ChatGPT Ads Hit $1B. The Media Buy Moved Into the Chat
- Aseem Singh
- 1 day ago
- 2 min read
OpenAI just said ChatGPT Ads crossed a $1 billion annualized run rate. Same week, GPT-6 Astra landed and Nvidia closed Hugging Face at $12.9 billion. The labs will keep shouting AGI. The money already moved into the chat.
If you run a marketing shop and you still treat AI as a writing assistant, you are measuring the wrong layer.
What actually changed this week
Astra can run multi-step computer work. Claude Fable 5.1 cut cache-read costs by 75 percent and started watermarking output. Gemini 3.8 Flash dropped again. Cheap open models keep climbing the leaderboard. Fine. Production got faster. That is not new.
The new part is the commercial surface. ChatGPT now has hundreds of millions of weekly users. About one in five queries already carry direct commercial intent. Ads Manager is live across India, Europe, the Middle East, and North Africa. Early partners report most of that paid traffic is new customers, not recycled clickers.
The marketing + AI cut agencies keep skipping
This is the section most newsletters will bury under model scores. Do not.
LAYER 1 — The funnel is a conversation. People are not bouncing from a search result to your landing page. They are asking an answer engine what to buy. If your brand is not named in that reply, the ad you bought on Meta never gets a chance.
LAYER 2 — Paid and GEO now share a brief. Generative engine optimization used to mean “write pages models can cite.” ChatGPT Ads means you can also buy the slot next to the answer. Organic citation plus a tight paid test is the 2026 media plan, not a Q4 experiment.
LAYER 3 — Volume is cheap. Taste is the product. Every shop can now generate twenty variations before lunch. The ones that win will ship fewer, sharper assets that still sound like the brand after the model has had its pass. That is the work we protect at Dzine Prodigy.
BEFORE: a content calendar, a Google Ads account, a hope that SEO holds. AFTER: an answer-engine audit, a ChatGPT Ads test with a real offer, and a human edit layer so the work does not look like everyone else’s mid prompt.
Do this before Friday
Pick one client. Run five buyer questions through ChatGPT and Gemini. Screenshot whether the brand is named. Then spend a small test budget inside ChatGPT Ads on the one offer those questions already imply.
The model wars will stay loud. The agencies that learn to buy — and get cited — inside the chat this month will own the next brief.


Comments