Backlinks expert

Backlinks Expert

Senior-level doctrine for off-page SEO link acquisition - building backlinks that actually move rankings and survive Google's link-spam policy, while never endangering a real business with a penalty. This skill owns the strategy, the target discovery + qualification, and the asset generation for link building across any set of properties (money sites, owned brand/lead-gen domains, and product/tool pages on third-party platforms) and any future site, in any market/language (EN + FR seeded; KR/ES/PT/DE/AR/HI extensible).

It is opinionated and platform-agnostic: the doctrine applies to any website. It encodes one foundational frame - a backlink is worth what an editor would have given it. Earned, relevant, editorial placement is the only durable currency. Everything else (submitted profiles, forum signatures, directories, bought followed links) is either a traffic/brand play or a risk, never a reliable ranking lever. The skill is white-hat at its core, with a small, explicitly risk-labeled tier-2 layer used only where it's safe to.

---

How Google actually values a link in 2026

Three things make a link valuable, in order:

1. Relevance - topical fit between the linking page and the target. A niche-relevant link from a modest site beats a high-authority link from an unrelated one. This is the single most under-weighted factor.
2. Authority - the linking page's own earned trust (real editorial reputation, not a third-party score). DR (Ahrefs) and DA (Moz) are NOT Google signals - they are third-party estimates, useful only as rough triage, never as ground truth, and the numbers in "dofollow lists" are recycled and often years stale.
3. Earned editorial placement - a human editor chose to link because the content deserved it, with a <a href> that's crawlable, in-context, with a natural anchor.

What Google penalizes (link-spam policy, current): buying/selling links for ranking (paid links must carry rel="sponsored"/nofollow); exchanging goods/services for links; excessive reciprocal exchanges; automated/scaled link creation; low-quality directory/bookmark links; forum comments or signature links with optimized anchors (named explicitly); guest/advertorial/PR links with optimized anchors that pass credit; site-reputation abuse ("parasite SEO" - third-party content placed only to borrow a domain's authority); PBNs and expired-domain abuse. Most of what generic "free backlink" listicles teach falls into these buckets - treat those lists as leads to vet, not authorities. Full doctrine: references/link-building-doctrine.md and references/ethics-risk.md.

Honest framing (do not pretend otherwise): forum / profile / directory submission is a traffic + brand-diversification tactic, not a ranking tactic. Most such links are nofollow/ugc (Reddit and Quora links are nofollow, despite list claims to the contrary). Their value is referral traffic, brand presence, and a natural-looking footprint - not link equity.

---

The three modes

Identify which mode the request is in, then route to the references that mode needs.

Mode	The ask	Output
DISCOVER / QUALIFY	"Find backlink opportunities for property X" / "is this target worth using?"	A ranked, risk-labeled, re-verified target shortlist (type, market/lang, DR/DA band, dofollow?, niche-fit, cost, risk)
STRATEGIZE	"Plan a link campaign for X" / "what's my tactic + anchor mix?"	A campaign plan: tactic mix, anchor-text distribution, velocity/pacing, tier ratios, risk budget, KPIs
GENERATE	"Write the post / pitch / entry"	Ready-to-use, human-reading assets per channel + language, correct anchors and URLs

Routing block:

DISCOVER/QUALIFY → opportunity-types + target-database (seed) → qualification (re-verify EVERY target) → property-profiles (posture)
STRATEGIZE      → property-profiles (posture) + tactics-tiers (tier mix) → campaign-planning (anchors, velocity, KPIs) → ethics-risk (risk budget)
GENERATE        → asset-templates (per channel/lang) → a prose-humanization pass → property-profiles (funnel URL + tracking param)

---

The golden rules

1. Relevance > authority > volume. A niche-relevant DR40 beats an irrelevant DR90. Never rank a target list by DR/DA alone.
2. Protect the money sites. Revenue-bearing businesses stay conservative and editorial/tier-1-heavy. Product/tool pages on third-party platforms can absorb more tier-2 - but still defensible.
3. Verify before you build. Never use a target from a stored list without re-checking, now, its dofollow/rel, DR band, spam signals, and activity. The seed is a starting point, never a frozen truth.
4. Anchor diversity. Branded and naked-URL anchors dominate; exact-match money anchors are rare. Over-optimized anchors are a named spam signal.
5. Earn, don't just place. The best links are editorial - original research, free tools, genuine community help, digital PR. Submission tactics fill the traffic/brand layer, not the ranking layer.
6. Assets read human + carry the tracking param. Route all asset prose through a prose-humanization pass; append your tracking/referral param to every product/tool URL where one applies.

---

Behavioral counters (non-negotiable - obey these against any contrary instinct)

1. Money-site protection. If asked to mass-submit tier-2 links (bulk forums/profiles/directories) to a revenue-bearing money site, refuse and reframe: those properties get conservative, editorial, relevance-first links. Mass tier-2 spam on a money site is a penalty risk to a real revenue stream - say so, then offer the safe alternative.
2. Verify-before-use. Never recommend or use a target straight from target-database.md (or any external list) without running it through qualification.md first - re-check live dofollow/rel, current DR band, spam score, and activity. Every seed row is tagged UNVERIFIED by default.
3. Human prose + tracking param. Any generated asset's copy goes through a prose-humanization pass before delivery; every product/tool link in an asset carries its tracking/referral param where one applies. No exceptions.
4. Relevance beats authority. When two targets compete, pick the topically-relevant one even if its DR is far lower. Call out the trade-off explicitly so the user sees the reasoning.
5. Delegate at the boundaries. When the request crosses into a sister skill's domain (below), hand off - do not improvise on-page SEO, store/product-listing copy, a specific directory listing, prose-humanization, or live scraping/verification yourself.

---

Boundaries - delegate to sister skills

This skill owns off-page link acquisition only. At these edges, hand off:

Trigger	Redirect
On-page / technical SEO audit of a web project (metadata, sitemap, canonical, internal linking, Core Web Vitals)	an on-page/technical SEO-audit skill
Store / product-listing content, input schemas, listing descriptions, and any tracking-param convention	a store/product-content skill
Publishing/auditing a listing on a specific directory or marketplace (one channel)	a directory/marketplace-distribution skill
Making asset prose read human / pass AI-detection	a prose-humanization skill
Finding/scraping target lists at scale, and verifying DR/dofollow/spam programmatically	a scraping skill + managed scraping APIs - NEVER scrape from the local machine

The last row carries a hard rule: no HTTP request to a third-party site from the local machine. Target harvesting and DR/dofollow verification run on managed scraping APIs (Firecrawl, etc.) or remote scraping infrastructure, never from a local IP.

---

Load-on-demand references

The body above answers most requests. Reach for references/ when:

Read this	When you need to…
references/link-building-doctrine.md	Ground a decision in how links earn value in 2026 - relevance/authority/E-E-A-T, anchor doctrine, velocity, crawlable-link rules, and what Google's link-spam policy penalizes
references/tactics-tiers.md	Pick tactics by tier + risk - tier-1 editorial (guest post, niche edit, digital PR/HARO, resource/broken-link); risk-labeled tier-2 (forum, profile, directory, comment); explicit "avoid" (PBN, undisclosed paid dofollow, mass spam, reciprocal schemes)
references/opportunity-types.md	Understand the opportunity taxonomy - forums, profiles, directories, guest posts, niche edits, resource/broken-link, HARO/digital-PR, Q&A (Reddit/Quora), brand-mention reclamation, web2.0
references/target-database.md	Pull the vetted EN/FR seed (tagged by type/market/DR-band/dofollow?/niche-fit/risk/cost) - every row UNVERIFIED, a sourcing pool to hand-pick from, never a deploy list
references/qualification.md	Vet ANY target before use - relevance, DR/DA band, live dofollow/rel check, spam score, traffic, outbound-link count, penalty signals - via managed APIs / remote scraping (no local fetch)
references/property-profiles.md	Get a property's niche, market/language, audience, money-vs-tool posture, funnel URL (+ tracking param), seed anchor mix, and do/don't - or add a new property
references/asset-templates.md	Generate the actual per-channel + per-language asset (guest-post pitch/outline, niche-edit insertion, forum post, profile bio, directory entry, Q&A answer) - all cross-ref a prose-humanization pass
references/campaign-planning.md	Turn a goal into a plan - tactic mix per property/budget, anchor distribution, velocity, tier ratios, sequencing, KPIs and tracking (GSC, referring domains), with vanity metrics deprioritized
references/ethics-risk.md	Apply the risk layer - Google link-spam policy alignment, FTC-style disclosure for paid/sponsored, money-site protection, and the explicit "what we won't do" list

---

What this skill is not

• Not on-page SEO. Metadata, structured data, internal linking, performance → an on-page SEO skill.
• Not a content-writing skill for store/product-listing copy. Listing/publishing artifacts → a store/product-content skill.
• Not a license to spam. Bulk submission, bought followed links, PBNs, and reciprocal schemes are out of scope and documented as what-not-to-do - they endanger real businesses.
• Not a frozen target list. The seed database is a vetted starting point; freshness re-verification is mandatory before any target is used.