← papernorth.ca

Why we built PaperNorth

And why "just ask ChatGPT" or "just Google it" fails exactly when the stakes are highest.

Every year, millions of Canadians receive a piece of paper that threatens their home, their money, or their peace of mind — an eviction notice, a CRA reassessment, a collections letter. Most can't afford the $150–400/hour a professional costs. So they turn to Google or a chatbot. Here's what the research says happens next.

The evidence against "just ask AI"

58–88%hallucination rate of general AI chatbots on specific legal questions (Stanford research on LLMs and legal queries)
75%+error rate when general models are asked about what a court actually ruled
5 minhow long it took the LTB to dismiss a real Ontario application filed on ChatGPT's advice — after months of waiting

The deeper problem isn't just wrong answers — it's confidently wrong answers. Research shows these models rarely flag their own uncertainty, and when a user starts from a mistaken assumption, the AI tends to build a plausible-sounding argument on top of the mistake instead of correcting it. In a legal situation, that's how people miss real deadlines while feeling fully informed.

The five failure modes — and how PaperNorth was built against each

1. Generic AI never sees your document

Ask ChatGPT about "an N4" and it answers about N4s in general. But cases are won on the specifics: is the termination date 14 days out? Is there a rent-period table? Are there illegal charges mixed in?

✓ PaperNorth reads your actual letter — every page — and checks it against a defect list built from real tribunal practice.

2. Wrong jurisdiction, wrong answer

Google serves whatever ranks — often US law, BC rules for an Ontario problem, or landlord-oriented advice. AI chatbots blend jurisdictions into confident mush. Tenancy and tax rules are ruthlessly local.

✓ Our rules engine is Canadian and province-specific, verified against Tribunals Ontario, the RTA, and canada.ca — and the AI is instructed to use those verified facts, not its general memory.

3. Outdated law

Chatbots have training cutoffs; old blog posts never update. Example: Ontario's N12 rules changed in 2026 — compensation is now waived for 120+ day notices, and from September 21, 2026 a landlord who doesn't move in within 60 days is presumed to have acted in bad faith. Generic AI doesn't know this. Most websites still state the old rules.

✓ Our knowledge base is a living dataset — when the law changes, we update one file and every guide page AND the decoder's brain update together. The 2026 N12 changes are already in.

4. Answers without action

Even a correct Google answer leaves you with homework: find the form, calculate the deadline, write the letter, remember the date. That gap between knowing and doing is where most people lose.

✓ PaperNorth closes the loop: deadlines into your calendar with reminders, a response letter drafted in 30 seconds, an evidence bundle builder, and a direct link to the exact guide for your letter.

5. English-only help in a multilingual country

The people most likely to be targeted by bad-faith notices and scams — newcomers, seniors — are the least served by English-only legal information.

✓ The decoder explains your letter in 12 languages, from Punjabi to Tagalog to Ukrainian, keeping the official form numbers intact.

What we are — and what we're not

PaperNorth is a self-help tool: legal information, document organization, and preparation — the things you could do yourself with unlimited time. We are not a law firm and don't replace one. That honesty is a feature: when your case genuinely needs a human, we say so and point you to the free options — community legal clinics, Tenant Duty Counsel, and volunteer tax clinics — instead of pretending an AI is enough.

The promise

Canada sends you paperwork. We send it back — filled out, on time, in your favour. Free to decode, free to understand, free to act. Because knowing your rights shouldn't depend on affording a professional.

See the difference in 20 seconds

Snap any official letter — get the plain-language truth, your deadlines, and your next move.

Decode a letter free

Research references: Stanford RegLab/HAI studies on hallucination rates of large language models in legal settings (general-purpose LLMs 58–88% on specific legal queries; ~75%+ on case holdings; specialized legal tools 17–43%); Landlord's Self Help Centre reporting on an Ontario LTB application based on chatbot advice dismissed at hearing. PaperNorth provides legal information, not legal advice.