ChatGPT Apps Aren't What You Think (And Why You Can't Sit This One Out)

ChatGPT's App Store launched six days ago to 900 million users, and most companies are about to make the same mistake: they'll compress their existing product into a chat interface, recreate their navigation flows, and wonder why nobody uses it.

They'll fail because they don't understand what just changed.

For decades, users went to software whenever they needed to get something done. You opened the finance app to move money, the project tool to update tasks, the CRM to log a call. Every task required remembering which app to open, navigating through menus, and adapting to how that particular tool wanted you to work.

Not anymore.

Software now comes to the user, mid-flow, exactly when intent exists. That's not an interface upgrade, it's a fundamental shift in how software works, and the companies that understand this distinction will disrupt the ones that don't.

Here's a screenshot of the ChatGPT App Store:

I click onto an App, like Photoshop:

And now I can work with the App in ChatGPT:

Here's what most teams will do: "We have a product. Let's bring it into ChatGPT". They'll try to own the screen, design flows, and recreate the experience they've spent years building.

This is the wrong approach.

ChatGPT is not a destination where users commit to your environment. Users don't open your app, don't land on your home screen, don't learn your navigation. They're mid-conversation when the model decides your app should appear (that'll come, today you invoke the app), which changes everything about how you need to think about design.

The best way to think about it: apps now live inside the conversation where intent already exists, not in separate destinations users have to remember to visit.

When someone says "I need to book a car", the software acts in that exact moment, not after they remember which icon to tap or navigate three menus. Right there, in the flow of the conversation.

Therefore, the interface disappears and the conversation becomes where work happens, mid-workflow.

This isn't about making prettier screens or faster features. This is all about designing outcomes inside intent, which means the fundamental design question shifts entirely. From "Where should the user go next?" to "What can we help accomplish right now?"

Not every product belongs in ChatGPT, and understanding why will save you months of wasted development. Here's how to evaluate if yours does:

Can you connect to live data from proprietary systems, permissioned information behind authentication, or user-specific context and transaction history? You're extending the model's reach into places it can't see on its own. A real estate app that pulls live NBA or NRL listings filtered by user preferences and financial pre-approvals is valuable because ChatGPT can't access that data alone.

Can you actually create records, send messages, and trigger workflows in real systems? Book appointments, approve transactions, deploy code, update databases, not just tell users how to do these things. An HR app that actually submits PTO requests, updates calendars, and notifies teams is genuinely useful, versus one that just explains the process.

Does your app create comparison tables, filtered shortlists, or visual representations of complex data that make trade-offs clearer? A travel app that shows flight options in sortable tables with price, time, and airline comparisons delivers more value than narrative descriptions of the same information. This is where generative UI becomes an interesting talking point for brands: do you let GPT render a UI of your company's data and experience? Or would you prefer to own that experience end-to-end within the GPT interface?

The filter is fairly straightforward to articulate: if your app doesn't clearly solve at least one of these three problems, it's probably too thin to be worth building.

The best ChatGPT apps will look almost trivial from the outside because they don't recreate the full product, they expose just the few things the product does exceptionally well. They don't build mini-products that replace the core product experience. Instead they build leverage.

Your first instinct might be: "This will end badly. Platforms always close down, suppress distribution, and take their cut. I'm not playing this game".

Brian Balfour (founder of Reforge and former VP of Growth at HubSpot) calls out that this creates a prisoner's dilemma, and he's studied enough platform cycles to know why opting out doesn't work.

There is no opting out. If you don't integrate, your competitors will, and customer expectations will shift regardless of your decision. You'll be forced to respond anyway, but from a position of weakness instead of strength.

Real example:

The expectation shifts. The question becomes "Why can't I do this in ChatGPT?" not "Do you have a ChatGPT app?".

We've seen this movie before.

Both wrong, and for the same reason: customer behaviour changes faster than companies want to believe.

It's the same reason agentic commerce is coming faster than brands can accept, often pushing back on whether the transaction will in-fact end up taking place on an AI platform like ChatGPT. Without a doubt it will (spoiler: it's already happening, just not in the ANZ backyard), and it will dwarf the size of the ecommerce market.

Back to ChatGPT apps.

The evidence this shift is already happening is everywhere. Publishers are seeing ChatGPT drive more traffic than Twitter. People are shopping, planning travel, and analysing spend inside conversations instead of opening dedicated dashboards. Not as tech demos but as default behaviour.

This isn't about distribution anymore. It's about proximity to decision-making. When someone asks a question, compares options, or decides what to do next, can your software act in that exact moment? That's the new competitive edge, and it matters more than having the best features or the prettiest interface.

Balfour has studied every major platform launch for over 15 years, from Facebook to Google to iOS, and he's identified a pattern that repeats with brutal consistency. Every platform follows a four-step cycle, and understanding where we are in that cycle right now is critical for making smart decisions.

There's consensus around the category, but the stakes are high because these markets tend toward monopoly or duopoly outcomes.

The winner identifies their defensibility (for ChatGPT, that's context and memory) and realises they can't build that moat alone. They need an ecosystem to accelerate growth, so they open a third-party platform with a clear value exchange: "Give us use cases and users to strengthen our moat, and we'll give you distribution you can't get anywhere else".

Developers flood in because the organic distribution is massive and free. Early movers capture compounding advantages that last for years. This is where startups win by moving faster than incumbents can respond.

Eventually, organic distribution gets suppressed, paid distribution becomes mandatory, and the platform absorbs the highest-value use cases into first-party features. Revenue share drops dramatically: Udemy went from 80% to creators down to 15-20% in a single move. Facebook killed notification access for third parties. Google turned search results into 90% ads and Google features. LinkedIn suppressed organic reach to force ad spend.

This step always comes, and the only question for brands playing in this space is timing.

ChatGPT just launched their App Store, which means we're entering Step 1-2, the opening. Balfour's prediction based on current platform velocity: we have about six months before the transition to Step 3 begins.

The urgency will become more real because, as Balfour notes, "the cycles seem to be getting shorter and shorter, so you actually have a smaller amount of time to play the game". Facebook took roughly five years to move through all four steps. This will take six months.

Why does being early matter so much? Zynga hitched to Facebook's platform in 2007 and became a billion-dollar company. Companies that bet iOS-only (despite a smaller user base than Android at the time) eventually captured 70% of mobile revenue. Cursor disrupted GitHub Copilot's market share in nine months by moving fast on AI capabilities.

By the time everyone calls this "obvious" in 2026/27, the best positions will already be taken. The compounding advantages of being early are real, measurable, and historically consistent across every platform cycle.

Balfour's framework for playing the platform game comes down to two core rules:

For startups, the strategy is clear: you can't spread resources across multiple platforms hoping to hedge your bets. You must pick one platform and go all in, because as Balfour observes, "all the failures are the ones that tried to play multiple games at once with scarce resources".

Retention and engagement over vanity metrics. ChatGPT shows the "smile curve": retention that actually goes UP over time as users build habits and context. Balfour notes he's only seen this retention pattern with Slack and other massive winners, which is a strong signal that ChatGPT has genuine staying power beyond hype.

User quality over raw numbers. The iOS vs Android example is instructive here: iOS had 30% of devices but captured 70% of revenue, while Android had 70% of devices but only 30% of revenue. If you'd only bet on Android because of the bigger user base, you lost. Quality matters more than quantity.

Understand the value exchange. What is the platform giving you in exchange for building on top of it? How can you leverage the rules before everyone else figures them out? The teams that win are the ones who understand the platform's incentives and optimise for them early.

Scale with momentum. ChatGPT has 10x the monthly active users of Claude, combined with better retention curves and the smile curve pattern. When you're choosing where to focus scarce resources, those signals matter.

Now here's the paradox in Balfour's exit strategy: You need to enter the game knowing you'll eventually need to exit it. Step 3 always comes, which means the platform will close, suppress distribution, or absorb your use case. So you build your exit strategy from day one by owning an important part of the workflow the platform can't easily replicate, accumulating specialised data and context they don't have access to, and creating micro network effects that make you sticky regardless of platform changes.

Be so good at one specific thing that the platform won't want to absorb it, or can't replicate it even if they try.

As Balfour puts it: "It's better to be early, know you need an exit strategy, and figure it out along the way versus waiting, being late, and then knowing the exit strategy but missing the opportunity".

This is the first new distribution channel since mobile launched in 2008, but with a critical difference: 900 million users are already there. There's no cold start, no convincing people to download a new app, no waiting for the platform to reach critical mass.

Distribution is dead or dying everywhere else. The traditional search channels as we know it are declining as AI answers replace clicks. Social platforms are suppressing external links to keep users on their properties. Paid acquisition costs keep rising as competition intensifies.

This is the window where early movers capture advantages that compound for years.

Balfour frames the fundamental startup game as "get distribution before the incumbent can copy", and incumbents move slower on new platforms because they have more to protect and lose. That's your advantage. That's the disruption opportunity. But the window is six months, not five years, which means you need to move now if you're going to move at all.

Three actions for January:

1. Evaluate your bet. Use the above evaluation criteria to assess whether ChatGPT is the right platform for your product and audience. Don't try to spread resources across multiple platforms. Make a focused decision and commit to it.

2. Filter your idea. Does your app provide access ChatGPT doesn't have, execute actions it can't do alone, or present information better than pure text? If not, refine your concept until it does at least one of these three things exceptionally well.

3. Plan your exit. What specialised value can you build that the platform won't replicate? Start accumulating that advantage now, while you're also building distribution on the platform.

The question isn't whether to play or not... Customer expectations are shifting whether you participate or not.