Why Publishing More Content Is Making Your SEO Worse

Search Engine Journal (Carolyn Shelby, Principal Consultant at CSHEL Search Strategies, Jan 2026): the old "more pages = more chances to rank" logic has inverted under AI retrieval — publishing volume now actively dilutes visibility. The argument is the supply-side mirror of brand-depth-ai-recommendations: where that page says brand depth wins recommendations, this one says undisciplined volume destroys it.

The inversion

Traditional search (pre-2015) rewarded coverage: "a site with 5,000 pages simply had more opportunities to appear than a site with 50" — the "blogging-for-dollars" model. AI systems evaluate chunks, not whole pages, and reward authority-density ("clarity, not volume") instead. So the same volume that used to compound now works against you.

Four mechanisms of harm

Semantic dilution. Overlapping articles "scatter signals across multiple weak or partially redundant pages" instead of reinforcing "one strong semantic center" — so "no single page accumulates dominant semantic weight."
Internal vector competition. "Your pages are not just competing for rankings. They are competing for embeddings" — near-duplicate pages cannibalize each other in embedding space, weakening retrieval dominance. This is keyword-cannibalization generalized from keyword overlap to semantic/vector overlap.
Crawl inefficiency. "Thin archives, redundant pages, obsolete content, tag explosions, faceted navigation problems" burn crawl budget and dilute internal-link equity. This bullet now has its own mechanism node — crawl economics (scaled-ai-content-crawl-economics): the supply-side half of the inversion, where the other three mechanisms are demand/retrieval-side. Google won't spend the compute to crawl, index, or retain the thin pages (burst-crawl → decay → de-index within ~75–140 days), so volume doesn't just lose in embedding space — it fails to stay in the index at all.
Entity-coherence degradation. "When sites publish endless disconnected content purely to chase search demand, they weaken their own entity coherence" — directly the entity coherence dimension of brand-depth-ai-recommendations, here being eroded rather than built.

The remedy — "publish with intent"

"Visibility is no longer a volume game. It is a clarity game." Shelby's prescription: audit the ecosystem honestly (which pages add unique value, which fragment a topic, which exist only for legacy SEO); consolidate aggressively ("one exceptional, highly structured, deeply authoritative page" beats "twenty mediocre supporting articles"); prioritise structural clarity (headings, lists, declarative language, focused paragraphs) for extractability; reframe success from publishing velocity to authority/usefulness; strengthen cornerstone assets and intentional internal linking.

This lands squarely on the optimize-vs-abandon fork — it is firmly an optimize-the-channel prescription (consolidate and win retrieval), the opposite pole from "stop chasing the channel." It also gives the "AI volume alone fails" thread a mechanism: not just that mass AI content doesn't help, but that it measurably hurts via vector competition.

Caveats

T4 trade-press op-ed, no quantitative evidence — the argument is entirely conceptual reasoning about how embedding/retrieval systems behave (no studies, benchmarks, or before/after data). The mechanisms (semantic dilution, vector competition) are plausible models, not measured effects, and the measurement gap applies in full: you can't observe whether consolidation actually improved AI consideration. Vendor-adjacent (a search consultant arguing for audit/consolidation work).

Why T4 (restated 2026-08-30)

The ## Caveats above were written when T4 read "trade press / SEO blogspam", and that reading is now wrong about this page. Since the 2026-08-30 tier decision T4 means unevidenced, and Carolyn Shelby's argument is the other kind of unevidenced: a named principal consultant reasoning from practice, not filler. Nothing about the grade says the mechanisms are wrong. Semantic dilution and internal vector competition are coherent predictions about how embedding retrieval behaves, and the tier records only that nobody has measured them, including the one that would matter most: nobody can yet observe whether consolidating twenty pages into one improved AI consideration, which is the piece's own prescription.

authority-density · brand-depth-ai-recommendations · keyword-cannibalization · great-content-no-longer-works · ai-content-seo-visibility · generative-engine-optimization · crawl-budget · scaled-ai-content-crawl-economics

Re-checked 2026-09-02 (absence-claim drain): the claim holds, and what turned up is worth recording because of what it is. A search for measurement of consolidation's effect on AI visibility returns plenty of numbers — a claimed 40–60% lift in citation rates from merging overlapping pages, a case study going 16% → 74% citation rate in 90 days by restructuring 15 pages, pruning write-ups quoting +23% to +104% organic traffic. Every one is an agency or vendor publishing its own client result: no control, no methodology, no sample frame, and each firm selling the service its number recommends. That is the same evidentiary class this page already discounts, arriving in greater volume, so it does not move the claim. Nobody has measured semantic dilution; several people are now selling the cure.