Can AI really clone any app?

AI can rebuild many features of simple apps quickly, and tools like Reverse Engineer Anything (REA) help coding agents study how an existing app works. But a working copy of the features is not a copy of the business. The customers, reviews, track record and distribution behind a product stay with the original.

What is a software company's moat if AI can copy the code?

The moat is what sits around the code: customers who stay when something breaks, reviews from real users, testimonials from people your buyers know, distribution and a long track record of not letting people down. None of that lives in the app, so none of it can be reverse engineered.

Can AI fake reviews and testimonials instead?

It can generate them, but in the UK fake reviews have been banned under the Digital Markets, Competition and Consumers Act since 6 April 2025, and the CMA can fine businesses up to 10% of global turnover. As fake reviews get easier to make, buyers trust them less, which makes genuine reviews more valuable.

AI can reverse engineer any app. Trust is the moat it can't copy

AI can now take apart an app and rebuild it.

Good. Copying software was never the hard part.

What AI can copy vs what it can’t - infographic

This month a GitHub repo called Reverse Engineer Anything (REA) went viral. It plugs your coding agent into decompilers like Ghidra and IDA, works out how an existing app does something, and helps you build your own version. It picked up hundreds of thousands of stars in about a week. My feed filled with “I cloned a $9M app” posts and one-prompt lookalikes of Calendly, Typeform and Trello.

The conclusion most people drew: software is dead, anyone can copy anything, moats are gone.

The conclusion I drew: nothing important changed.

The short answer

If you’re worried a competitor will clone your product with AI, they probably can. They could before, too. What they can’t clone is everything around the product: the customers who stay, the reviews written by real people, the reputation that makes a stranger trust you with their money. That has always been the hard part. AI makes it more valuable, not less.

Was copying software ever actually hard?

No.

Every successful software product has had copycats for as long as there has been software. A decent team could always rebuild the feature list of a simple tool in a few months. The internet is full of “Uber for X” clones nobody remembers.

Even the people testing REA make this point. One YouTuber raced an REA-powered rebuild of the Mac Preview app against plain Claude with no special tools, and said the base model gets most of the way on its own. And the viral Calendly and Trello lookalikes weren’t built with REA at all. They came from a screenshot and a prompt in an app builder, and none of them is a full product.

So copying got cheaper. It was already cheap.

If the code is free, what are customers paying for?

They’re paying to not worry.

When an event organiser picks a ticketing platform, they don’t compare feature lists line by line. They ask simpler questions. Will the payments work on the night? Will someone answer when the door scanner breaks at 7pm? Has anyone like me used this and been fine?

I build Presso Events, mostly with AI agents. I could rebuild the feature list of a big ticketing platform. What I can’t rebuild is the organisers who have trusted it with their ticket money for years. That gap is the business. The code is the smallest part of it.

Before I built my own products, I spent years in enterprise customer success. Accounts didn’t stay because of a feature. They stayed because they trusted the people behind the product when something went wrong.

What can’t AI reverse engineer?

You can decompile an app. You can’t decompile a relationship. Here’s what stays with the original:

Customers who stay when it breaks. Loyalty is earned in the bad weeks, not on launch day.
Reviews from real people. Written by someone who paid, used it and came back.
Testimonials from someone your buyer knows. A name they recognise beats any landing page.
Distribution. The audience, the email list, the partners who send you work.
Track record. Years of not losing anyone’s money or data.

None of these live in the binary. That’s why no tool can pull them out of it.

Doesn’t AI fake reviews and testimonials too?

It can. That’s the twist.

When anyone can generate a glowing review in seconds, people trust reviews less. And the genuine ones become worth more.

Regulators have caught up too. In the UK, fake reviews have been banned under the Digital Markets, Competition and Consumers Act since 6 April 2025, and the CMA can now fine businesses up to 10% of global turnover. The stakes are big: the CMA has estimated that £23 billion a year of UK consumer spending is influenced by online reviews, and a 2023 government study found up to 15% of reviews in three common product categories were fake.

So the shortcut to trust is now illegal and easier to spot. The slow route is the only one left.

So what should founders do?

Last year I wrote that AI didn’t democratise product, it drowned it. REA is the same flood, one level up. More copies, same scarcity.

Here’s how I’d play it:

Stop treating code as the moat. Ship it, and assume it will be copied.
Spend the time AI saved you on customers. Calls, onboarding, support. That’s where the gap opens up.
Collect proof properly. Ask for reviews after someone got real value, from real users, with real names.
Show up for years. A clone can copy your product. It can’t copy three years of you being there.

The bottom line

REA is a clever tool and the hype is understandable. But if your plan is “we’ll just build it”, you’re competing on the cheapest thing in the stack.

Code is a commodity. Trust is the moat.

AI can clone your app. It can't clone your customers.