Advertising in the Answer Era: Paid Media in AI Search (2026)
A new category of advertising is being invented in front of us — and it’s not search.
ChatGPT’s advertising business is running toward a $1 billion annualized pace barely a year into testing (OpenAI began trials in early 2026, added CPC buying in the spring). Google has pushed ads into AI Overviews and AI Mode — with “Sponsored” labeled merchants and recommendations — and shipped a whole 2026 stack around it: Conversational Discovery, Highlighted Answers, AI Shopping, and a lead-gen Business Agent. Microsoft’s “AI Max for Search” landed in August.
Even the contrarian made news: Perplexity, after early tests with sponsored follow-ups and answer-adjacent paid content, dropped advertising entirely in February — betting its revenue on AI agents instead. The lesson isn’t “ads don’t work.” It’s that everyone is choosing a different answer to a brand-new question.
How answer ads behave (they’re not search ads)
The early Similarweb data draws the line clearly:
- Google AI Mode behaves like search: 98.5% of its ads appear in response to the user’s first query.
- ChatGPT behaves like a salesperson: less than half its ads arrive in the first response, and nearly a third come after the tenth turn of the conversation — waiting for intent to become clear before pitching.
One is a shelf. The other is a conversation with a commission. “This is not search. This is a new category of advertising.”
What marketers should do now
1. Win the organic answer first. Ads in AI answers currently sit beside answers assembled from sources — being cited is the prerequisite layer. (See our GEO guide: structured facts, consistent footprint, citable claims.)
2. Test where your buyers already ask. If your audience lives in ChatGPT, allocate test budget to its ad products as they roll out; if they search Google, AI Overviews coverage rides on your existing search ads hygiene. Small, measured pilots — the formats are weeks old.
3. Watch the disclosure boundary. OpenAI states ads won’t influence ChatGPT’s independent answers; Google and Perplexity emphasize the sponsored/natural boundary. Trust in the answer is the entire product — expect labeling to stay strict and sudden changes to be frequent.
4. Relearn attribution. A recommendation inside a conversation doesn’t behave like a click on a blue link. Track branded search lift, direct traffic, and assisted conversions around your answer-layer presence — not last-click alone.
5. Keep the message citable. Paid or not, the sentence that survives in an answer is the specific one. Ads don’t excuse vagueness.
The 90-day plan
Month 1: audit your presence in AI answers across major assistants (manual checks per category query). Month 2: fix the citation layer (facts, footprint, pages). Month 3: run first paid pilots where inventory exists, with holdout measurement.
The bottom line
The answer layer is becoming a paid surface — but the currency is still trust and specificity. Get cited first, buy second, measure differently.
When your answer-layer wins deserve amplification, spread them everywhere with Flownib — one message, every platform, native format. The assistants are learning what to recommend. Teach them well.
Share Article