Background & myths · Knowledge
Link attribute vs. meta robots nofollow – what differs?
Same word, two levels: rel=“nofollow” sits on a single a-tag and withdraws one vote; meta name=“robots” content=“nofollow” sits in a page's head and withdraws every vote the page casts – scalpel versus sledgehammer.
Both are hints in the modern era, and both exist as HTTP headers too (X-Robots-Tag) – useful for PDFs and other non-HTML files.
When which level fits – and the popular pair that is usually a mistake
The link attribute is the everyday tool, because vouching is a per-link question: this comment link no, that source link yes – differentiated policy is its whole point, and site-wide enforcement with exceptions (the plugin approach) automates exactly that differentiation. Page-level meta nofollow answers a different, much rarer question: “should any link on this page count as my endorsement?” – defensible for fully user-generated pages you cannot moderate, auto-generated aggregations, or archive corners of dubious inherited content; even there, per-link ugc/nofollow usually does the same job with more honesty and less collateral. The pairing people actually mean is mostly about indexing, which brings in nofollow's stronger sibling noindex – and the popular combination noindex,nofollow that deserves its warning: pages kept out of the index usually should still pass their internal links (navigation flows through them!), so noindex,follow is the sane default for filtered lists, tag pages and thin utility pages – adding nofollow cuts internal circulation for no gain; the practical difference has narrowed over the years (long-noindexed pages fade as link sources anyway), but choosing the harsher pair on purpose still needs a reason. Completing the toolbox row: robots.txt governs fetching (the crawl-budget question), noindex governs appearing, nofollow – on either level – governs vouching; most “robots problems” dissolve the moment those three verbs are kept apart. And the X-Robots-Tag header carries all of it to non-HTML files, which is how a PDF gets its noindex.
Key facts
- Levels: rel=nofollow = one link; meta robots nofollow = every link on the page.
- Page level is rarely right – per-link ugc/nofollow does the job with less collateral.
- Sane default for index control: noindex,follow – adding nofollow cuts internal circulation.
- Three verbs, three tools: fetching = robots.txt · appearing = noindex · vouching = nofollow.
- X-Robots-Tag: the header variant brings all directives to PDFs and other non-HTML files.