技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
seo-keyword-strategist【数据与研究】Use when researching keywords, mapping search intent, building topic clusters, detecting keyword cannibalization, or optimizing content for entity-based search. Covers keyword strategy, intent classification, clustering, and cannibalization diagnosis.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Keywords are entity pointers, not density targets. Google's Knowledge Graph holds 54+ billion entities and 1.6+ trillion facts. Modern keyword strategy is about mapping content to entities and intent, not counting keyword occurrences.
| Myth | Reality |
|---|---|
| Keyword density 0.5-1.5% | No correlation with rankings (2026 study, 1,536 results). Google uses NLP, not word counting |
| LSI keywords | Confirmed myth. John Mueller: "There's no such thing as LSI keywords." Use "semantically related terms" instead |
| Content scoring tools predict rankings | Correlation 0.17-0.30 (weak). Use for gap finding, not score chasing |
| Search volume is accurate | Google Keyword Planner overestimates 54% of the time. Use volumes as directional only |
| Intent | % of Searches | Content Type | AI Overview Risk |
|---|---|---|---|
| Informational | 52.65% | Guides, how-tos, explainers | HIGH — AI answers directly |
| Navigational | 32.15% | Brand/product pages | LOW — users want specific site |
| Commercial | 14.51% | Comparisons, reviews, "best X" | MEDIUM — sweet spot for clicks |
| Transactional | 0.69% | Product/checkout pages | LOW — users want to buy |
Priority: Target commercial intent keywords — high conversion potential, not fully absorbed by AI Overviews. Informational head terms are increasingly zero-click (58.5% US, 83% when AI Overviews appear).
Google shifted from "strings" to "things." Optimize for entities, not keyword strings.
Three pillars:
mainEntityOfPagePractical steps:
SERP-based clustering is the gold standard. If 3+ of the same URLs appear in top 10 for two keywords, they belong in one cluster.
Rule: One search intent = one keyword cluster = one page.
Method:
Not every instance of multiple pages ranking for the same keyword is a problem. Ahrefs reports 9,700 cases on their own site — almost none need fixing.
It's a problem when:
It's NOT a problem when:
Detection via GSC API:
Query dimensions: ["query", "page"]
Group by query → count distinct pages
Flag: queries where count(page) > 1 AND positions swap frequently
Resolution decision:
| Signal | Action |
|---|---|
| Position swapping, same intent | Consolidate — merge weaker into stronger, 301 redirect. Case studies show +200-466% traffic |
| Different intents, both stable | Leave it — keyword diversification is beneficial |
| Same intent, one clearly dominant | Re-optimize — change weaker page to target different cluster |
58.5% of US searches result in zero clicks. When AI Overviews appear, 83% zero-click.
| Don't | Why |
|---|---|
| Target keyword density percentages | Dead metric — no ranking correlation |
| Use "LSI keyword" tools | LSI is a myth; use topical gap analysis instead |
| Chase content scoring tool numbers | 0.17-0.30 correlation; score ≠ ranking |
| Trust search volume precision | 54% overestimated; use as directional only |
| Create separate pages for same-cluster keywords | Causes cannibalization; one cluster = one page |
| Treat all multi-page rankings as cannibalization | Often beneficial; check for actual harm first |
| Ignore zero-click data | 58.5% of searches get no clicks; factor into strategy |