For twenty years, paid media worked like a guessing game. You picked keywords. You placed bids. You hoped the person who typed "project management software" was actually in the market to buy, not writing a blog post, comparing options for a class, or satisfying a curiosity after reading an article at lunch.
Google made that guessing game remarkably profitable. Not because keyword intent is accurate. Because at scale, the noise produces enough signal to justify the spend. You waste 60 cents to find the 40 cents that converts. You build a business on that math, optimize toward it, and call it performance marketing.
February 9, 2026 changed the math.
That's the date OpenAI officially launched ads inside ChatGPT, starting with Free and Go tier users in the United States. By March 26, 2026, the pilot was already expanding to Canada, Australia, and New Zealand. Within six weeks of launch, OpenAI reported $100 million in annualized ad revenue with several hundred advertisers in the program. The company projects $2.5 billion in ad revenue for 2026 alone.
Those numbers are not what matters. What matters is why the format works differently from anything that came before it.
When someone opens ChatGPT and asks a question, they don't type two keywords. They describe their situation. They explain their problem. They ask follow-up questions. They narrow. They compare. By the time a sponsored placement appears, the platform has read a conversation, not a search query. The user has already done the qualification work your landing page was never able to do.
That's not a better version of Google Ads. That's a different category of advertising entirely.
The conversation does the qualification work your landing page never could. That changes everything about what a lead actually is.
Four numbers that explain
why this channel is different.
How contextual matching works.
And why it changes lead quality.
Google built its empire on one insight: when someone types a query, they reveal intent. "Best CRM for small business" tells you something about where a buyer is in their journey. You bid on that signal. You write headlines that match it. The problem is that a keyword is a single data point. It tells you what someone typed, not why they typed it. A user who searches "CRM software pricing" might be ready to buy, or benchmarking a competitor, or preparing a budget presentation. The keyword is the same. The intent is completely different. Your ad hits all four scenarios. You pay for all four. Three don't convert.
Google's machine learning has spent two decades approximating the missing context from behavioral data and audience signals. It has gotten good at approximating. But it is still approximating. Every keyword match is an educated guess about what someone actually wants.
Meta builds detailed audience profiles from browsing behavior, content engagement, social connections, and demographic data. You don't target what someone typed. You target who they are based on how they've behaved across the network. This is powerful for brand building and demand generation. It is poor for bottom-of-funnel conversion because behavioral signals tell you someone might be interested. They don't tell you someone is actively deciding right now. The click that arrives from a Meta ad interrupts someone looking at their cousin's vacation photos. Converting that curiosity into a purchase requires retargeting sequences, landing page optimization, email nurture, and a lot of patience.
Behavioral targeting tells you who. Contextual targeting tells you what they're deciding right now. For high-consideration purchases, that gap is the difference between a lead and a browser.
ChatGPT's ad system reads the full conversation before placing a sponsored result. Not a keyword. Not a demographic profile. The actual dialogue. A user who starts with "what are the best tools for managing a remote sales team," asks "how does HubSpot compare to Salesforce for a 12-person team," then asks "what's a realistic budget for CRM implementation" has self-qualified through their own questions. By the time an ad appears, the platform knows: active evaluation mode, team of roughly 12 people, comparing specific products, budget conversation ahead. The ad that appears doesn't interrupt. It arrives at the right moment in an already-established decision process. OpenAI's system evaluates topic category, intent stage (research vs. comparison vs. decision), query specificity, and conversation depth before determining ad relevance. The system is matching ads to moments, not queries.
Early B2B pilot data suggests cost-per-acquisition can be competitive with Google Search despite higher CPCs, particularly for complex products where buyers need explanation before deciding. The conversation does the qualification work your landing page never could.
Paid media has always worked against context.
ChatGPT Ads work with it.
Why most brands will get
ChatGPT Ads wrong.
The channel is new. The mistakes are already showing up in early pilot data. Four failure patterns. All of them preventable.
01 / Repurposing Google Ad copy
02 / Ignoring the attribution problem
03 / Measuring a new channel with mature ROAS expectations
04 / Skipping organic AI presence before buying paid
ChatGPT Ads vs Google Ads.
The honest comparison.
This is not a takedown of Google. Google Ads remains the most powerful intent-capture channel in the history of paid media. The comparison matters because the two formats answer different questions.
What real performance
looks like right now.
The channel is ten weeks old. Controlled case studies are sparse. But the data points that exist are instructive.
CPMs dropped from $60 at launch to approximately $25 by April 2026. The minimum spend commitment fell from $250,000 to $50,000. CPC pricing launched at $3-5 per click in the same week. Digiday, April 2026. This follows the exact pattern of Facebook Ads in 2012 and LinkedIn Ads in 2015. Early entrants build audience data and creative learnings while the auction is under-competed.
The $50,000 minimum spend is now within quarterly budget for mid-market B2B SaaS companies running paid search. Financial services, B2B software, professional services, education, and technology categories have the highest alignment with ChatGPT's user base. Monks, January 2026. These are categories where users spend extended time researching in AI before deciding.
At THE UN KNOWN, we build GEO programs for clients in aerospace, retail, and defence that establish brand citation in AI-generated answers before any paid spend is placed. When Bombardier's content already appears in ChatGPT answers about aircraft maintenance planning or supplier qualification, a paid placement in that same category carries a Contextual Relevance Score advantage over a brand with no organic AI presence. The paid and organic signals reinforce each other.
Brands navigating measurement well are running three-layer measurement: platform-provided impressions and clicks as a base, branded search volume tracking to catch indirect influence, and direct traffic monitoring with CRM correlation to capture delayed conversions. Adventure PPC, March 2026. Use a 14 to 30 day lookback window minimum. The methodology requires more manual work than a Google Ads dashboard but produces a defensible read on true channel contribution within 60 to 90 days.
Build the foundation. Then buy the placement.
Build the foundation first.
Then buy the placement.
THE UN KNOWN builds the GEO foundation that makes organic AI presence possible, and the paid media strategy that amplifies it. Both together. Not either/or.
Build Your AI Advertising Strategy"The brands that win on ChatGPT Ads are the ones that showed up in ChatGPT answers before they started paying for placement. If you want to build that foundation and own the first-mover window, let's talk."
