why tube tag pages are broken and who they're actually for

tag pages are the part of a tube nobody designs on purpose. they’re a byproduct of the seo strategy, not a feature built for the person clicking through. and you can feel it the second you land on one.

what a tag page is supposed to do

click a tag and you should get a filtered, ranked set of scenes that match what that tag actually means. that’s the whole job. one job.

pornhub, xvideos, and xhamster all fail it in slightly different ways. pornhub’s tag pages default-sort to “relevance,” which in practice means whatever their engagement model thinks will keep you clicking, not whatever best matches the tag. xvideos tag pages mix in autoplay-adjacent recommendations before you’ve even scrolled past the fold, so half the real estate isn’t the tag at all, it’s just more of whatever’s trending site-wide. xhamster’s tag pages are probably the cleanest of the three, but even there the tag often just means “this scene’s uploader typed this word into a box,” with zero curation on whether it’s accurate.

none of these are filtering for relevance. they’re filtering for retention. a tag page on a tube isn’t a search result, it’s a session-length lever with a label on it that looks like a category.

the seo long-tail trap

here’s the thing that actually explains the mess: these tag counts aren’t accidents, they’re strategy. pornhub has thousands of tags in its system. so does xvideos. most of them return under 20 scenes. some return three.

that’s not a taxonomy failure, that’s the taxonomy working exactly as designed, just not for you. every tag is a landing page. every landing page is a shot at ranking for some long-tail search query nobody else bothered to target. “tag has 20,000 scenes” is a bad seo bet because you’re competing against every other tube for that same head term. “tag has 14 scenes and an exact-match url” is a great seo bet because you might be the only page on the internet that ranks for it.

so the incentive is to mint tags, not to consolidate them. more tags means more indexed urls means more chances to catch a stray google click. whether the tag is useful to a human looking for something specific is not part of that calculation at all. it’s a side effect, sometimes a good one, usually not.

this is why tube category taxonomy looks the way it does when you actually dig into it: broad head-term tags with a firehose of content sorted by an engagement algorithm, sitting next to thousands of near-empty tags that exist purely to catch search traffic. two completely different products wearing the same ui.

dead tags, duplicate tags, and tags that don’t mean the same thing twice

spend twenty minutes clicking around any big tube’s tag list and you’ll hit three failure modes back to back.

dead tags: pages that return zero or near-zero results, still linked from somewhere, still indexed, still eating a crawl budget slot that could’ve gone to something with content behind it.

duplicate tags: the same concept split across three or four spellings because nobody ever ran a merge pass. singular versus plural, hyphenated versus not, an abbreviation next to the spelled-out version. each one has its own tiny, mostly-empty scene pool instead of one page with real depth.

and the worst one: tags that mean genuinely different things depending on which studio or uploader typed them in. a tag can mean one specific niche on one source and something adjacent-but-different on another, with no site-side normalization between them. you click through expecting one thing and the results are a coin flip. multiply that across five or six major sources and you start to understand why “dedup” is such a real problem at scale, not a nice-to-have. the same scene can show up on four to seven different platforms under four to seven different tag sets, and nothing upstream is reconciling that for the viewer.

none of this is a conspiracy. it’s just what happens when tag creation is unmoderated and the incentive is “more indexed pages,” not “more useful pages.” nobody’s evil here, the taxonomy is just optimizing for a different customer than you. the customer is googlebot.

what a tag taxonomy for viewers actually looks like

if you were building this for the person clicking, not the crawler, it’d look almost nothing like what’s live today.

first, a hard floor. a tag doesn’t exist as a browsable page until it clears a real minimum, something like 10 scenes, not 3. under that, it’s not a category, it’s noise wearing a url.

second, one canonical tag per concept, with the near-duplicate spellings folded in behind it instead of split into competing pages. you shouldn’t have to guess which of four spelling variants has the deeper pool.

third, source and tag transparency on the page itself. tell me where each scene actually came from and what it was tagged as at the source, because that’s the only way a tag stays honest when five different platforms mean five different things by the same word.

fourth, sort by something closer to actual signal, not by whatever maximizes time-on-page. relevance should mean relevance, not “the algorithm thinks you’ll stick around for this.”

that’s roughly the shape i built prowl’s category pages around. ~35 categories currently clear a real scene-count floor, each one shows a source breakdown and tag chips so you can see where the content is actually coming from instead of trusting a single unlabeled pool. it’s not trying to have 3,000 categories. it’s trying to have the right 35.

the honest version of a tag page is smaller than the seo-optimized version. that’s the tradeoff nobody selling you a directory wants to say out loud, and it’s exactly why most of them don’t make it.

– alex