Canadian Gold Miners Cohort — 40-F filings dated 2026-02-18 through 2026-03-31

Lede

No event-level disclosures surfaced in this cycle across the six-name Canadian gold cohort — no bullets extracted from any of the GOLD, AEM, KGC, FNV, PAAS, or WPM 40-Fs. The differentiation is entirely linguistic, and it sorts the cohort along a streamer/operator axis. Franco-Nevada is the hedging concentrate; Wheaton and Pan American sit at opposite ends of the certainty register; Agnico Eagle's MD&A reads operationally rather than narratively; and Barrick's AIF carries the highest boilerplate signature in the group.

Findings

1. FNV clusters on the upper tail across six distinct signals

Franco-Nevada posts above the 95th percentile readings on [Annual Information Form Hedging language = 8.490], [Annual Information Form Passive voice ratio = 1.394], [Annual Information Form Forward-looking framing = 15.198], [Management's Discussion & Analysis Hedging language = 4.388], [Management's Discussion & Analysis Disclaimer density = 0.984], and [Management's Discussion & Analysis Cliché density = 0.109]. No other ticker in the cohort clusters this densely on the hedging tail. The pairing of forward-tilt and tentative phrasing is the signature, and it is concentrated in one name.

2. WPM and PAAS define opposite ends of the certainty spectrum

[Annual Information Form Certainty / hedge imbalance]: WPM 0.510 (above the 95th percentile) against PAAS 0.231 (bottom 10%). The same split repeats in [Management's Discussion & Analysis Certainty / hedge imbalance]: WPM 0.528 (above the 95th percentile) against PAAS 0.230 (bottom 10%). PAAS additionally posts [Annual Information Form Disclaimer density = 0.000 (bottom 10%)] — low certainty without the disclaimer vocabulary that would normally accompany it. That combination is unusual and worth naming explicitly.

3. AEM's MD&A reads operationally, with caveats

Agnico Eagle runs bottom 10% on [Management's Discussion & Analysis Passive voice ratio = 0.734], [Management's Discussion & Analysis Forward-looking framing = 6.907], [Management's Discussion & Analysis Hedging language = 1.601], and [Annual Information Form Hedging language = 2.649] — active voice, present-tense, low hedging. But the same filing posts above the 95th percentile on [Management's Discussion & Analysis Boilerplate callbacks to prior periods = 1.960] and [Annual Information Form Vague attributions = 0.568]. Confident phrasing leaning on prior-period framing and unattributed sourcing.

4. GOLD's AIF carries the cohort's highest boilerplate markers

Barrick posts above the 95th percentile on [Annual Information Form Cliché density = 0.038] and [Annual Information Form Boilerplate callbacks to prior periods = 0.064], alongside above the 95th percentile readings on [Management's Discussion & Analysis Forward-looking framing = 14.367] and [Management's Discussion & Analysis Vague attributions = 0.174]. More recycled phrasing, more forward-tilt, less specific sourcing — concentrated in the AIF.

5. KGC carries heavy disclaimers in active voice

Kinross posts [Annual Information Form Disclaimer density = 5.682 (above the 95th percentile)] against a cohort mean of 2.145, while [Annual Information Form Passive voice ratio = 1.038 (bottom 10%)] runs below the cohort. Disclaimer load is the highest in the cohort, but the voice carrying it is unusually direct. That pairing is uncommon and deserves the annotation.

Forensic note

The absence of bullets is the editorial frame. With no event-level disclosure to anchor, the linguistic dispersion across the cohort is doing all the differentiation work, and it sorts cleanly along business-model lines. The streamers (FNV, WPM) and the operators (GOLD, AEM, KGC, PAAS) inhabit measurably different disclosure registers — FNV's six-signal cluster on the hedging tail and WPM's Certainty / hedge imbalance extremes ([Annual Information Form = 0.510], [Management's Discussion & Analysis = 0.528]) are not noise. Within the operator subset, AEM's combination of low tentative markers and high thermo callback density ([Management's Discussion & Analysis Boilerplate callbacks to prior periods = 1.960]) is the inverse of FNV's profile: confident phrasing built on backward reference rather than forward hedge.

What to watch

Next cycle, the test is whether FNV's hedging cluster persists or compresses — persistence implies structural disclosure practice tied to the streamer model; compression implies something specific to this filing year. PAAS's pairing of bottom 10% Certainty / hedge imbalance with zero AIF disclaimer density is the second item to track: if disclaimer language returns to the next filing, the current reading was a drafting artifact rather than a posture. AEM's [Management's Discussion & Analysis Boilerplate callbacks to prior periods = 1.960] warrants monitoring against any future operational disclosure — heavy reliance on prior-period framing is the kind of pattern that becomes legible only when something new has to be said. And if GOLD's [Annual Information Form Cliché density] and [Annual Information Form Boilerplate callbacks to prior periods] both remain above the 95th percentile next cycle, the AIF is on autopilot.