ChatGPT Ads Just Hit $1 Billion — And Most CMOs Buying It Are Making a Mistake from 1997
On August 31, OpenAI announced that ChatGPT Ads had crossed $1 billion in annualized revenue run rate in under 200 days.
They opened self-service buying to advertisers across India, Europe, the Middle East, and North Africa. Tens of thousands of advertisers are already in. Omnicom alone has placed more than thirty clients into the pilot. Every week since has brought another capability — custom audience uploads, conversion optimization, geo-targeting.
The trade press has settled on the same framing every new channel gets: treat it as a test budget, layer it on top of the stable media plan, and see if it performs.
Table of Contents
That framing is the mistake, and it’s not a new one. Anderson, Day, and Rangan named it in MIT Sloan Management Review in 1997.
They wrote about channel strategy in general terms, decades before anyone had heard of a large language model. Companies routinely treat channel design as logistics — a distribution problem to bolt onto existing capability — when it is actually a strategic decision that must be designed from the customer back.
Most CMOs approving a ChatGPT Ads test budget right now are doing exactly what that paper warned against. The ones who get it right this quarter will be running a fundamentally different playbook than the ones who just add a new line item to the media plan.
Channel Design Is Strategy, Not Logistics — and Almost Nobody Treats It That Way
Anderson, Day, and Rangan’s core argument is that every channel performs a specific function in the customer’s buying journey — awareness, education, comparison, transaction, support.
The design question isn’t ‘how do we get into this channel,’ it’s ‘which function does this channel actually serve, and does our existing playbook match it.’
Search ads intercept a customer who already knows what they want and is comparing options. Social ads interrupt someone who wasn’t looking and has to be stopped mid-scroll.
ChatGPT is neither. A person typing into ChatGPT has already explained their goal, their constraints, and what matters to them, in their own words. They are asking the system to help them reason through the decision.
That’s a presales, decision-support function — closer to a conversation with a knowledgeable salesperson than an impression buy.
It’s a function most media plans don’t have a slot for, because it didn’t exist as a purchasable channel eighteen months ago.
The mistake OpenAI’s own rollout data quietly reveals is that most advertisers are still buying it like search.
They use the same last-click attribution logic, the same creative built for a ten-word query, and the same CPA target ported over from the Google account.
A handful of advertisers are already reporting real numbers doing it differently — one ecommerce brand hit 3x ROAS over 28 days, and a technology partner found more than 80% of its ad-driven ChatGPT traffic was net-new customers.
These are the accounts that decomposed the channel’s actual function before deciding how to buy it, not the ones that copy-pasted last quarter’s search campaign into a new interface.
What I Learned Building the Channel Nobody Wanted to Measure Correctly
I watched this exact error almost happen at Bolt Goodly.
A mid-market client wanted us to fold a new conversational ad product into their existing $40,000-a-month paid search budget.
They wanted to hold it to the same last-click ROAS target we used for branded search terms.
On paper, that made sense to the client’s finance team — one media line, one target, one dashboard.
I pushed back because the channel wasn’t intercepting someone at the moment of purchase decision; it was reaching people earlier, while they were still forming the criteria they’d use to decide.
If held to a search-grade last-click bar, it would have looked like it was failing by week six and gotten killed before it had a chance to do the job it was actually suited for.
We split the budget instead. We gave the new channel its own funnel-stage objective tied to assisted conversions and downstream pipeline rather than immediate ROAS.
We reported it separately to the board for two quarters before folding any of it back into the blended number.
It cost us an uncomfortable conversation about ‘why isn’t this on the same scorecard as everything else.’
It also meant that channel survived long enough to become one of the client’s three largest sources of qualified pipeline eighteen months later.
This was a channel that would have been dead in a quarter if we’d let the existing measurement framework make the call.
The Provocative Implication: You’re Not Choosing Whether to Buy This Channel — You’re Choosing Whether to Understand It First
Here’s where this gets uncomfortable for a lot of boardrooms this fall.
The CMOs approving ChatGPT Ads budgets right now are, almost without exception, being asked ‘how much should we spend’ before anyone has answered ‘what job is this channel actually doing for us.’
That’s backward, and it’s the same sequencing error that shows up whenever a genuinely new channel arrives.
The tactical decision (budget, creative, targeting) gets made before the strategic one (function, fit, measurement) is even asked.
Anderson, Day, and Rangan’s second warning is just as relevant here: channel power is shifting toward whoever controls the interaction.
In this channel, that’s unambiguously the customer, who frames the entire conversation before your brand even enters it.
A media plan built for channels where you control the framing doesn’t transfer cleanly to one where the customer does.
The CMOs who matter in this channel two years from now won’t be the ones who moved fastest into the $1 billion run-rate story everyone’s citing this week.
They’ll be the ones who refused to let procurement buy it like search.
They built a measurement framework around the actual function before the first dollar went in.
They were willing to look slower for two quarters while a channel built for decision-support proved itself on decision-support terms.
Before your team allocates a single dollar to chatgpt Ads this quarter, ask the only question that actually matters.
What function is this channel performing in our customer’s decision, and does our current measurement framework even have a category for it?
If nobody in the room can answer that, you’re not running a channel strategy.
You’re running a media buy with better technology behind it.
chatgpt
Related Posts
Discover more from
Subscribe to get the latest posts sent to your email.
