Links as Information Flow (2006)

A hyperlink does not merely pass a scalar popularity score (a raw quantitative boost of PageRank). In the conceptual model originally published in «Links come passaggio d'informazione», a link between two documents behaves exactly like a physical vector force characterized by three inseparable properties:

• Magnitude (Intensity): the scalar weight, determined by source page authority divided across all outbound links.
• Direction: the semantic and topical axis, defined by language, surrounding text context, and named entities.
• Sense (Orientation): the direction of flow, traveling from the referring document to the target resource.

Treating hyperlinks as mere scalar numbers is structurally flawed: two forces identical in magnitude but divergent in direction do not double total velocity—they pull the resultant vector off course.

In September 2006, this vector hypothesis was tested by observing an Italian-language website that suddenly received a concentrated influx of 565 inbound links originating exclusively from English-language domains.

Conventional SEO theory assumed any backlink was an absolute positive. The empirical outcome was the exact opposite: the site lost ranking visibility for its primary Italian search terms.

This demonstrated relative penalization: the 565 English links carried no manual spam penalties, but their magnitude, exerted along an English semantic axis, deflected the website's resultant vector far away from its native linguistic center of gravity. Applying strong vector force along a misaligned axis neutralized its authority along the primary trajectory.

Every link—inbound or outbound—modifies a document's coordinate within informational space. When a web page links to heterogeneous topics, it does not merely leak authority: it fragments its vector force across multiple directions, diluting its magnitude along the primary axis.

Likewise, accumulating citations from unrelated contexts scatters the resultant vector. In distributed graph structures, authority is not volumetric accumulation—it is coherent vector alignment.

This physical vector approach maps directly to modern dense vector embeddings utilized by neural retrieval engines. Contemporary Information Retrieval models represent documents, queries, and entities as coordinates in high-dimensional semantic spaces, evaluating relevance through cosine similarity between vector directions.

In production, the governing dynamic remains identical to what was observed in 2006: topological coherence consistently outperforms raw link volume. A handful of contextually aligned citations transmit a far cleaner relevance signal than a diffuse scatter of generic mentions.