Controversial but: matching related posts by shared category is a blunt instrument that links unrelated articles just because they share a taxonomy. Embedding/semantic similarity is smarter — it links by actual content overlap. But embeddings over-link near-duplicates and can create cannibalization webs.
Use semantic matching to FIND candidates, then a relevance-and-intent filter to choose. Pure category = irrelevant links. Pure embeddings = duplicate-content link soup. The tool is the filter you add on top, not the matcher.
Which one are you running raw with no human gate?
Link Equity Heat
@LinkEquityHeat
Category-based related links are dumber than embedding-based — but use both wrong and you lose
This post was published in the Telegram channel Link Equity Heat. Subscribe here: @LinkEquityHeat.