ChatGPT has plenty of advertisers, but not much revenue

OpenAI has many advertisers for ChatGPT, but it has not yet turned that interest into much advertising revenue.

According to Dan Salmon, a partner at New Street Research, the number of ChatGPT advertisers grew from about 20,000 at the end of August to around 60,000 in early October. Each click costs about $5, compared with $30 to $50 for similar business software ads on Google. Despite this growth, OpenAI is still likely to fall short of the $2.4 billion in advertising revenue it told investors it expected to earn in 2026, according to The Information.

This gives us a useful picture of where ChatGPT advertising stands today. OpenAI has made it easier for businesses to try its ads, expanded the service to more than 60 countries, introduced a self-service buying system, and started building the targeting and measurement tools that advertisers expect.

Marketers seem willing to give ChatGPT ads a try. The bigger challenge is getting them to keep spending money on them over time.

ChatGPT ads are starting to look like a real advertising platform

When ChatGPT ads first launched, the program was more like a small, closely controlled test. OpenAI initially charged advertisers based on every 1,000 times an ad was shown, known as CPM. It required a minimum commitment of $200,000, similar to the large budgets often needed for Super Bowl advertising, and allowed only a small group of major advertisers to take part.

That approach changed quickly.

OpenAI now allows advertisers to bid based on CPM or CPC, which means paying for every click. It recommends CPC bids of $3 to $5. The company has also tested CPA bidding with selected advertisers, although it has not publicly confirmed that this pricing option is available.

More businesses can also buy ads themselves. When OpenAI expanded ChatGPT Ads into Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam, and Taiwan in September, businesses in those markets could use Ads Manager to buy ads from the start. This expansion brought the service to more than 60 countries.

These changes make it easier for advertisers to test the service. A $200,000 trial needs to be discussed in a major planning meeting. A $5 click can fit into a small test budget. The fact that OpenAI has lowered the cost and made access easier also suggests that advertisers are not rushing to join in huge numbers.

All this raises more questions about OpenAI's claim that, less than 200 days after launch, ChatGPT Ads had reached a $1 billion annualized revenue run rate. The company did not provide evidence to prove the claim. Because OpenAI is privately owned, there is no way to confirm whether it is true. Also, ARR, or annual recurring revenue, is not a legally accepted accounting method. It estimates what revenue might be over 12 months by taking a current growth rate, measured over a day, a week, or a month, and projecting it across a full year. Both OpenAI and its competitor Anthropic often use this measure.

Ad targeting starts with the conversation

ChatGPT gives OpenAI a kind of targeting information that traditional ad platforms do not have in quite the same way: the conversation a person has just before an ad appears.

OpenAI says it can choose ads based on the subject and purpose of the current conversation, descriptions of relevant topics supplied by advertisers, the ad itself, its landing page, and other signals when personalization is turned on. Advertisers can provide “context hints” that describe conversations or topics where their ads might be useful. This gives them another option besides targeting exact keywords.

Advertisers can also upload email addresses and phone numbers to match people against their customer lists. According to Gadget Review's account of OpenAI's documentation, these matched audiences must contain at least 25,000 users.

For marketers, ChatGPT falls somewhere between search ads and ads aimed at specific groups of people. The conversation can show what someone wants right now, while matching customer lists adds a familiar way to use information businesses already have about their customers.

But this also creates a different challenge for protecting a brand's reputation. An ad does not simply appear next to a webpage whose topic can be identified ahead of time. Instead, it may appear during a private conversation that can change direction with every message.

OpenAI says it is working with DoubleVerify and Integral Ad Science to check whether ads appear in suitable settings for brands. It also says it has safeguards to keep ads away from sensitive conversations and conversations with people who may be emotionally vulnerable. According to OpenAI, these partners do not get access to private conversations when checking these protections.

OpenAI is trying to move beyond the usual ad click

Some of the more interesting experiments focus on what happens after someone sees an ad.

OpenAI is testing Sponsored Agents in the U.S. These let people click an ad and start a separate conversation with an advertiser's AI agent. In an example from OpenAI, someone shopping for a dining table can ask the advertiser's agent whether the table will fit in their room, ask more questions, and then follow a link to the company's website.

This changes what an ad does. Instead of sending every person who clicks straight to a landing page, an advertiser can let them have a conversation before visiting the website.

For marketing teams, this creates work beyond buying ads. Someone needs to decide what the branded AI agent knows, what it is allowed to claim, how up to date its product information is, when it should send a person to another place, and how to measure these conversations.

OpenAI has also connected Ads Manager to HubSpot and Shopify. It has added tools that let advertisers create or change campaigns by giving instructions in everyday language. An optional feature can also adjust an advertiser's existing headlines and descriptions to fit the context of a conversation.

These features bring ChatGPT advertising closer to the other marketing tools businesses already use. But they also create more points where campaign rules and oversight could fail.

Visual ads give advertisers more places to appear

OpenAI is also moving beyond ads made up of text.

This month, the company announced a test in the U.S. of visual ads that appear alongside image-generation results. OpenAI says these ads will be clearly marked and kept separate from the generated images. It also says the ads will not affect ChatGPT's answers.

This format gives advertisers another setting in which to reach people. Someone creating an image might be exploring a product, a design, a trip, an event, or another idea where an image-based ad could work better than a text ad.

OpenAI is also expanding its measurement tools. Its partners now include Hightouch, Tealium, and LiveRamp for conversion data. It also works with companies that help track which ads lead to results and measure campaign performance, including AppsFlyer, Adjust, Branch, Triple Whale, Kochava, Haus, Measured, and others.

This work matters because the number of advertisers alone tells marketers very little about how well ads perform. Sixty thousand advertisers could mean that many businesses keep coming back to spend more money. Or it could mean that 60,000 companies are each running small, inexpensive tests.

The reported revenue shortfall makes this difference important to watch. OpenAI has made ChatGPT ads affordable and easy to access, which has helped attract many businesses willing to test them.

The next challenge is harder: getting marketers to keep spending money on ChatGPT ads after they have enough data to compare conversions, determine whether the ads create extra results that would not otherwise have happened, and judge the quality of the customers they bring in against other advertising channels they already pay for.