Caveats — vendor quarantine and the correlational-vs-causal divide

Honest gaps:

1. Correlation vs causation pervades this field. Almost all ranking-factor evidence (links, age, brand) is correlational. The rare causal exceptions: review-impact studies (Luca regression-discontinuity, Spiegel conversion experiments); antitrust-confirmed NavBoost existence.
2. Survivorship/selection bias is rampant. SEO case studies and "we outranked a 2015 competitor in 8 months" stories feature only winners. The Ahrefs 1.74% figure is the closest thing to a base rate and it is sobering.
3. Vendor incentive bias. Large share of all published data in this domain comes from companies selling SEO tools or services. Numbers quarantined accordingly.
4. Ground is shifting fast. AI Overviews, AI Mode, and zero-click behavior are changing quarter to quarter; any difficulty model must be re-baselined regularly. Google's own claims that AI Overviews keep click volume "relatively stable" conflict with independent publisher data and Pew — unresolved dispute.
5. Most data is US-centric. Whitespark (Edmonton, Alberta) and BrightLocal provide some Canadian-relevant local-search grounding, but vertical- and country-specific difficulty can differ materially.
6. Tool numbers are not measurements. KD / DA / DR are proprietary estimates, not Google signals; Google does not use them.

Quarantined vendor sources: Ahrefs, Semrush, Moz, SE Ranking, Keywords Everywhere, SpyFu, Serpstat, Mangools, Ubersuggest, Whitespark, BrightLocal, Uberall, Sterling Sky, GatherUp, Seer Interactive, NP Digital, SparkToro.

Non-vendor / higher-trust sources used: Pew Research Center (Pew 2025 browsing-data study), Michael Luca / Harvard Business School (peer-reviewed-grade Yelp study), Northwestern Spiegel Research Center (academic), US DOJ v. Google trial testimony, Google's own documentation and spokespeople (Mueller, Ohye, Nayak), and peer-reviewed review-impact literature.