ChatGPT Ads Are Here. So, What Do We Actually Know?

There has been no shortage of opinions about what ChatGPT is going to do to advertising. Search is dead. AI is the new search. Brands need to figure out ChatGPT immediately. If you aren't there now, you're already behind. 

You get the idea.

The problem is that there has been a lot more conversation about what ChatGPT could become than data showing what advertising on ChatGPT actually looks like today. 

We have now run ChatGPT Ads across a number of clients, and while our testing is still relatively small, we are starting to have enough data to separate some of the reality from the hype.

The short version? There are some encouraging signs. There are also some less encouraging signs. And, perhaps most importantly, ChatGPT Ads have behaved a lot more like a normal advertising channel than you might expect.

First, What Do the Results Actually Look Like?

Let's start with the big caveat: volume is low. Part of that is the network itself, and part of it is us. We have intentionally kept these tests relatively small while we figure out what works and how much confidence we should have in the platform.

Within that limited volume, results have been mixed. For some clients, ChatGPT has performed extremely well. We have seen efficient leads, continued engagement further down the funnel and results that compare favorably with more established channels. There is a catch, though. Some of our strongest results have come from highly recognizable brands.

For lesser-known brands, the picture hasn't been nearly as clean. CPLs have generally been higher than average. We have seen people continue to engage after becoming a lead, which is encouraging, but that engagement hasn't consistently turned into customers.

Does that mean ChatGPT Ads only work for big brands? No. But it does make us a little hesitant to look at an efficient CPL from a household name and declare that ChatGPT has cracked advertising. After all, recognizable brands tend to perform pretty well on Google too.

So, Does It Live Up to the Hype?

Not yet. That doesn't mean ChatGPT Ads are bad. It means they are advertising.

There is an understandable assumption that advertising within a ChatGPT conversation should be uniquely powerful. Someone is actively explaining what they want, asking questions and evaluating their options. In theory, that is an incredibly valuable moment for an advertiser to enter the conversation. And it might eventually be.

But so far, we haven't found a cheat code. The same things that matter everywhere else still seem to matter here. People are more likely to engage with brands they recognize. An inexpensive lead isn't particularly valuable if that person never buys anything. Your offer still matters. Your landing page still matters. And the platform still needs enough conversion data to figure out who it should be showing ads to.

In some ways, the most interesting thing about our ChatGPT testing has been how normal it has been. There are good campaigns. There are bad campaigns. There are clients where we want to spend more. There are clients where we don't. That's not quite as exciting as "AI is reinventing advertising," but it is probably a more useful way to think about the channel today.

Can We Actually Track Any of This?

This was one of our bigger questions going into the tests. ChatGPT is obviously a very different environment from Google Search and other well-known networks, so would the reporting actually correspond with what was happening after someone left ChatGPT?

So far, the answer has mostly been yes. When we compare what ChatGPT is reporting against what we can see in our clients' backend data, the numbers have lined up relatively closely. There are differences, but nothing that looks particularly unusual compared with what we would expect from Google or another advertising platform. In other words, we haven't found a glaring ChatGPT attribution black hole.

That's important. There is always going to be some debate over attribution, and no ad platform should get to grade its own homework. But at least so far, we haven't seen anything that makes us fundamentally distrust the conversion data coming out of ChatGPT. For a product this new, "the tracking seems to work" is actually a pretty meaningful finding.

Do ChatGPT Ads Have A Broader Impact?

This is probably the question where we have the least data, but it is also one of the more interesting ones. If ChatGPT is becoming an increasingly important part of how people research companies and products, does advertising there create some broader lift that isn't captured directly by the campaign?

Maybe. But we haven't seen much evidence of it yet.

When we turn campaigns on, we can see the additional traffic and conversions attributable to those campaigns. What we haven't seen is some huge corresponding jump elsewhere that would suggest ChatGPT Ads are creating a massive halo effect. There appears to be a small, trackable lift. Outside of that, things generally continue to look pretty normal.

Again, there is a very important caveat: these tests are small. If you dip your toe into a new advertising network, you probably shouldn't expect your entire business to suddenly move up and to the right. A larger investment could produce a larger secondary effect. We simply don't have enough data to say that yet.

For now, we would be very cautious about assigning some mysterious, untrackable value to ChatGPT Ads just because they are ChatGPT Ads. If there is a major halo effect, we haven't found it yet.

Does That Mean You Shouldn't Test ChatGPT?

No. Actually, we think the opposite. There is enough here to be interesting.

We have seen efficient performance for certain advertisers. We can track what is happening. Users aren't just clicking ads and disappearing. And the basic premise of reaching someone while they are actively researching or trying to solve a problem is still incredibly compelling.

But there is a pretty big gap between "this is worth testing" and "we need to put a ton of money into this immediately." Right now, we're firmly in the first camp.

ChatGPT's advertising product is young. Conversion tracking is young. Its bidding and optimization systems are young. Advertisers are still figuring out what creative works, OpenAI is still figuring out how to serve the right ads, and everyone involved needs more data. Think about how much data Google has accumulated to figure out which person should see which ad at which moment. ChatGPT is at the beginning of that process.

That doesn't mean marketers should wait five years before doing anything. In fact, there is a pretty good argument for learning alongside the platform. Test it. Figure out how your customers respond. Build some conversion history. See whether performance improves as the product improves. Just don't confuse being early with having found an easy win.

The Bottom Line

If you came here hoping we had discovered that ChatGPT Ads are either the greatest advertising opportunity since Google Search or a complete waste of money, sorry. The answer is less exciting: they're promising.

We've seen enough good performance to keep testing. We've seen enough poor performance to avoid blindly scaling. Tracking has worked better than we might have expected. And we haven't seen evidence that simply turning ChatGPT Ads on creates some enormous, invisible boost to the rest of the business.

Most importantly, many of the rules marketers already know still seem to apply. Brand matters. Conversion quality matters. Measurement matters. Scale matters. And new advertising algorithms need time and data before they're particularly good at their jobs.

Maybe ChatGPT eventually changes the way we think about digital advertising. For now? It's another channel that has to earn your budget.