ChatGPT ads have evolved from a novelty to a significant revenue stream in just six months. Recent research from Graphite and Similarweb reveals that 43% of ChatGPT’s ad-eligible users in the U.S. have been exposed to an advertisement. Graphite estimates the channel’s annualized revenue run rate at $1 billion and that it has attracted tens of thousands of advertisers.
The study is less certain about the future, as advertisers test creative formats, guess at landing pages, and determine the optimal conversation context for an ad to appear. There is no one playbook, and the researchers do not pretend there is.
Search Engine Land first reported on the findings, which are based on a large sample of live ads and related landing pages.
Reach Grew Quickly in Six Months
Forty-three percent (43%) is an incredible amount for a product so new. The figure covers ad-eligible users, not all those who encounter the ChatGPT, underscoring the volume at which the advertising is being distributed throughout the United States within just six months.
The number is noteworthy given the small run rate, but it should be noted that a run rate is essentially a projection based on the current level of performance, not a figure for actual revenue yet. Nevertheless, $1 billion in annualized revenue and tens of thousands of advertisers suggests a channel that has been rapidly scaled.
The performance seems to be improving, as well. Graphite reports that their click-through rates have doubled since launch, a sign of improvement as the ad product matures. With that said, the researchers recommend that ChatGPT not be considered an acquisition channel for the time being. It is a strange message, but one that makes sense within the confines of the burgeoning market.
A Look at the Data Behind the Findings
Graphite and Similarweb combined audience and performance data with a close read of the ads and landing pages themselves.
- Scale: more than 98,000 ads and 8,000 landing pages analyzed
- U.S. dataset: over 95,000 ads from 500+ advertisers, from Graphite alone
- Collection window: April 1 to Aug. 16
The dates are also worth keeping in mind. The dataset ends in mid-August, which means any changes advertisers have made after that date will not be reflected in the results. It’s a crucial consideration for such a dynamic channel.
Top Advertisers Write Fresh Copy for ChatGPT
One of the more interesting findings is around creative. Graphite found that top advertisers create 98% of new copy for ChatGPT, as opposed to lifting ads from other platforms. Messaging is focused on the use cases that people are debating in their conversations with ChatGPT.
That fits with the general caution that brands can’t simply repurpose their Google or Meta approaches. A search ad is a response to a query, while a ChatGPT ad appears inside a conversation, so conversational context, tailored creative and timing can all impact performance.
In practice that could mean ads that are focused around a specific task being undertaken in chat, rather than a broad brand message. Graphite’s finding doesn’t give examples, but the direction is clear: relevance to the conversation is key, and reusing old assets is a shortcut that the leaders aren’t taking.
More Variations Help, but There’s No Magic Number
Those who use more creative variations tend to collect more impressions. Graphite did not find an ideal number of ads, but there is no formula to follow.
The volume is already high among the biggest spenders. Half of the top 10 advertisers run more ads on ChatGPT than they do on Meta. Meta is an established channel, and that is a significant detail, and it means that the largest advertisers are testing heavily while the format is new.
Landing Pages Stay Simple
On the landing page side, advertisers are not overbuilding. Feature, pricing, and product pages are the most common destinations and where most advertisers are focused their activity, utilizing 5 or fewer landing pages.
Graphite’s advice is to begin with existing high-intent pages to test before building out a set of ChatGPT-specific landing pages for better learning and a more budget-friendly experiment. Keeping the number of landing pages limited also makes it simpler to understand the results, particularly since most traffic will go through 5 or fewer.
Therefore, having a small set of landing pages allows you to see which offer or page is more successful.
The First Few Turns Matter Most
On the landing page side, advertisers are not overbuilding. Feature, pricing, and product pages are the most common destinations and where most advertisers are focusing their efforts, using 5 or fewer landing pages.
Graphite’s advice is to start with the high-intent pages to test before building out a set of ChatGPT-specific landing pages to improve learning and reduce costs for a more budget-friendly experiment. A smaller set of landing pages also helps keep things simple and easy to understand, especially since most of your traffic is likely to come from 5 or fewer.
Therefore, you can see which offer or page is resonating better.
Clicks Are Only Part of the Story
Click-through rate is the metric most talked about in the given article, along with impression share, and both are definitely relevant. However, the actual explanation does not go into details regarding how these clicks translate into sales or sign-ups and what advertisers are paying for.
This metric leaves room for interpretation and an advertiser might have a good CTR but a bad ratio of clicks to sales or another desired action. Therefore, every company that wants to advertise through ChatGPT should set up their own conversion tracking to assess the value of their investment. The article suggests keeping budgets experimental which makes more sense now.
Google Still Carries More Volume
None of this places ChatGPT ahead of search. Graphite’s analysis suggests Google carries significantly more volume in terms of advertising spend, so ChatGPT is more of an incremental channel than a substitute.
This has budget implications. Marketers looking to trial ChatGPT Ads do not necessarily have to reallocate from search spend to try them. It makes more sense to set aside a small test budget and evaluate it independently.
A Slice of the Market, Not the Whole
The analysis also needs to consider how big a footprint the market covered by the data can provide. Graphite’s United States dataset covers more than 500 advertisers, and the channel gathered tens of thousands of them; both are significant sample sizes, but 95000 examined ads still represent only a part of the overall market.
Moreover, the tendencies found might not be applicable to every company that wants to launch an advertising campaign. A small regional business that lists a single product page for a niche audience will most likely have a different experience than the top 10 advertisers that place thousands of ads, including many in the ChatGPT channel. Thus, the research provides a useful basis for experiments but should not be considered a universal recipe for success.
How OpenAI Says Ads Get Selected
OpenAI states that it uses multiple signals for choosing the ads, including the context and intent of the conversation, landing page, and the creative and context hints provided by the advertiser. The ads are separate from the answers provided by ChatGPT.
The context hints are particularly notable because they grant advertisers any control over how their ads are matched. Separateness from the answers is a detail that will shape the perception of the advertisers and how they feel about the ad appearing in chatbot format in the first place. It is something that they will want to follow as the format evolves.
Where That Leaves Advertisers
The early evidence favors advertisers who treat ChatGPT as its own environment. This means that they have been testing native creative with a focused budget and adapting as the auction and user behavior evolve.
Growth is the headline, but effort is what the data seems to reward. Top advertisers wrote new copy, kept their landing pages lean and paid attention to where the ads land in the conversation. None of these depend on outspending anyone.
The opportunity, as Search Engine Land frames it, is to learn the channel early while competition and best practices are still developing.
FAQs
Q1: What did the Graphite and Similarweb research about ChatGPT Ads say?
Six months after launch, 43% of US-based eligible users for ChatGPT received an ad. Graphite claims that the channel delivers a billion-dollar annualized revenue run rate and has already attracted tens of thousands of advertisers.
Q2: How much data is there to analyze?
Graphite and Similarweb processed over 98 thousand ads and 8 thousand landing pages. Graphite’s US-advertiser database contains more than 95 thousand ads, including the inventory from over 500 advertisers, from April 1 to August 16.
Q3: Has CTR improved since the launch?
According to Graphite’s data, click-through rates (CTR) per campaign have doubled since launch, and the researchers advise viewing ChatGPT as an experimental channel for now.
Q4: Can advertisers repurpose their Google and Meta Creatives for ChatGPT?
The researchers found that top advertisers use 98% fresh copy for ChatGPT, which is developed around the conversation’s use cases.
Q5: How many variations of creatives should advertisers test?
The researchers identified no particular number of variations to test, but those who performed best utilized a higher quantity of creative variations to capture impression share. Moreover, 50% of the top 10 advertisers used more ads per user in ChatGPT than in Meta.
Q6: What landing pages do advertisers use most often?
Feature pages, pricing, and product pages are the most popular ones, and most advertisers only use 5 or fewer pages. Graphite advises testing with existing high-intent pages.
Q7: What turn of conversation is the best to place an ad?
The best position for ads is within the first few turns, where they secure the highest CTR and impression share, but conversations tend to be long.
Q8: Will ChatGPT replace Google Search for advertisers?
Probably not, as Google has a far greater volume of advertising inventory, according to Graphite’s analysis, and ChatGPT should rather be seen as an addition to the existing channels.
Q9: How does OpenAI make decisions about what ads to show?
Representatives from OpenAI stated that the company uses signals from the conversation’s context and intent, the landing page, and the creatives and context hints provided by the advertiser, while ads do not appear within ChatGPT’s responses.
