It was 2008 when Apple reinvented the world of applications with its App Store.
By imposing new rules, strengthening quality, and establishing a level of trust, Apple created a marketplace that shaped user expectations, developer actions, and revenue models. The App Store has standardized the way users discover and purchase their apps, while providing developers with a unified framework for publishing and updating them.
And now in 2025, OpenAI is proposing a change that could well prove to be just as structuring, by transforming the way humans communicate with applications. Indeed, its recent foray into commerce (with instant payment) and the integration of applications into ChatGPT (via the new Apps SDK) should not be interpreted as a simple expansion of its products. This is an attempt to transform ChatGPT into a new operating system (OS) to manage the interaction between humans and applications. In doing so, ChatGPT will change the way applications are used.
Focus on Instant Payment and SDK Apps
OpenAI very recently publicly launched:
- Instant Checkout: Users can now purchase products directly in ChatGPT (currently only in the US from Etsy sellers, soon from many Shopify merchants). Through the “Buy” button, they confirm delivery and payment, with the order sent directly to the merchant’s backend. ChatGPT here acts as an “agent” between the user and the merchant via the open Agentic Commerce Protocol (ACP), developed in collaboration with Stripe.
- Agentic Commerce Protocol (ACP): ACP is an open standard that allows AI agents, like ChatGPT, to interact with merchant systems (payment, ordering, fulfillment) while maintaining merchant control and limiting data sharing to a minimum. The ACP benefits all participants: companies remain the official merchant and maintain control of the products, presentation and execution; AI agents can integrate commerce directly into conversations without becoming merchants themselves; and payment providers manage transactions securely using encrypted tokens. This structure ensures transparency, interoperability and scalability, paving the way for a true agent-driven commerce layer.
- The Apps SDK in ChatGPT: OpenAI is preparing a system where third-party services can integrate “applications” into ChatGPT. The application receives the context of the conversation and responds with structured outputs, user interface elements and user journeys. Developers will connect via server APIs, manage authentication, maintain sessions, etc.
Combining all of this, the boundaries between search, application, web and commerce are gradually blurring. Imagine asking your assistant to order your usual stock of dietary supplements and seeing that your assistant immediately finds the right seller, places the order for you, and tracks the delivery — without you having to switch apps or navigate manually. Or, you ask him to spot a certain model of sneakers in your usual size as soon as it becomes available. Repetitive purchases will become frictionless, and searching for rare items will cease to be a manual hunting trip and become a simple automated routine.
Why does this seem like a new OS?
- Abstracting the user interface: In the world of ChatGPT apps, the user interface is conversational and made of structured responses. Custom UI skins, home screens, tailored navigations and comprehensive user experiences matter less. What matters is content, value, task accuracy and responsiveness.
- Protocols and standards: Apple gave us the rules of the application ecosystem; now OpenAI provides (or will provide) apps in ChatGPT with a specification, SDKs, security rules and marketplace policies. In the same way that iOS apps must comply with Apple’s guidelines, here apps must comply with those of ChatGPT.
- The role of the gatekeeper: Apple is the arbiter of its application ecosystem; ChatGPT will be able to manage discovery, ranking, monetization, permissions and access.
- Conversational routing: Rather than opening a brand’s app, the user can simply name the task, and ChatGPT will decide which app to forward it to. For example: “Book me a flight to Tokyo” and ChatGPT will choose one or more travel apps in the background.
We can clearly see that ChatGPT does not “kill” applications: it changes the way they are used by emphasizing the standardization of tasks and efficiency.
What apps and brands will need to do differently
As ChatGPT becomes the primary interface for digital interaction, apps and brands will need to rethink how they are designed, accessed and monetized in this new ecosystem. Here is what they will have to do:
1. Focus on the core service, not the user interface shell: The “skin” of your app becomes less important than what it accomplishes. Product research, booking, analytics, support — what matters is your core logic, your APIs and your data.
2. Build a server-side API designed for AI: You’ll expose endpoints that allow your app to accept ChatGPT requests, send structured responses, track state from round to round, handle errors, and ensure graceful recovery. You will manage authentication, rate limits, versioning and throttling.
3. Comply with ChatGPT app standards: OpenAI will require apps to meet security standards, privacy standards, UI guidelines, response formats, error handling, authorization pathways, throughput quotas, and more. (similar to how Apple evaluates apps for guidelines, security, UI and performance).
4. Map deep linking, routing and canonical entry points: Even inside ChatGPT, you will need to link tasks to your internal flows. For example: ChatGPT says “Go to product catalog → filter → add to cart → proceed to checkout. » Deep links become “links of intent” from the conversation to your internal logic. You will be able to provide templates indicating how ChatGPT should “open” a context in your app logic (with parameters). This is similar to the current use of deep links (eg: myapp://product/123?ref=chatgpt).
5. Monetize, get discovered and promote yourself: Once apps work natively in ChatGPT, brands will need to invest to stand out. They will pay for: highlighting or priority ranking, sponsored suggestions for certain tasks, promotional content integrated into the ChatGPT interface and upselling modules exposed in the app’s feeds. This is similar to what happened in search (Google Ads) and in the App Store (search ads, app promotion). ChatGPT will become a new channel for “in-chat promotion”.
6. Provide signals for better targeting and personalization: As ChatGPT knows the user’s conversation history, preferences and context, it is able to choose which services, products or application offers to present. This gives great targeting power to apps integrated into ChatGPT, but also increases the requirements: you will need to optimize conversion in this agentic environment. Because OpenAI will see user interactions across multiple apps/feeds, it will be able to help match users to offers more intelligently. Brands will need to provide signals, feedback and engagement metrics to drive better recommendation logic.
Measurement and attribution in the age of AI applications
Measurement has always been the cornerstone of smart growth strategies, and that won’t change. What will change is what we measure and how we measure it. In a world where users no longer install apps, but interact with them in AI environments, the traditional funnel will be replaced by seamless, intent-driven experiences.
Attribution will go beyond just counting installs to understanding which prompts and interactions are actually driving results. Instead of measuring downloads, we will track actions taken. Instead of mapping user journeys through screens, we’ll do it through conversations, agent calls, and tasks.
This future requires attribution models that can track intent across multi-step, multi-agent journeys, accurately linking upstream triggers to downstream impact. Just as important: it will rely on first-rate feedback loops — where conversions, cancellations or refunds re-enter the ecosystem. These signals will help AI agents classify, recommend and personalize experiences more effectively, turning every action into intelligence that refines the system.
Measurement in this new era will be lightweight, contextual, real-time, and deeply integrated into the way users and agents interact — enabling smarter decisions for both advertisers and platforms.
The era of apps is entering a new phase. ChatGPT does not mark the end of applications, but on the contrary, it presents the next stage of their evolution. And the big winners will be the brands and developers who truly view ChatGPT as a new OS.
