SpamBrain is Google's machine-learning spam system. It doesn't need a manual reviewer to notice that your inbound anchor distribution is unnatural — it notices algorithmically, and it demotes algorithmically.
The giveaway isn't one bad anchor. It's the ratio. A page with 43% of its inbound anchors saying "buy widgets online" has been manipulated; a page with 4% saying that is simply well-optimized. The threshold is fuzzy but it exists, and SpamBrain has crossed it long before a human manual action lands.
The Anchor-Text Manipulation Detector is an on-page version of that audit. It scans every anchor pointing out from your page and every anchor pointing in to your page (where crawlable), profiles the distribution, and flags the patterns SpamBrain cares about.
What the tool flags
- EMA over-optimization — exact-match commercial phrases repeated past ~8-12% of total anchors
- PMA stacking — partial-match anchors ("best widgets for small business") stacked alongside EMA such that combined share exceeds ~25%
- Branded-anchor starvation — natural link profiles are 30-60% branded; pages at <15% branded are suspicious
- Naked-URL deficiency — natural profiles include 10-30% raw URLs; zero naked-URLs = curated = suspicious
- Generic-anchor starvation — "click here", "read more", "this article" should be 5-15% of the mix
- Velocity anomalies — sudden clusters of new EMA anchors in a short window (harder to detect from on-page; flagged when found in a fresh link profile)
- Keyword co-occurrence — the same commercial phrase appearing across different linking domains in identical form
Why the fix isn't "just remove bad links"
Removing links signals to SpamBrain that you knew they were manipulative. The better play:
- Diversify new link building — push inbound anchors toward branded and generic for the next 3-6 months until the ratio normalizes
- Internal-anchor correction — you control 100% of your internal anchors; re-balance those first
- Only use Disavow for paid/penalty-risk links — Google's guidance is that Disavow is a last resort, not a routine hygiene tool
How to use it
- Go to /tools/anchor-text-manipulation-detector/
- Paste a URL
- The tool fetches the page, extracts every anchor, and profiles the distribution
- Read the score + findings
- Copy the fix prompt into Claude or ChatGPT to get a rewrite plan for your internal anchors
Typical runtime: ~3-5 seconds per page. No API key, no signup.
Related reading
- Anchor-Text Entropy Scanner — complementary tool measuring diversity via Shannon entropy
- Internal Link Equity Flow — PageRank-style flow analysis
- Every new AEO tool for 2026
Fact-check notes and sources
- SpamBrain (Google's ML spam system): Google Search Central — Webmaster Guidelines (Spam).
- Exact-match anchor penalty history: Google Penguin update (2012).
- Disavow guidance (last-resort tool): Google Search Central — Disavow links.
- Natural anchor-text distribution patterns: Ahrefs — Anchor Text Study.
This post is informational, not SEO-consulting advice. Mentions of Google, Ahrefs, Claude, ChatGPT, and similar products are nominative fair use. No affiliation is implied.
