Four AI Music Sites, One Codebase: How to Spot a Review Farm
One of our articles recommended four AI music generators as four independent choices. They are not. They share 23 byte-identical JavaScript bundles, the same FAQ sentence reproduced typo-for-typo, the same unrendered template placeholder in their terms, and four differently-named shell companies on one arbitration clause. They are branded instances of a single product. Comparing them against each other tells a reader nothing, and treating them as four suppliers to fall back on is a mistake.
This post is how we found it and how you can run the same check in a browser, in about two minutes, on any set of suspiciously similar sites.
Key Takeaways
- 23 JavaScript bundle filenames were byte-identical across all four sites. Modern web frameworks derive those filenames from a hash of the file's contents, so identical filenames mean identical compiled code (checked 2026-09-03, ANALYSIS).
- The same FAQ sentence appears on all four, with the same typo, which is not something independent teams produce.
- All four leak the same unrendered template variable in their upload terms — a placeholder that was meant to be replaced by the brand name and never was.
- Four different company names share one arbitration clause, with only the entity name swapped between them.
- Separate analytics IDs per brand. This is a white-label operation running several storefronts, not one site with four domains pointed at it.
What we originally published
Our article named four AI music generators, gave each its own section, and presented them as separate options with different strengths. That is the normal shape of a round-up and it is what a reader expects: four choices, compare and pick.
When we re-checked the article we started with the ordinary questions — do these products exist, are the prices right, are the licensing claims accurate. They all existed. But the sites looked alike in a way that was hard to ignore.
The evidence
Identical code
Modern JavaScript frameworks split an application into bundles and name each file using a hash of its contents. Two files with the same hashed filename contain the same bytes. That is the point of the naming scheme — it lets a browser cache aggressively and know when something has genuinely changed.
We extracted the bundle filenames from each of the four sites and intersected the lists. Twenty-three filenames appear on all four, hash and all. Independent products built by independent teams do not produce twenty-three identical compiled files.
An identical typo
Every one of the four carries the same FAQ sentence about commercial licensing, with the same missing word, in the same place. Shared marketing copy between two brands is common. A shared error is a copy.
An unrendered placeholder
All four sites' upload terms contain a template variable that was never filled in — the spot where the brand name was supposed to be substituted still shows the placeholder. The same bug, in the same sentence, on four sites.
That single detail is the most telling thing we found. It is direct evidence of a template: one document, meant to be personalised per brand, deployed four times with the substitution step broken.
Four companies, one arbitration clause
The legal entities differ by name. The arbitration sentence they sit in is otherwise identical, word for word, with only the company name swapped. Four shell names on one legal template.
Separate tracking
Each brand has its own analytics container. That is what distinguishes this from a single site with four domains pointed at it: someone is measuring these storefronts separately, which is what you would do if you were running them as distinct brands on purpose.
Why this matters to a reader
A comparison between them is meaningless. If three of the four are the same product wearing different names, ranking them against each other is theatre. Any difference you find is branding.
It is not four suppliers. The practical reason to compare vendors is redundancy — if one raises prices, disappears or changes its licence, you move. Four instances of one operation give you none of that. If the operator stops, all four stop.
Licensing risk concentrates in the same way. Whatever the terms permit or forbid, you are agreeing to one set of terms four times, from one entity structure, not diversifying across four independent legal positions.
None of this means the product is bad. It may be perfectly good. It means the article's framing — four choices, pick one — was wrong, and we have corrected it.
How to run this check yourself
Two minutes, in a browser, no tooling required.
- Open two of the sites and view the page source (
Ctrl+UorCmd+Option+U). - Search the source for
/_next/static/chunks/or/static/js/— the paths where bundled JavaScript lives. Copy a handful of the filenames, which look likemain-app-e5ad82c522e9e590.js. - Compare them across the sites. A few shared filenames can happen when two sites use the same framework version. Twenty-plus identical hashes cannot.
- Read the terms and privacy pages side by side. Look for identical sentences, and especially for identical mistakes: a typo, a missing word, an unreplaced placeholder.
- Compare the company names in the arbitration and governing-law clauses. Different names in an otherwise identical paragraph is the signature.
The shortcut: search each site's source for { followed by a word and } — an unrendered template variable. It is a strong signal on its own, and it takes seconds.
What this does not prove
We should be precise about the limits, because it is easy to overread this.
It does not prove deception. Running several brands from one platform is a legitimate business model, and white-label products are everywhere. Nothing we found says the operator claimed the four were independent — our article claimed that, and that was our error.
It also does not tell you who owns them. Shell entity names in a terms page are not a corporate registry. We can say the four share a codebase and a legal template; we cannot say who is behind them, and we have not tried to.
How This Guide Was Sourced
Written and maintained by the LogicBalls editorial team (logicballs.com). Disclosure: LogicBalls builds AI writing tools. The article that presented these four as independent products is ours, and it has been corrected.
Method. The four sites were fetched on 2026-09-03. JavaScript bundle filenames were extracted from each page's source and intersected. Terms, privacy and FAQ pages were compared for shared wording, shared errors and entity names. All findings are from the sites' own published pages.
The sites are not named here. The correction is published on the article that made the error, which is where a reader who acted on it will see it. This post is about the detection method, which works on any set of similar sites and is more useful than a list.
What is checkable and what is not. The bundle-hash and shared-template findings are reproducible by anyone following the steps above, as of 2026-09-03 — sites get rebuilt, and a later check may look different. The inference that these are one operation is ours, marked ANALYSIS, and rests on the evidence described rather than on any disclosure by the operator.
AI involvement. The original article was AI-assisted and published without this check. This post was AI-assisted and reviewed before publication.
No LogicBalls telemetry is used in this guide.
Frequently Asked Questions
Is running multiple brands from one codebase wrong?
No. It is a normal business model, and plenty of reputable products are white-labelled. The problem is a review presenting them as independent alternatives, which is what ours did.
Could the identical files just mean they use the same framework?
Two or three shared filenames, yes — common libraries at the same version can produce that. Twenty-three, including application-specific bundles, cannot. The hash covers the file's contents, so a match means the compiled code is the same.
How would I have spotted this without viewing source?
The FAQ typo and the unrendered placeholder are visible on the page itself, in ordinary reading. So is a very similar layout across sites that claim no relationship. Viewing source makes it conclusive; reading carefully raises the suspicion.
Does this apply outside AI music tools?
The technique applies to any set of sites you suspect are related — directories, review sites, tool listings. Shared hashes and shared template errors are not category-specific.
Conclusion
Four AI music generators, presented in our own article as four independent products, share twenty-three identical code bundles, one FAQ typo, one unreplaced template placeholder and one arbitration clause across four company names.
The check takes two minutes: view source, compare bundle filenames, read the terms side by side, look for an error that repeats. It is worth running any time a set of competing products feels a little too alike — and worth running on round-ups that recommend several of them at once.
