01
ChatGPT has been running its advertising business for about three months since it announced the start of its advertising test on January 26. The initial price for the advertising pilot was about $60 per CPM.
According to recent reports from overseas media, the actual transaction price for advertisers has fallen to between $25 and $45, and even lower on some channels. Looking at the numbers alone, ChatGPT's advertising prices can be described as having been halved.
Of course, the price drop should not be interpreted as a setback for OpenAI's advertising business.
The initial price of $60 per ad was a gesture of entry into the market, or rather, a strategic offer: when the platform first started selling ads, it was selling a sense of scarcity: ChatGPT's ads were entirely new conversational ad placements. For the market, this type of resource was both scarce and carried a strong expectation of entry, so it's not surprising that the early offer included a premium.
Netflix's advertising business in recent years has also followed this market entry strategy.
When its advertising version was first launched in 2022, the CPM was as high as $55 to $65. In the following year, as inventory expanded and Amazon offered a large amount of cheaper inventory, the price fell back to the $20 to $30 range.
ChatGPT and Netflix differ in their ad inventory attributes. Netflix's advertising model is based on a clear value exchange: users accept ads in exchange for a cheaper subscription price.
This contractual relationship is clear, therefore the existence of advertising will not fundamentally affect the product definition.
ChatGPT's value lies in its user experience, where answers are given directly without intermediaries. Therefore, once an ad enters the conversation, the user experience becomes unclear: is the recommendation based on model judgment or a commercially paid result?
This ambiguity at the perceptual level can directly undermine the foundation of users' trust.
Therefore, Netflix's CPM decline is mainly due to increased supply and competition. ChatGPT's CPM decline, in addition to increased supply, also reflects a deeper issue of trust erosion.
Furthermore, according to Digiday, the minimum spending threshold for ChatGPT has been reduced from $250,000 to $50,000. This suggests that OpenAI's current goal is to expand its advertiser base, likely paving the way for a broader auction mechanism and global expansion.
Once the entry barrier is lowered, the price will naturally follow suit.
At this point, the platform begins to face a more practical question: is the advertising market willing to continue paying for this conversational advertising environment, and what is the acceptable price range for them?
This is perhaps the aspect of this round of price changes worth noting.
02
One of the strategic challenges OpenAI faces in developing its advertising business is that the nature of its advertising inventory has not yet been clearly defined by the industry. In other words, OpenAI's niche in the advertising market remains ambiguous.
For example, some people view it as "search-like advertising." Since it resembles search, the market will naturally ask whether it can produce results like performance advertising.
Some argue that ChatGPT's ad inventory relies on a high-quality content environment. Following this logic, advertisers value the user's attention span and decision-making atmosphere, and are willing to pay a premium for such an environment. This is closer to the characteristics of media platforms like Netflix or LinkedIn.
The former is the click logic, and the latter is the exposure logic.
OpenAI, on the other hand, clearly prefers that the outside world understand ChatGPT as a decision-making assistant. According to this positioning, it needs to demonstrate the platform's influence on users' judgment and selection processes. This requires a process of building an advertising value system centered around the influence of decision-making from scratch.
To date, all three market perceptions exist, but none have been validated by the market to the point of supporting long-term, stable, and high prices.
LinkedIn can serve as a reference.
According to data from US digital marketing company Gupta Media, LinkedIn's current average ad price is approximately $39.19 per CPM. In the same comparison, Facebook is around $4.82, Instagram around $7.63, and Google Display around $10.33. LinkedIn is clearly in the higher price range.
Moreover, this high price is not temporary; LinkedIn's advertising prices have been in the high range for a long time.
One important reason is that the market has long accepted its inventory attributes: a mature media environment with clear professional identities, a clear business context, and advertisers knowing how to assess value.
What OpenAI wants to tell now is a similar story: the questions in ChatGPT are more direct, the demands are expressed more proactively, and the decision-making state of users is more focused.
However, unlike LinkedIn, the latter's pricing is the result of more than a decade of education and repeated validation in the professional advertising market, and has sufficient market consensus to support it.
ChatGPT is still in the stage of prioritizing value propositions and lagging behind in validation systems. It is difficult to simply classify it as a mature performance advertising asset, nor can it be fully regarded as a stable and established high-end media environment.
It's more like a new type of experimental asset that exists between brand and effect, and whose definition is not yet complete.
Since inventory attributes are not fixed, prices are more susceptible to the influence of sales strategies and market expectations, making it difficult for pricing to truly stabilize.
03
Based on current product progress, OpenAI is addressing its shortcomings in conversion tracking, aiming to allow advertisers to see what happens to users after they interact with ChatGPT ads.
However, given the current capabilities of advertising systems, charging is primarily based on impressions, while CPC and CPA models, which are closer to the logic of performance-based advertising, have not yet been truly implemented. Targeting capabilities are still relatively rudimentary, and the reports available to advertisers mainly focus on impressions and clicks.
I think the importance of performance measurement has been underestimated by the market.
The reason why budgets continue to flow to Google and Meta is not only due to factors such as traffic and data, but also because advertisers and agencies know how to analyze clicks, keyword returns, and conversion paths on Google and Meta. The process of campaigning, reviewing, optimizing, and increasing investment has been highly refined by the platforms.
This means that once the measurement system is established and a market consensus is reached, the platform gains the right to interpret the budget.
The basic approach to traffic trading is to provide traffic to the market, the market sets the price for me, and I'm left to my own devices. Modern advertising transactions, in addition to sending traffic out to boost sales, also involve defining which results can be priced, which conversions can be attributed, and which metrics are sufficient to support budget expansion.
On the surface, advertisers are still buying traffic, but in reality, they are buying a complete framework of verifiable, replicable, and scalable results.
Looking at ChatGPT at this point, the attribution system is still incomplete and lacks the support of an advertising measurement system. High CPM is difficult to sustain in the long term, and low CPM may not be enough to create appeal.
Because the real concern of advertisers remains unchanged: whether every penny spent can ultimately be proven effective.
04
Of course, ChatGPT is just starting to develop its advertising business, so there's no need to hold it to the standards of mature platforms like Google and Meta.
However, OpenAI itself couldn't wait any longer.
Reuters reported in April, citing Axios, that OpenAI expects its advertising revenue to be approximately $2.5 billion in 2026, $11 billion in 2027, $25 billion in 2028, $53 billion in 2029, and reach $100 billion in 2030.
According to The Information, OpenAI internally estimates that advertising could contribute approximately $102 billion in revenue by 2030, accounting for 36% of the company's total revenue of $300 billion. In other words, advertising will be a crucial source of cash flow in OpenAI's future revenue structure.
It is estimated that OpenAI burned through about $8 billion in cash last year, and this figure may rise to $25 billion this year and further to $57 billion next year. It is not expected to achieve positive cash flow until around 2030.
Even if these predicted figures are revised in the future, they at least illustrate one thing: OpenAI does not have a long buffer period, nor does it have the luxury of slowly cultivating an advertising ecosystem and waiting for commercialization to mature naturally.
Therefore, we can see that:
First, the evolution of OpenAI's advertising products will not completely follow the classic internet path of "experience first, then commercialization," but will exhibit a stronger financial orientation.
It must "make money while running," and the rapid development of self-service systems, channel partnerships, lowering the threshold for advertising, and conversion tracking tools are all preparations to boost advertising revenue.
Secondly, the execution difficulty of OpenAI's current advertising business is actually higher than expected. For ordinary media platforms, monetization usually only requires solving the problem of selling inventory. For example, Netflix adds an advertising system on top of a stable and mature CTV media traffic pool; this layout is very smooth.
OpenAI is now facing the challenge of simultaneously defining its product, advertising format, measurement system, and revenue realization, which has significantly increased its difficulty.
05
However, before a mature measurement system is established, it will be difficult for OpenAI to systematically allocate large-scale effects budgets.
Another reason that cannot be ignored is that the advertising market that OpenAI faces is a fiercely competitive arena, an industry that is already highly concentrated and deeply locked in by Google, Meta, and Amazon.
In this context, it will be difficult for OpenAI to directly challenge the core budget pools of tech giants in the short term. A more realistic approach is to start by targeting the budget margins where pricing logic is not yet fully solidified and advertisers are willing to try new things.
Therefore, I am more inclined to believe that OpenAI's most realistic budget source in the short to medium term is to focus on three types of budgets:
The first category is brand experimentation budgets. Advertisers are willing to pay a premium for a new traffic environment, provided that the budget is manageable, the story is novel enough, and the external communication value is strong enough.
The second category consists of mid-to-high funnel budgets in industries with high average order value and high consideration levels, such as finance, education, B2B software, and tourism. These industries share the characteristic that users often require more time to gather, compare, and judge information before making a decision, making them more receptive to the influence and guidance of a conversational environment.
The third category consists of budgets already spread across LinkedIn, YouTube, and native advertising platforms. The logic behind these budget allocations is based on the user's current reading and thinking context. Advertisers buy these resources because they value users' more focused attention, deeper content consumption, and more complete decision-making processes. From this perspective, OpenAI would have a relatively lower barrier to entry in handling this portion of the budget.
To summarize briefly, ChatGPT's halved ad prices do not negate the inherent value of advertising. On the contrary, it demonstrates that OpenAI has identified advertising as one of its core future revenue engines and is aggressively pursuing it.
After all, advertising is different from APIs and subscriptions; it's an industrialized business that heavily relies on sales, measurement, attribution, agency relationships, and budget inertia. OpenAI's technological advantages can be established quickly, but its advertising capabilities may not be replicated at the same pace.
The most crucial variable for the coming year is whether the advertising measurement system can mature quickly . Once it proves its value, the market will transform it from an experimental inventory positioning into a high-value, essential inventory.
This article is from the WeChat official account "DaokeDoc" , author: Daokedoc, and is published with authorization from 36Kr.
