Human-reviewed. AI-assisted. Never raw AI.
Most finance sites now run on AI and won't tell you. We will. This page shows the exact process behind every WillStreet article, where the machine helps, and where a human who spent six years inside Canadian banking draws a hard line.
AI drafts and researches
It structures first drafts, cross-checks Canadian data, and surfaces sources faster than a human working nights and weekends could alone.
A human edits every line
Nothing publishes on autopilot. Every figure is recomputed by hand, every AI tell is stripped, every claim is checked against a primary Canadian source.
The judgment is human
The verdict in a review, the insider section, the call on what's worth your money. That's written by a person and answerable to a person.
Watch the tells
get stripped out.
Raw AI writes in a recognizable dialect: em-dashes, "here's the thing," padded triplets, hollow reframes. Below is a real paragraph as the model produced it. Run the review and watch what a human takes out before it reaches you.
This is not a metaphor. The banned-pattern list you just watched get applied is the same pre-publish checklist every WillStreet draft runs through. Dash count has to hit two or fewer across a full article. If a sentence sounds like something a finfluencer would post, it gets rewritten.
What AI does. What it never touches.
- Build the first-draft structure of an article from an approved outline
- Cross-check Canadian figures: CRA limits, contribution room, FX rates
- Surface primary sources and flag where a claim needs one
- Generate headline and social variants to test against each other
- Recompute payoff and tax math as a second independent check
- Write "How the System Actually Works", the insider section, a human writes it or it stays blank
- Decide a WillStreet Score or the verdict on whether a tool earns your money
- Publish a single figure that a human hasn't recomputed by hand first
- Present US financial content dressed up with Canadian labels
- Have the final word on anything customer-facing. Ever.
Every tool, and the leash it's on.
No mystery black box. This is exactly which model does what, and what a human does after each one.
The $0 rule: the stack runs on free tiers by design. One paid exception. AI-generated financial math is always independently verified before it's published, because a wrong contribution limit isn't a typo, it's a reader making a decision on bad information.
One wrong number is one wrong decision.
A concrete example from an actual review: one model reconciled 32 transactions to the dollar. Another was off by $192.55 on the same set. That gap is invisible until you're modeling a TFSA contribution against it. This is the entire reason a human recomputes the math.
AI writes fast.
A human is still accountable.
That's the whole policy. Speed from the machine, judgment from a person, and full disclosure to you. If a WillStreet article ever reads like raw AI, it failed its own checklist, and you should hold it to that.
WillStreet is an independent Canadian publisher. Content is for informational purposes only and is not professional financial, tax, or legal advice. Consult a Canadian CPA or licensed advisor before acting on anything here.
WillStreet Scores are never influenced by affiliate relationships. Some links may be affiliate links. Claude and ChatGPT have no public affiliate program, so WillStreet earns no commission from either. Written with operational experience in Canadian banking. No confidential information. No employer named.
All pricing is shown in CAD where it appears and verified on the publish date. AI tool features and pricing change quickly, so figures reflect the time of writing. Verify current figures against primary sources before deciding.