How do you make AI products go viral?

referral programs, secret hacks, and why the 2010s playbook doesn't work here

One of the most powerful growth engines I watched up close at Uber was the referral program. Give $10, get $10 (or more). The math and psychology were simple: you got a free ride, your friend got a free ride, and Uber got a new user who would likely become a repeat rider. Everyone won.

Referral programs were the growth playbook of the 2010s. Uber, Dropbox, Airbnb, PayPal. The core insight was that marketplaces have a cold start problem that referrals solve beautifully. You need riders to attract drivers and drivers to attract riders, and the whole thing doesn’t work until you hit a minimum threshold of both sides. And referrals were the spark that got the flywheel going.

So what does this look like in 2026 with 2026’s hottest products? Why doesn’t ChatGPT have a referral program? Why doesn’t Claude? These are some of the fastest-growing products in history and none of them are using the mechanic that defined the last decade of consumer growth. I’ve been sitting with this question and I think the answer has three layers.

There’s no financial incentive

The simplest answer is the structural one. Uber was a two-sided marketplace, and your experience as a rider got tangibly better when more drivers joined. Shorter wait times, lower prices, more coverage. You could feel it in every ride.

ChatGPT is a single-player product. My experience doesn’t get better because you signed up. There’s no second side of the marketplace that benefits from more users joining. The model might improve over time with more data, but that improvement is invisible to me as a user. I’m certainly not sitting there thinking “thank god more people are using this, my answers are getting better.”

So the fundamental incentive structure that made referral programs work in the 2010s just doesn’t exist here. There’s nobody to pay, nothing to split, and no reason for a growth team to build a referral loop around a product that doesn’t get better with more people on it.

There’s no natural trigger

With Uber, the shareable moment was built into the experience. You’d get out of a car and your friend would say “wait, how did you just do that?” The product was visible. It happened in real life. People saw you using it (similar to Waymo today?)

These products work quietly. The output usually lives in your Whatsapp convo, your email draft, or your private conversation. There’s no episode to recommend, no meal to photograph, no moment where someone next to you says “what is that?” The product does its best work in private, and private things don’t spread.

Now, that’s not universally true. People are genuinely raving about things like Claude Code and Codex in a way that does feel like “oh my god you have to try this.” So shareable moments do exist for some AI products. The question is why those moments happen for some experiences and not others, and how you create more of them intentionally.

The learning curve is steep

This might be the most underrated reason of all. With Uber, the gap between “you should try this” and getting value was essentially zero. Download the app, tap a button, get a ride. The referral was the whole onboarding.

With AI, that gap is enormous. Someone has to learn how to prompt well, figure out what the tool is actually good at, build it into their existing workflow, and develop enough intuition to know when to use it and when not to. That takes time and it takes patience and most people don’t get there on their own.

I saw this firsthand recently. I asked my friend to use an AI note taker at work. She was so grateful afterwards and she uses it regularly now. But the referral wasn’t “try this.” The referral was me trying to be a better friend by reducing her workload. That’s friendship, and it’s completely unscalable lol.

Which is the core problem. AI referral, when it works, requires a guide. And right now that guide is almost always a patient friend or family member, not a feature inside the product.

So what does the AI growth playbook actually look like?

If there’s no financial incentive, no natural trigger, and a steep learning curve, how do these products grow?

Netflix didn’t grow through referral codes. It grew by becoming part of culture. You watched a show, you talked about it at dinner, you posted about it, you told your group chat about it. The product created moments that were so shareable that people naturally pulled others in. There was no incentive structure and there didn’t need to be. There was just an experience worth talking about.

AI products haven’t fully cracked this yet, and I think it’s because the output is often invisible and the path to value is long. A Netflix show gives you something to point at, to quote, to screenshot, to argue about over drinks. The productivity gains from an AI tool stay buried in your workflow, and that makes them really hard to spread culturally.

What could this actually look like as a product?

A few ideas if I was talking to a PM building these products:

The gift prompt. Set up a personalized AI workflow for someone you love and send it to them. “I made this for you” hits differently than “here’s $10 off.”

The screenshot moment. AI generates something so visually surprising or personally accurate that people can’t help sharing it. The product builds in its own distribution because the output is so shareable.

The collaborative session. Two people, one AI conversation. Planning a trip together, building a brief together. ChatGPT has a version of this already, and the instinct is right even if the habit hasn’t stuck yet.

The “here’s how I did it” replay. You built something cool with AI. Instead of just sharing the output, one tap lets you share the whole conversation as a walkthrough. Your friend sees the meal plan and can tap “see how she made this” and follow your exact process. The learning curve disappears because someone already climbed it for you.

The before and after. Show me what my first draft looked like vs. what AI helped me turn it into. That transformation is inherently shareable (same psychology that makes home renovation content addictive).

The goal is to turn product into something people want to share because of how it makes them feel.

The AI products that figure out how to create visible, shareable, cultural moments are the ones that will win the next phase of growth. The path forward is making the experience of using AI something people genuinely want to talk about the way they talk about a great show, a recipe that changed their weeknight dinners, or a book they couldn’t put down.

While the playbook can’t be “give $10, get $10”, it is “create something so good that people can’t help telling someone about it.” Which, when you think about it, is the oldest form of marketing there is!

Until next time,
Shrikala