You Should Never Go Viral With Your AI App

Victor Stepanov does growth marketing at Every and has spent years building audiences at Netflix and BuzzFeed. He understands the seductive pull of virality, and why AI founders should resist it. In this piece, he argues that for AI products, sudden viral growth starves the feedback loops that make them better and chases away the very users that matter most. His three rules of “boring” growth—don’t overpromise, build in public, and talk to users constantly—offer a counterintuitive playbook for building an enduring business.—Kate Lee

Was this newsletter forwarded to you? Sign up to get it in your inbox.

If you’re building an AI product, I hope you never go viral with it.

I hope you never feel the surge of thousands of new signups overnight, the rush of rocketing up Apple’s App Store downloads chart, the unmistakable jolt after your launch post blows up on TikTok or clocks a million impressions on X.

As counterintuitive as it sounds, if you’re building an AI product—especially an agent-native one, where an AI agent in the app can do anything that a human user can do—obscurity can be a hidden advantage, the quiet space where the shape of the product emerges, your best shot at finding true product-market fit.

I come from the mobile app world, but I’ve also spent years doing social media and marketing at companies like Netflix and BuzzFeed. In entertainment, the entire business depends on winning the attention game over and over again. If people aren’t watching or reading, nothing else matters. Today, that belief seems to be everywhere. My X feed is full of people explaining how to make your app go viral. The promise is that if a product gets attention quickly, the rest will sort itself out.

That mindset is powerful. It treats attention as proof and reach as validation. While that works in entertainment, where the product is the content, we’ve seen prominent examples in tech of how sudden, unexpected growth can backfire.

This is an even greater risk with AI products, which reveal and even increase their value through repeated interaction. They need the same users to return again and again—enough for the AI model or agent to learn them and vice versa. When a surge of one-time users arrives and quickly abandons, the churn starves the system of the feedback loops it needs to improve.

So what kind of marketing helps AI products succeed? It’s not really glamorous, and might even seem boring. If instead of reach, AI products thrive on retention and depth of relationship, then your growth strategy has to do the same. It comes down to a few practices that feel almost too simple but work because they compound over time, in the same way your product (hopefully) does.

The rules of ‘boring’ growth

1. Don’t overpromise

A lot of viral growth playbooks start with the same directive: Make your product look as magical as possible. Cut out all the struggle and show the most extreme transformation, an instant life upgrade. Some apps go so far in this direction that they design their core workflows to look amazing on TikTok.

That approach works great if your only goal is to get people to tap “Download.” Flashy product demos inevitably showcase a narrow interaction and a specific use case. This fits traditional apps like Slack and TikTok, which have a set of core flows and success is about repetition. But AI apps, especially agent-native ones, tend to be non-linear, unpredictable, and sometimes have a virtually infinite number of use cases. Instead of a strict recipe of pre-defined instructions to follow, these apps are more like a kitchen stocked with tools that AI agents can use to make your requested dish. For every instance of magic there might be countless others that feel completely mundane—and cause people to quickly bail.

So your marketing has to do almost the opposite of what the viral playbooks say. You don’t want to show one most impressive thing. You want to show the boring but recognizable moments potential customers experience every day. If you’re building an AI writing assistant, you don’t show a perfectly polished article generated on the first try. You show the blank page. You explain the writing principles behind your prompts. You work with the “I don’t know where to start” feeling your audience has and slowly guide them to writing their first paragraph with the agent’s help.

What you’re really optimizing for here isn’t a single “wow” moment but a series of small “aha”s. Ideally all your growth efforts would empower users to discover what Every cofounder and CEO Dan Shipper calls “emergent capabilities”—things the product can accomplish for the user that weren’t explicitly defined by the developer.

2. Build in public

Repetition is one of the oldest rules in marketing. People rarely act the first time they see something. They need to trip over your product enough times to recognize it, trust it, and place it in the context of their lives. That’s why the best time to start talking about your product is while you’re building it.

Some posts will perform better than others, and the product might not always look impressive at first. But this becomes a natural way of attracting the right people—people who care about your approach and the problems you’re trying to solve. Over time, this transparency builds familiarity with your product and trust in you as its builder.

But another—a far more important—reason this works for an AI product is because the builder is uniquely positioned to show how to use it. Boris Cherny, one of the creators of Claude Code, publicly shared how he and various team members at Anthropic use the tool—not as “the” way to use Claude Code but more as a set of patterns and common practices. By building in public, you’re simultaneously marketing the product and teaching people how to work with it, what kinds of questions to ask, and what good outcomes look like. Your example inspires them to tinker with it themselves.

3. Talk to users constantly

Traditional metrics like session length, sessions per user, or “time to value” were designed for software with fixed flows. AI systems behave more fluidly, and the capabilities that make agents useful also make them difficult to evaluate. The meaningful moments are often qualitative, unexpected, and highly contextual—things like a user discovering a new way to work with an agent, or trusting it enough to keep trying after it fails.

That’s why having conversations with users matters a lot, especially early on. Find time to direct message people and hop on calls. Email them. Look for patterns in their feedback. Schedule more conversations to dig deeper. Make small, targeted improvements, then go back to the same users and see if their experience has changed. Talking to users isn’t an occasional activity but a critical part of the loop.

Showing up personally to those conversations does something else. It turns your product into a shared story, one in which users feel like they’re co-creating the product with you—which they are. Figma CEO and cofounder Dylan Field reads and regularly responds to customer feedback on X. I’d argue that this kind of direct communication is even more important for solo founders and small teams because that shared story becomes your brand. It signals care, taste, and accountability in a way no automated response can. In AI products, where behavior is fluid and still forming, that human trust is often what convinces people to stick around long enough for the product to become genuinely useful.

The trade worth making

For consumer AI apps, what’s even more important than network effects—where your product gets more valuable as more people use it—is relationship effects. The memory, personalization, and trust that develops between user and AI—particularly with AI assistants and companions but not limited to them—is part of the product’s value. Like any meaningful relationship, it takes time to form, deepen, and compound. You can’t rush it. And you can’t fake it by driving downloads from thousands of virality-sourced users who were never going to build that relationship in the first place.

This kind of relationship is not trivial for any software product, let alone an AI one. My big bet for 2026 is that for an AI app to succeed it will need to “pre-activate” users—to have already built that relationship with them before they sign up and download it. So by the time they do sign up, they already know what the product is, and they’ve seen the work you put into it. They’ve already decided that this thing that you’ve built is for them because, in a way, you built it together.

If you need to trade virality for these kinds of users, so be it. For a shot at building a truly durable consumer AI product, it’s a trade worth making.

Victor Stepanov runs growth marketing at Every. Previously he ran regional content and marketing at Netflix, BuzzFeed, and Flo Health. You can follow him on X at @vicngmi.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Discover Every’s upcoming workshops and camps, and access recordings from past events.

For sponsorship opportunities, reach out to [email protected].