algo-seo-backlink
評估反向連結品質，找出有毒連結並規劃連結建置
中文翻譯
---
name: "algo-seo-backlink"
description: "用 Domain Authority、Domain Rating 和信任度指標評估反向連結的品質。當使用者需要評估連結檔案的健康度、找出有毒的反向連結、或規劃連結建置策略時使用，即使他們說的是「檢查我的反向連結」、「連結建置」或「網域權威分析」也一樣。"
metadata:
category: "WP-35 SEO 演算法"
tags: ["seo", "backlinks", "domain-authority", "link-building"]
---
# 反向連結品質評估（Backlink Quality Assessment）
## 概述
反向連結分析會用品質指標（DA/DR、相關性、錨點文字多樣性、毒性）評估外部連入的連結，藉此評估網站站外 SEO 的強度。品質評估是啟發式的，用第三方指標（Moz DA、Ahrefs DR）當作 PageRank 的代理。
## 何時使用
**觸發條件：**
- 要稽核網站的反向連結檔案，檢視 SEO 健康度
- 要找出並 disavow 有毒或垃圾的連結
- 要依競爭對手分析來規劃連結建置策略
**何時不要用：**
- 要最佳化站內內容時（用內容 SEO）
- 要從原始連結圖計算真正的 PageRank 時（用 PageRank 演算法）
## 演算法
```
鐵律：反向連結的「品質」重於數量
一個來自高權威、主題相關網域的連結，
比數百個來自低品質網站的連結更有價值。
每個連結都要從以下幾點評估：
1. 權威度（連出網域的 DA/DR）
2. 相關性（連出頁面和目標頁面的主題吻合度）
3. 位置（編輯內文中的 > 頁尾／側邊欄）
4. 錨點文字（自然多樣 > 完全比對的關鍵字堆砌）
```
### 階段 1：輸入驗證
從 Ahrefs、Moz 或 Search Console 匯出反向連結資料。必要欄位：參照網域、DA/DR、錨點文字、連結類型（dofollow/nofollow）、首次發現日期。
**關卡：** 完整的反向連結匯出，並附有權威度指標。
### 階段 2：核心演算法
1. 依參照網域去重（分析時每個網域只算一個連結）
2. 替每個連結評分：權威度（0-100）× 相關性（0-1）× 位置權重
3. 標記有毒連結：DA < 10、不相關的外語、連結農場模式、PBN 跡象
4. 計算檔案指標：參照網域總數、DR 分布、錨點文字多樣性指數
### 階段 3：驗證
把標記為有毒的連結，和已知的垃圾資料庫交叉比對。驗證錨點文字的分布符合自然的模式（品牌 > URL > 關鍵字 > 其他）。
**關卡：** 有毒連結已辨識、錨點檔案已分析。
### 階段 4：輸出
回傳檔案評估，附上連結品質分布和行動項目。
## 輸出格式
```json
{
"profile": {"referring_domains": 450, "avg_dr": 35, "toxic_count": 23, "anchor_diversity": 0.78},
"actions": [{"type": "disavow", "domains": ["spam1.com"], "reason": "link farm pattern"}],
"metadata": {"tool": "ahrefs", "export_date": "2025-01-15"}
}
```
## 範例
### 輸入輸出範例
**輸入：** 500 個反向連結、200 個參照網域
**預期：** 分布：15% DR 60 以上、40% DR 20-59、45% DR 0-19。標記 23 個有毒網域要 disavow。
### 邊界情況
| 輸入 | 預期 | 原因 |
|-------|----------|-----|
| 所有連結都來自同一個網域 | 檔案多樣性低 | 依賴單一來源有風險 |
| 90% 是完全比對的錨點 | 有錨點文字被懲罰的風險 | 不自然的錨點模式 |
| 零反向連結 | 先專注在內容 | 不存在的東西沒辦法最佳化 |
## 常見陷阱
- **DA/DR 是第三方的估計值**：它們是 PageRank 的近似，但「不是」Google 的指標。兩個工具對同一個網域的權威度常常有出入。
- **Nofollow 仍然重要**：Google 把 nofollow 當作「提示」。來自 DR 90 網站的 nofollow 連結仍有 SEO 價值，只是比 dofollow 少。
- **謹慎 disavow**：Google 的 disavow 工具是最後手段。disavow 正當的連結會傷害你自己的檔案。只對明確有毒／垃圾的連結做 disavow。
- **錨點文字的操弄**：完全比對的錨點文字以前是排名因素，現在是垃圾訊號。自然的檔案，錨點大多是品牌和 URL。
- **時間模式**：反向連結突然暴增（例如一天 100 個）會觸發垃圾過濾器。自然的連結取得是漸進的。
## 參考資料
- 連結毒性評分的方法，見 `references/toxicity-scoring.md`
- 競爭對手的反向連結缺口分析，見 `references/competitor-gap.md`英文原文
---
name: "algo-seo-backlink"
description: "Evaluate backlink quality using Domain Authority, Domain Rating, and trust metrics. Use this skill when the user needs to assess link profile health, identify toxic backlinks, or plan link building strategy — even if they say 'check my backlinks', 'link building', or 'domain authority analysis'."
metadata:
category: "WP-35 SEO 演算法"
tags: ["seo", "backlinks", "domain-authority", "link-building"]
---
# Backlink Quality Assessment
## Overview
Backlink analysis evaluates incoming links by quality metrics (DA/DR, relevance, anchor text diversity, toxicity) to assess a site's off-page SEO strength. Quality assessment is heuristic-based using third-party metrics (Moz DA, Ahrefs DR) as PageRank proxies.
## When to Use
**Trigger conditions:**
- Auditing a site's backlink profile for SEO health
- Identifying and disavowing toxic or spammy links
- Planning link building strategy based on competitor analysis
**When NOT to use:**
- When optimizing on-page content (use content SEO)
- When computing actual PageRank from raw link graphs (use PageRank algorithm)
## Algorithm
```
IRON LAW: Backlink QUALITY Outweighs Quantity
One link from a high-authority, topically relevant domain is worth
more than hundreds from low-quality sites. Evaluate every link on:
1. Authority (DA/DR of linking domain)
2. Relevance (topical match between linking and target pages)
3. Placement (editorial in-content > footer/sidebar)
4. Anchor text (natural diversity > exact-match keyword stuffing)
```
### Phase 1: Input Validation
Export backlink data from Ahrefs, Moz, or Search Console. Required fields: referring domain, DA/DR, anchor text, link type (dofollow/nofollow), first seen date.
**Gate:** Complete backlink export with authority metrics.
### Phase 2: Core Algorithm
1. Deduplicate by referring domain (one link per domain for analysis)
2. Score each link: authority (0-100) × relevance (0-1) × placement weight
3. Flag toxic links: DA < 10, irrelevant foreign language, link farm patterns, PBN indicators
4. Compute profile metrics: total referring domains, DR distribution, anchor text diversity index
### Phase 3: Verification
Cross-reference flagged toxic links against known spam databases. Verify anchor text distribution follows natural pattern (branded > URL > keyword > misc).
**Gate:** Toxic links identified, anchor profile analyzed.
### Phase 4: Output
Return profile assessment with link quality distribution and action items.
## Output Format
```json
{
"profile": {"referring_domains": 450, "avg_dr": 35, "toxic_count": 23, "anchor_diversity": 0.78},
"actions": [{"type": "disavow", "domains": ["spam1.com"], "reason": "link farm pattern"}],
"metadata": {"tool": "ahrefs", "export_date": "2025-01-15"}
}
```
## Examples
### Sample I/O
**Input:** 500 backlinks, 200 referring domains
**Expected:** Distribution: 15% DR 60+, 40% DR 20-59, 45% DR 0-19. Flag 23 toxic domains for disavow.
### Edge Cases
| Input | Expected | Why |
|-------|----------|-----|
| All links from one domain | Low profile diversity | Single-source dependency is risky |
| 90% exact-match anchors | Anchor text penalty risk | Unnatural anchor pattern |
| Zero backlinks | Focus on content first | Can't optimize what doesn't exist |
## Gotchas
- **DA/DR are third-party estimates**: They approximate PageRank but are NOT Google metrics. Two tools often disagree on the same domain's authority.
- **Nofollow still matters**: Google treats nofollow as a "hint." A nofollow link from a DR 90 site still has SEO value, just less than dofollow.
- **Disavow carefully**: Google's disavow tool is a last resort. Disavowing legitimate links harms your own profile. Only disavow clearly toxic/spammy links.
- **Anchor text manipulation**: Exact-match anchor text used to be a ranking factor; now it's a spam signal. Natural profiles have mostly branded and URL anchors.
- **Temporal patterns**: Sudden spikes in backlinks (e.g., 100 links in one day) trigger spam filters. Natural link acquisition is gradual.
## References
- For link toxicity scoring methodology, see `references/toxicity-scoring.md`
- For competitor backlink gap analysis, see `references/competitor-gap.md`