The product-led growth playbook for consumer AI apps

product-led growth for a consumer AI app is mostly not the invite-a-friend tab. it is building the product so that using it spreads it: outputs people want to show other people, onboarding that reaches that output fast, and referral incentives added only after retention proves people stay. this playbook covers that stack in order, and why it beats bolting a referral program onto a leaky app.

What PLG means when your product makes things

the term comes from b2b saas, where it means the product sells itself through usage. for a consumer AI app the translation is more literal: your product makes things, and things travel. every image, plan, song, score or answer your app produces is a potential ad that a real person chose to show another real person, and no media budget buys that kind of credibility. that is why a loop built into the product itself usually outperforms the identical loop bolted on next to the settings tab, where it sits waiting for a motivation that never arrives.

The strongest loop is the thing your app makes

the strongest viral loop in consumer AI is the shareable output, not the referral link. 2025 proved it at a scale nobody plans for: when GPT-4o image generation landed, a16z’s State of Consumer AI 2025 report put the peak at roughly a million new users an hour, and Nano Banana pulled 10M new users in a single week off shareable images, per the same report. nobody was forwarding invite codes during either of those runs. people posted what the product made, and the outputs carried the product to everyone who saw them.

you will not replicate that scale, and you do not need to. the mechanism is the same at any size: if some fraction of your users shows an output to their feed or their group chat every week, you have a compounding channel that costs nothing per impression and arrives with a friend’s implicit endorsement attached. the craft is making the output worth showing, which is product work before it is marketing work.

The hierarchy: retention first, shareable output second, referral last

the order is fixed, and most founders run it backwards. retention comes first because a loop attached to an app people abandon multiplies nothing; new users arrive, churn, and the loop starves. the shareable output comes second because it is the loop users run for their own reasons, status, usefulness or delight, with no bribe required. the incentivized referral program comes last because it is the weakest motive of the three. it can amplify love that already exists. it cannot create it, and shipped too early it mostly measures how little love there is.

How I actually run this

as a supporting mechanic inside a mix, never as the whole plan. at ZuAI the share loops supported the machine while paid ads and UGC did the volume; the loop made every acquired user worth a little more, and that is the honest job description for most apps. the tradeoffs are real too: loops are slow to design, impossible to force, and easy to fake with a share button nobody presses. they only compound users who stay, which is why a good chunk of this playbook is about retention and onboarding rather than anything with the word viral in it.

if your app’s output is private by nature, journaling, therapy, finance, the public loop is mostly closed to you, and pretending otherwise wastes a quarter. you still get the rest of the stack: fast activation, a clean web-to-app funnel, referral timing done right, and incentives paid in product instead of cash. plg is a stack, not a single trick, and the pages in this playbook take the layers one at a time.

want the loop mapped against your actual product? the breakdown call is free. Let’s talk Growth.