When OpenAI introduced its platform for custom GPTs, Josh Brent Villocido was thrilled to find out that one of his innovations would be highlighted.
The emerging AI company revealed during its November 2023 developer day that it would launch a marketplace dedicated to hosting GPTs, customizable tools that utilize its proprietary ChatGPT technology. Users can develop GPTs for various purposes, such as analyzing data, generating tattoo designs, or assisting with customer service. Furthermore, when OpenAI’s CEO Sam Altman addressed the audience at the developer day, he discussed the potential revenue opportunities for developers.
“Sharing revenue is significant to us,” Altman remarked. “We’re committed to compensating those who create the most helpful and widely-used GPTs with a share of our revenue.”
In certain respects, GPTs resemble applications, although OpenAI differentiates between the simpler GPTs and the enterprise applications built on its API. OpenAI’s initiative to create a marketplace for developers is part of its broader strategy to reestablish itself not merely as a chatbot provider but as a leading platform in the AI landscape.
Villocido, a 22-year-old medical student from the Philippines, recognized the potential of GPTs as a means to earn additional income. With no need for advanced coding skills, he successfully developed over 250 GPTs. His most popular creation, Books GPT, offers personalized book recommendations and gained attention when OpenAI featured it during the Store’s launch.
However, ten months following the launch, it appears that revenue-sharing opportunities are limited to a select group of developers participating in an invite-only pilot program managed by OpenAI. Villocido, despite his dedication, was not among those chosen.
As Villocido and other small developers shared with WIRED, their experiences with OpenAI’s GPT Store have had both positives and negatives. Many expressed frustration over the insufficient analytics tools provided by OpenAI, leaving them uncertain about how well their GPTs are actually performing. OpenAI has clarified that creators located outside of the US, like Villocido, do not qualify for revenue-sharing.
Developers who do manage to generate income from their GPTs often create alternative methods, such as incorporating affiliate links or advertisements within their offerings. Additionally, some small developers have leveraged the success of their GPTs to enhance their visibility and attract external funding. For example, the AI-driven scientific research application Consensus, which operates a well-received GPT, secured millions in venture capital financing in August.
Nearly a year ago, OpenAI introduced the opportunity for individuals to create custom GPTs. The company described these GPTs as an automated, low-code method for developing specialized experiences on the ChatGPT platform, expressing confidence that remarkable GPTs would emerge from nontraditional developers such as educators, coaches, and hobbyists.
The Store would be available to customers and creators who subscribed to ChatGPT Plus, Teams, or Enterprise. OpenAI also announced plans to implement a revenue program for GPT builders, stating that participants in the US would receive compensation based on user engagement with their GPT creations. They assured users that further details regarding the payment criteria would be provided.
The official launch of the GPT Store took place in January 2024. By late March, developer Nick Dobos posted on X a screenshot of an email he received from OpenAI, inviting him to join its revenue-sharing pilot program. His coding GPT, Grimoire, has become one of the most popular GPTs in the Store, engaging in over 2 million “conversations” through the app.
The email shared by Dobos indicated that the pilot program ensured a minimum monthly payout of $1,000 for him, with opportunities to earn more based on usage levels. Dobos has not replied to multiple emails and direct messages inquiring about his experience with the GPT Store since the pilot program’s inception.
When inquired about the progress of the program, OpenAI directed WIRED to its help page, which provided a concise overview of the program and mentioned the company’s aspiration to enable more developers to monetize their GPTs in the future.
Similar to Villocido, Adrian Lin, a resident of Singapore, expressed hope regarding the GPT Store. Lin is a dedicated AI researcher who has established an app lab, Adrian AI Lab, as a side project. Although he lacks formal training in web or mobile app development, the announcement from OpenAI that creating a GPT would not necessitate any coding skills encouraged Lin to seize the chance to create and oversee his own products.
According to Lin, fine-tuning his models in alignment with his objectives was a straightforward process.
Lin believes that his Copywriter GPT, designed to draft advertising content, has accumulated between 500,000 and 600,000 interactions. Much like Villocido’s Books GPT, Lin’s version has also been highlighted on the homepage of OpenAI’s Store.
However, Lin cannot specify how popular his GPTs have become or how often they are utilized, as OpenAI only provides “rough estimations” to smaller developers like himself. Additionally, being located in Singapore means he will not receive any payments from OpenAI for the usage of his application.
OpenAI has chosen not to provide comments on the record regarding this matter or share details on GPT usage.
The company has made it clear that there is a difference between GPTs and full-scale applications, as creating a GPT does not necessitate expert coding skills, while developing an application that incorporates OpenAI’s technology is more intricate (and costlier). Some enterprise clients of OpenAI, such as Moderna, engage in both practices: Moderna invests in access to OpenAI’s models for creating its own applications while also producing over 750 GPTs for internal purposes.
Recently, the world’s top-valued AI “startup” concluded a funding round totaling $6.6 billion. It anticipates generating around $3.7 billion in sales this year, according to a report from The New York Times that refers to financial documents shared with potential investors. (It’s expected that its losses will surpass its earnings this year.) In June, The Information disclosed that a significant portion of this revenue stems from various subscription services for its chatbots and the fees for accessing its API.
During a recent press briefing coinciding with OpenAI’s developer day two weeks ago, OpenAI’s Chief Product Officer, Kevin Weil, shared insights on the growing use of GPTs, particularly within their enterprise offerings. “We’ve observed a significant increase in the utilization of GPTs overall,” he noted.
Weil mentioned, “Many of our clients are developing thousands of GPTs internally. There is substantial usage occurring; however, we don’t have any updates regarding the current product and program at this time.”
Even with OpenAI’s ambiguous intentions regarding compensation for smaller GPT creators and the lackluster analytics feedback provided to developers, some creators remain optimistic about the GPT Store as a viable platform.
One such creator, Dmitry Khanukov, developed SQL Expert, a highly-rated GPT that he claims has engaged in over 200,000 chat conversations. Similar to others interviewed for this article, he has not earned any direct revenue from the OpenAI GPT Store, considering his work a “volunteer project.” Khanukov, who also serves as the cofounder and CTO of a property management firm named Dwelly, is based in Florida, but he was not selected for OpenAI’s revenue-sharing initiative.
In a message to WIRED, Khanukov mentions that OpenAI “doesn’t provide any useful statistics” regarding the performance of his GPT.
However, Khanukov notes that his GPT has opened up new opportunities for him. He has formed connections with individuals who require SQL for their work but do not have engineering backgrounds. He has gained recognition on LinkedIn; he states that since launching his GPT, he has welcomed approximately 3,500 new followers on the platform. He also shares that another GPT developer expressed interest in purchasing his SQL GPT, but he turned down the offer because he preferred to retain its original form.
In a similar vein, Adrian Lin expresses gratitude for the direct user feedback he has received and the “attention” it has brought to his small app lab. He mentions that he is now compensated for promoting the app Notion through his GPT.
“I believe this is pertinent to my users since many are involved in digital marketing, and Notion serves as an effective tool for productivity and project management,” Lin explains.
For Eric Olson, creating a GPT for the OpenAI GPT Store has proven to be remarkably rewarding.
His GPT, named Consensus, was showcased on the Store’s launch day in January, alongside Villocido’s Books GPT. Since its debut, the Consensus GPT has engaged in over 5 million conversations and provides a link to Consensus’ official website in each interaction.
This functionality has transformed the GPT into a promotional platform for the startup, which leverages AI to extract information from complex scientific research papers—think of it as a Google Scholar for the era of artificial intelligence. Approximately 10 to 15 percent of Consensus’ new subscribers are acquired through its GPT. “Our experience is likely more favorable compared to others in the GPT Store, as we’ve consistently maintained a presence on the Store’s explore page,” Olson remarks.
This past summer, Consensus secured $11.5 million in Series A funding from prominent investors, such as Union Square Ventures and Draper Associates.
Villocido, the innovator behind the Books GPT showcased during the Store launch, realized that he could no longer rationalize the $20 monthly fee for the ChatGPT subscription that was essential for the creation and upkeep of his customized GPTs.
Currently, he earns a small sum each month by incorporating advertisements into the GPTs he has already developed, utilizing a chatbot advertising tool known as Adzedek. On a favorable month, he can make around $200 in revenue. However, he prefers not to reinvest that back into ChatGPT.