04/14/2026
London, Ontario Startup Builds Blockchain Shield Against AI Misinformation
A Canadian tech company is using distributed ledger technology to stop AI search engines from spreading false information about businesses
By The Pirate Monster Crew | Technology Section
LONDON, ONT. — When a Canadian software firm discovered that three major AI search engines were confidently telling the public it was headquartered in New York City, the company's founders did what most business owners do: they submitted correction requests and waited.
Weeks passed. One engine fixed it. The other two didn't. And by the time the second correction went through, a fourth engine had picked up the bad information from the others.
"That's the moment it became clear this wasn't a customer service problem," said the team at Pirate Monster, a London, Ontario-based AI optimization platform that has since built what may be the first blockchain-backed identity verification system designed specifically for the age of AI-generated search results.
The problem they've identified — and claim to have solved — strikes at a vulnerability in the architecture of modern AI search that few outside the industry have been talking about.
The Echo Chamber Problem
AI-powered search engines like ChatGPT, Google's Gemini, Perplexity, and Meta AI don't operate in isolation. Their training data and real-time retrieval systems routinely index content from across the web — including, in some cases, answers generated by other AI systems.
The result is a phenomenon researchers have begun calling "model collapse" or "data contamination," but which Pirate Monster describes in more specific terms: AI-to-AI citation chains.
The mechanics are straightforward. One AI engine generates an answer containing an error — a wrong address, a misattributed founder, an incorrect service description. A second engine, crawling the web for information, encounters that answer and incorporates it into its own knowledge. A third engine does the same. Within days, the original error has been laundered through multiple systems, each one lending it additional credibility by repetition.
"It's the telephone game, except the players have perfect memory and no skepticism," the Pirate Monster team said in a technical briefing reviewed by this newspaper. "Once a hallucinated fact enters the loop, it becomes consensus."
For the businesses caught in the middle, the stakes are not trivial. Incorrect information in AI search results can affect customer trust, local search visibility, vendor relationships, and in regulated industries, compliance standing.
The Blockchain Solution
Pirate Monster's response is a system it calls the Truth Layer — a five-part architecture built on top of the Hedera Hashgraph network, a distributed ledger governed by a council that includes Google, IBM, Boeing, and several other multinationals.
The system works as follows:
A business first verifies ownership of its domain through a standard DNS or file-based proof. It then submits what the company calls an Entity Truth Record — a structured declaration of its legal name, jurisdiction, service area, related web properties, and notably, a list of things the business explicitly is not. That declaration is cryptographically hashed, timestamped, and permanently recorded on Hedera's public ledger.
The record cannot be edited or deleted. When a business needs to update its information, a new version is appended that references the previous one by transaction ID, creating an unbroken chain of identity updates that anyone — human or machine — can independently verify.
"The key design decision was that corrections don't replace old records," the team explained. "They supersede them. The full history is always visible."
The system also provides a structured dispute mechanism. When an AI engine publishes incorrect information about a business, the dispute filing, the evidence (stored as a cryptographic hash, never raw content), the correction submission, and the eventual resolution are all recorded on the ledger. If the same engine repeats the same error, there is now permanent, timestamped, tamper-proof evidence that a correction was previously issued.
Detecting the Contamination
Perhaps the most technically ambitious component of the system is what the company calls its Synthetic Firewall.
The platform's Citation Probe Service actively scans responses from AI search engines and classifies every cited source into one of three categories: human-originated, AI-summarized, or synthetic — meaning one AI engine citing another AI engine as its source.
When synthetic citations are detected, the system tracks the depth of the chain (how many AI-to-AI hops have occurred) and the spread (how many engines are repeating the information). Contamination events that exceed defined thresholds — a chain depth of three or more, or spread across four or more engines — trigger automated alerts and are permanently recorded on the Hedera ledger.
According to internal data shared by the company, a synthetic citation propagates to an average of 3.2 additional AI engines within 72 hours of its initial appearance.
"The contamination is fast," the team said. "A single hallucinated fact can reach majority consensus across AI platforms in under a week. Our system catches it at the first hop."
What the System Won't Record
The company emphasized that the integrity of its ledger depends as much on what it excludes as what it includes.
An internal policy document obtained by this newspaper lists five categories of data that are permanently blocked from the chain: raw AI-generated citations, synthetic citation content, unverified negative claims about businesses, content that has not been explicitly reviewed and attested by the business owner, and any bulk or automated submissions.
"If you let everything onto the chain, the chain becomes meaningless," the team said. "The discipline is in the curation."
Five categories are permitted: verified entity declarations, versioned profile updates, hashed dispute filings, hashed legal attestations from recognized sources, and documented contamination events. Each permitted category has a defined data classification — public, confidential, or restricted — enforced at the code level.
Industry Context
The problem Pirate Monster is addressing sits at the intersection of two larger trends in technology: the rapid adoption of AI-generated search results by consumers and businesses, and the growing body of academic research documenting the risks of AI models training on AI-generated content.
A 2025 paper from researchers at the University of Oxford and the University of Cambridge warned that "model collapse" — the degradation of AI model quality when trained on synthetic data — could become a significant challenge as AI-generated content becomes an ever-larger share of the internet.
Pirate Monster's approach differs from the academic framing in that it focuses not on the degradation of AI models themselves, but on the downstream impact on real businesses whose information is being distorted by the contamination cycle.
The company currently operates under Canadian law, specifically the Province of Ontario, and says it has designed its system to comply with both PIPEDA (Canada's federal privacy law) and GDPR (the European Union's data protection regulation). Entity truth records are classified as public data. Dispute filings and legal attestations are classified as confidential. Personal identifying information is subject to a 90-day retention limit.
The Business Model
Pirate Monster offers its platform on a tiered subscription basis, ranging from a free tier with limited AI analysis credits to enterprise plans with unlimited usage. The Truth Layer is integrated into the broader platform, which includes AI search engine scoring, content optimization, social media management, and website monitoring tools.
The company has indicated it is exploring the introduction of a native utility token — described as a credit system for platform actions — though no timeline for that feature has been publicly announced.
The Hedera network was chosen, the team said, for its low transaction costs (fractions of a cent per record), fast finality (three to five seconds), and the governance credibility of its council members — factors they considered more important than the decentralization characteristics prioritized by other blockchain networks.
What's Next
The company said it is working toward making its verified entity records portable — a credential that businesses could present to any platform, partner, or auditor for independent verification.
"The end state is that your business identity isn't something AI engines guess at," the team said. "It's something you declare, verify, and prove. And the proof doesn't live on our servers. It lives on a public ledger that we don't control."
Whether that vision scales beyond the platform's current user base — and whether AI search providers will choose to check these records — remains an open question. But as AI-generated search results become the primary way millions of people find information about businesses, the question of who controls the truth is no longer theoretical.
It's infrastructure.
Pirate Monster is a privately held technology company based in London, Ontario. The Hedera Hashgraph network is a public distributed ledger. This newspaper has no financial relationship with either entity.
Contact: [email protected]
