ChatGPT App Store Struggles Six Months In as Developers Balk

A platform bet that hasn't caught fire yet

When OpenAI used its October 2025 DevDay to introduce the ChatGPT Apps SDK alongside Agent Kit, the message was clear. ChatGPT would stop being only a chatbot and become a platform where third parties could build interactive apps and autonomous agents 5. Coverage at the time framed it as ChatGPT becoming "the new app store" 5.

Six months on, the results are underwhelming. According to Bloomberg reporting summarized by PYMNTS, participating companies are giving the mini-app experience mixed reviews 1. The Los Angeles Times, drawing on interviews with app makers, put it more bluntly: the initiative is off to a sluggish start 2. The catalog has grown to more than 300 integrations, but they are hard to find inside ChatGPT, and partners have deliberately limited what those apps can do 2.

Partners are holding back

The most telling complaint is not technical. Companies are reluctant to hand their customer relationships and payment flows to OpenAI, so many have built pared-down versions of their services rather than full experiences 12. PYMNTS also notes that some partners find consumers hesitant to give credit card details to an AI assistant 1.

Consumer behavior supports that caution. A study cited by the Los Angeles Times found that 96% of shoppers who regularly use AI chatbots also turn to other channels, such as social media, traditional search and retailer websites, along the way 2. For a brand, that makes ChatGPT one stop among many rather than the place where purchases happen. Giving up direct ownership of the customer for a channel that rarely closes the sale is a hard trade to justify.

Developer friction compounds the problem

The two reports agree on the practical irritants. Developers describe a slow approval process and say they get little data on how their apps are performing with users 12. The Los Angeles Times adds complaints about a buggy coding system 2. Hanh Nguyen, CEO of Fractal, a startup that makes chatbot-app tools for non-coders, said publishing apps has been a lengthy ordeal. She said some rejections stemmed from "false signals" raised by OpenAI's AI-assisted review system, which then needed human support staff to resolve 2.

The lack of usage data may matter most over time. Platforms such as Apple's and Google's app stores won developer loyalty partly by giving them analytics to iterate and justify investment. Without engagement numbers, partners have no evidence to bring back to their own leadership when they ask for more resources. That makes the cautious, limited-functionality approach a reasonable response.

The security undercurrent

The app push sits on top of a broader opening of ChatGPT to outside systems. In September 2025, OpenAI added support for connecting ChatGPT to remote Model Context Protocol (MCP) servers in Developer Mode, a feature it described as "powerful but dangerous" 5. OpenAI's documentation labels Developer Mode "elevated risk" and names specific threats: prompt injection from malicious or compromised servers, model errors on write actions that could destroy data, and servers designed to exfiltrate whatever enters the model's context 4. The company offers per-app approval levels. The default lets reads happen automatically and requires confirmation for consequential writes, and a "Never ask" setting is itself flagged as elevated risk 4.

Security firm Noma argued that the feature is easier to use than its "Developer Mode" label suggests. In its view, any employee with ChatGPT access could connect to remote MCP servers without training or oversight, creating an immediate governance problem for security teams 3. MCP connectors and the curated app store are not identical, but the shared concern explains OpenAI's position. A company that wants to keep tight control over what runs inside ChatGPT will likely keep a strict, and therefore slower, review process. Developers experience that strictness as friction.

Reading the moment

Six months is early for any platform, and 300-plus integrations is not nothing. Still, the problems in the reporting are structural rather than cosmetic. Partners don't trust the commercial terms, consumers don't treat chatbots as checkout counters, and developers lack the visibility and tooling that make a store worth investing in. PYMNTS frames OpenAI as aiming to improve the developer experience 1. Faster approvals and better analytics are fixable. Persuading major brands to give up the customer relationship is a harder problem.

For now, ChatGPT's app store looks less like a new storefront and more like a showroom. Companies keep their real business, and their payments, at home. Whether that changes depends on OpenAI showing partners that users who arrive through ChatGPT actually convert, and on giving them the data to prove it.