Copy, Paste, Rug: Tracing the Ghost Writer Behind America's Boilerplate Token Whitepapers
Photo by Photo by Vitaly Gariev on Unsplash on Unsplash
The document looked impressive. Fourteen pages, a tokenomics pie chart, a roadmap stretching to 2027, and a section on 'novel consensus mechanisms' that name-dropped three academic papers nobody on the team had apparently read. It was the whitepaper for a new AI-minting protocol that launched out of a Delaware LLC last spring.
It was also, nearly word for word, the whitepaper for at least forty other tokens.
That's the discovery that turned Marcus Idowu — a former fraud analyst turned independent on-chain investigator — into something of a document detective. His specialty: automated whitepaper generators, the one-click tools that let anyone with a laptop and a Solana wallet produce a polished, plausible-sounding token thesis before their coffee gets cold.
The Assembly Line Nobody Talks About
Whitepaper generators aren't new. What's new is how good they've gotten, and how cheap. A paid tier on several popular services will produce a full document — executive summary, technical architecture, vesting schedule, even a fabricated 'advisory board' — in minutes. The output reads like a real whitepaper because it's trained on real whitepapers. That's also the problem.
'When you train on a corpus of token documents, you inherit their clichés,' Idowu explains. 'Every generated paper has the same skeleton. The same phrases show up again and again, like fingerprints.'
Those fingerprints are exactly what he hunts. By running token documents through similarity-detection software originally built for academic plagiarism, he's mapped clusters of launches that share long, verbatim passages — whole paragraphs on 'community-driven value accrual' or 'frictionless yield optimization' copied across projects that have never acknowledged each other's existence.
Some of these clusters are genuinely funny. One phrase — a specific, oddly poetic sentence about 'the marriage of neural architectures and mint economies' — appears in over two hundred token documents he's catalogued. Two hundred projects, one sentence, zero attribution.
Why This Matters More Than a Bad Grammar Day
It'd be easy to shrug this off. A whitepaper is marketing, right? Everyone knows the docs are aspirational.
Here's why it's not that simple. In the US, a whitepaper isn't just a brochure — it's potentially a securities offering document. The SEC has repeatedly signaled that statements in token materials matter, and projects that lean on 'AI-driven' claims without substance are squarely in the crosshairs of an enforcement environment that's gotten considerably less forgiving about tech-washing.
For retail investors, the stakes are blunter. Idowu's clusters correlate strongly with bad outcomes. 'In my dataset, documents sharing boilerplate above a certain similarity threshold have a rug rate that's honestly depressing,' he says. 'The lazy document is a symptom. If the team couldn't be bothered to write — or even meaningfully edit — their own thesis, what exactly did they build?'
There's a deeper irony here for a site like ours. The whole pitch of AI-meets-crypto is that machines can do what humans can't. But when a machine writes the whitepaper, a machine designs the tokenomics, and a machine mints the supply, there's no human judgment left anywhere in the pipeline — just a smooth chute from prompt to pump.
Inside the Reverse-Engineering Process
Idowu walked us through his workflow, and it's more accessible than you'd think.
First, he pulls the whitepaper text — stripping PDFs down to plain prose. Then he runs n-gram analysis, looking for matching sequences of eight or more words across his database of thousands of documents. Common crypto jargon gets filtered out; 'decentralized' matching isn't interesting. What's interesting is when two projects share a fifteen-word sentence about their 'modular liquidity flywheel.'
Next, he traces the wallets. If two projects share language, do they also share deployer addresses, funding wallets, or liquidity providers? Frequently, they do. One cluster he flagged last fall turned out to be the same four wallets relaunching a 'new' protocol every six weeks under fresh names — each with a freshly generated, freshly plagiarized whitepaper.
Finally, he checks the citations. Generated documents love to reference papers that either don't exist or have nothing to do with the claims. 'If the bibliography is hallucinated, the rest of it probably is too,' he says flatly.
A Five-Minute Checklist Before You Buy
You don't need Idowu's tooling to protect yourself. Here's the gut-check list he recommends to US retail investors:
- Google the weird sentences. Pick the most specific, unusual phrase in the whitepaper and search it in quotes. If it shows up in other token docs, you've found the template.
- Check the bibliography. Do the cited papers exist? Have the named authors actually published on this topic? A hallucinated citation list is a five-alarm fire.
- Look up the entity. A Delaware or Wyoming LLC registration is trivially easy to file and tells you almost nothing — but the absence of any legal entity at all tells you even less in your favor.
- Compare the roadmap to the wallet activity. Claims about 'audits underway' or 'partnerships pending' should be verifiable. If nothing on the roadmap has a receipt, the roadmap is decoration.
- Ask who edited this. A genuinely AI-assisted whitepaper written by a real team will still have a voice — specific numbers, named engineers, opinions. Generated boilerplate hedges everything and commits to nothing.
The Uncomfortable Middle Ground
To be clear: not every AI-assisted document is a scam, and not every scam uses a generator. Plenty of legitimate small teams use these tools to draft, then edit heavily. Idowu is careful on this point. 'The tool isn't the crime,' he says. 'The crime is when the tool is the whole company.'
The real tell isn't AI involvement — it's the absence of a human willing to stand behind the words. In a market where anyone can mint a document as easily as a token, the scarcest asset isn't compute or capital. It's accountability.
So the next time a 'revolutionary' token thesis lands in your feed, do what the investigator does: read it twice, search it once, and ask the only question that matters. If this thing went to zero tomorrow, who signed their name to it?
If the answer is a prompt, you have yours.