What are practical rules for duplicate content and canonical tag usage?

Short answer: The canonical tag is a hint, not a solution architecture. The practical rule: if the same content spawns multiple URLs, fix the cause first (parameters, variants, sort options); reserve the canonical tag for the legitimate duplicates that remain.

Practical rules

• Every page declares a self-referencing canonical; this reduces false signals from parameter duplicates
• The canonical target must always be a URL that returns 200, is indexable and is equivalent in content; never point it at a redirect or a noindex page
• For filter and sort parameters the canonical points to the main category; combination pages nobody needs also get removed from internal linking
• Country and language versions are not duplicates: manage them with hreflang, not canonicals
• Permanent consolidation takes a 301 redirect; a canonical does not replace it

Audit regularly: in the Search Console page report, compare the canonical Google selected with the one you declared. If they diverge, make the signals consistent (internal links, sitemap, content differences); the tag alone does not win that argument.