Business strategy

Matt Espinoza on Cloning Videos and Seeding AI Search

Clover’s founder describes video variants, phone networks and planted Reddit comments, with limits on his claims about reach and AI citations.

Listen to the audio analysis

Create your own

Written for builders and creators.

App growth isn't about going viral by luck anymore. It's actually about a high-velocity, engineered cycle.

Editor's note

Why this matters now

Matt Espinoza runs Clover, a company that distributes app marketing content. In his Superwall Podcast interview, he says about 90% of its campaigns consist of variations on videos that have already worked. The company also places Reddit posts and comments in an attempt to influence sources cited by AI search.

The campaign therefore involves more than generating videos. Espinoza describes physical phone networks, accounts made to appear local and promotional comments placed in existing discussions. Those distribution tactics are essential context for a growth operator assessing his reported results. Superwall, which makes the podcast, is a Clover customer.

The source

What it says

Distilled from the original. The notes above and below are the editor's own.

What Clover Calls a Clone

Espinoza describes keeping a successful video's structure while changing an element such as the opening hook, the product demonstration, the on-screen character or a visual detail. A working hook can be paired with several demos, and a working demo can be tried with several hooks.

He says the median clone attracts about 20% of the original video's views, with some variants reaching 2 to 6 times the original. These are his reported results, without a sample size, measurement window or view-to-sale comparison.

The distinction between a median and a total matters. Producing 10 variants does not mean each will get the median result, and adding their views does not establish unique audience reach. The account offers a reason to test variations, rather than a dependable multiplication rule.

Espinoza also describes changing small visual details to avoid duplicate-content detection. That is part of his stated tactic. The interview provides no platform evidence establishing which changes avoid detection or for how long.

Localisation Still Needs an Audience

Clover uses subtitles, dubbing and altered mouth movements to adapt videos for different languages. Espinoza describes launching accounts in about 20 countries and reports gains from taking successful content into additional markets.

His examples do not establish a standard percentage lift per country. A translation changes who can understand a video. It does not by itself establish demand for the app, appropriate pricing or the cost of supporting customers in that market.

The practical attraction is reuse: the team can try an existing concept in another language without filming an entirely new scene. The commercial result remains a separate question.

The Phone Network Is Part of the Method

Espinoza describes a network of physical phones used to create and run accounts that appear to be local users in different countries. Bios and activity are constructed to suit the market.

That detail changes the meaning of the distribution claim. The campaign relies on account infrastructure and deliberate presentation of those accounts, as well as on generating video variants. It is not evidence that posting translated content from an ordinary brand account would produce the same reach.

The interview does not provide the full operating cost, account-loss rate or a comparison with other distribution methods. It therefore cannot establish the profitability or durability of the phone network.

Planted Comments as a Route Into AI Answers

Espinoza describes turning common searches into chatbot prompts, examining the pages cited in the answers, then placing Reddit posts or comments that promote a product. He says an answer may cite an individual comment within a thread.

The intended route is clear: place a favourable statement where an AI search system has previously found material for a relevant question. The statement is planted promotion, rather than an independent customer recommendation.

A previously cited thread does not guarantee that a new comment will be retrieved, selected or repeated. The interview supplies no controlled before-and-after test establishing that Clover can determine an answer.

Daniel Kwon's stitched-video account concerns another kind of reuse: combining existing clips to construct a hook. Clover's account extends from variants into how and where the resulting promotion is distributed.

Editor's note

What to do with this

Clover's pitch joins production with placement. Clones reduce the amount of new material required for each attempt. Local-looking accounts and seeded comments try to put that material where viewers or search systems may encounter it.

Those mechanisms can be described without accepting the stronger claim that distribution is now predictable. Espinoza's reported view ranges and citation examples leave conversion, operating cost and sustained access unresolved. More variants and more placements are additional attempts, not proof of a controlled outcome.

Explore the Superwall Podcast creator marketing playlist.

The original

The NEW App Content Strategy Dominating in 2026

The Superwall Podcast · 13 September 2026

All posts