want, Google stays unbeatable, and will keep doing so.
First, the highly explicit transactional query, such as “buy black running shoes delivered tomorrow”: the intent is complete, the keyword captures all of it, and a conversation adds nothing. “Brand + price” queries follow the same logic, since the user has already chosen and is comparing a price. Google has captured that demand for years, with price extensions and Shopping. ChatGPT has no equivalent at the time of writing.
Then there is urgent local demand. “Plumber Lyon Sunday” stays with Search: you pick a country at campaign level, but a context hint is not a geographic rule, and “only show in Lyon” is not a valid hint.
The last case is the senior B2B target who pays for a subscription. Ads are shown only to signed-in adults on the Free and Go plans, so a director on a paid plan sees no ad at all. We cover this case in ChatGPT Ads for B2B: the Free/Go limit.
When does ChatGPT Ads win?
When the need is described before the category: the user explains their situation, and the situation is what you target.
Take decision in context: “I’m launching an online store and I don’t know which payment solution to pick for international sales.” No keyword carries that whole sentence. A context hint does.
Comparison works the same way. “What’s the difference between a simple CRM and a prospecting tool for a small team?” The user has no brand in mind yet; on Google you would have to buy every competitor’s brand to be present, while on ChatGPT you are present on the need.
The most interesting case is the need described before the category even exists in the user’s head: “My team spends too much time chasing candidates by email.” That person does not yet know they are looking for HR software, and this is where a useful ad changes a decision. Google is absent at that point, because the query has not been typed.
Why are CPCs lower today?
Because a few tens of thousands of advertisers share ChatGPT Ads, against several million on Google Ads. The explanation is mechanical and says nothing about the quality of the channel.
Entry CPCs are observed between $3 and $5, CPMs between $25 and $60. The minimum budget is $25 per day per campaign, with no overall minimum since May 2026. The auction is second price, weighted by relevance: a more relevant creative can win with a lower bid, which turns it into a cost lever.
The gap is temporary. Self-serve opened on 5 and 6 May 2026, and buying opened in Europe on 31 August 2026, across 31 markets at once. Every week, new advertisers enter the auction, and price follows.
Reading the two together, on cost per opportunity
On cost per opportunity rather than on CPC: a click on “brand + price” and a click on a comparison conversation are not worth the same, and a CPC will not tell you so.
The method comes down to three decisions. First, one definition of an opportunity, whether that is a qualified demo, a completed basket or a sales-accepted lead, and the same for both channels, otherwise you are comparing different units.
Second, the same measurement chain: UTMs on every link, the OpenAI Pixel, the Conversions API, and the CRM as the final judge. ChatGPT Ads’ native report stops at its six columns, so quality has to be read downstream.
Third, two roles instead of one winner. Google takes explicit demand, the kind the user already phrases in your words; ChatGPT takes described demand, the kind they phrase in their own. Cutting one to fund the other means giving up a moment in the journey.
One example, with no generalisation. A French HR SaaS company we work with obtained more than 30 qualified demos per month, 12 SQLs, a CPL of around €250, and a demo-to-SQL conversion rate 70% higher than its Google Ads benchmark. Its final customer CAC came in 22% below Google Ads. The details are in the B2B lead generation use case.
Do not read this case as a promise of generalisable results. It only shows that reading on cost per opportunity produces a different ranking than reading on CPC.