LinkGuard cover — Anchor text portfolio health checklist (what 2026 data actually shows)

Your agency dashboard shows the client's portfolio at 14% exact-match commercial anchors. The 2018 SEO blog post you're reading says "anything above 5% is dangerous". Your client just had a Core Update drop. You're about to file a disavow file.

Stop. The 2018 rule is folk wisdom, not 2026 data. Ahrefs published two relevant anchor-distribution studies. The bigger one (384,614 pages) found that the median exact-match-anchor percentage across all ranking positions is zero, and the correlation between exact-match share and ranking is 0.1436 — weak enough that exact-match share alone explains essentially nothing about where pages rank. A separate older Ahrefs analysis of 16,000 keywords (~320,000 pages) found that top-5 ranking pages routinely show 6-13% exact-match anchors, the same band that gets clients panic-disavowing under the 2018 rule. Two studies, both Ahrefs, both pointing the same way: the old ceiling is wrong.

Two things changed since the old playbook. First, Penguin 4.0 in 2016 moved Google from "demote sites with bad-anchor patterns" to "ignore the bad anchors". Second, SpamBrain (Google's AI link-spam classifier deployed December 2022) looks like it treats anchor distribution as evidence about donor cluster quality, not as a direct ranking signal on the target — the mechanism isn't publicly documented at that level of detail, but the published case studies are consistent with it. The cleanest example is Joy Hawkins's August 2025 write-up: "Referring domains going up, organic traffic going down" because the donor cluster was quietly devalued retroactively. The operator pain looks like a penalty; the mechanism doesn't seem to be one.

Gary Illyes summarized the position more bluntly at a Bulgarian SEO conference in April 2024: "We need very few links to rank pages... we've made links less important than they used to be". Illyes later said he shouldn't have phrased it that way, but the underlying directional claim — links matter less than they did — is consistent with Google quietly dropping the word "important" from its links documentation in March 2024. John Mueller, asked about identical anchor texts across many backlinks in April 2025: "fine & common to me". The 2018 anchor-ratio playbook is no longer the right tool.

This checklist is the 2026 version: how to audit anchor distribution, what's worth cleaning, what's wasted effort, and what target ratios are operator-defensible. 21 items across five stages: get the mental model right (pre-flight), see what you've got (audit), decide what to touch (action), prevent the same mess next year (build), avoid three classic mistakes (anti-patterns). Tier filter and browser-saved progress.

Vocabulary, before we start

SpamBrain: Google's AI link-spam classifier deployed in December 2022. Now does most of the link-spam neutralization the disavow tool used to do.
Penguin 4.0: 2016 algorithm update that shifted Google's anchor-spam response from "demote target sites" (push the target down in rankings) to "ignore bad links" (the link stops counting). Devalue, not demote.
PBN: private blog network. A cluster of sites (often on expired domains) built to manufacture backlinks to a money site. Black-hat tactic; Google's reviewers know the pattern.
Manual action: a human reviewer at Google has flagged your site for a specific policy violation. You'd see it in Search Console → Security & Manual Actions.
Core Update: Google's broad-strokes ranking refresh, typically 3-4 per year. Affects content quality scoring, not link signals.
Helpful Content update: Google's content-quality classifier targeting thin or unhelpful pages. Has nothing to do with backlinks.
Disavow tool: Search Console feature that tells Google to ignore specified inbound links. Useful for active manual actions or confessed paid-link history. See our disavow checklist for when it actually helps.
DR (Domain Rating): Ahrefs's 0-100 score for a domain's overall backlink authority. Used as a cheap pre-filter in donor qualification; not a Google signal.
Exact-match anchor: an inbound link whose text exactly matches your target keyword (e.g. "backlink monitoring tool" linking to a backlink-monitoring product page).
Partial-match anchor: contains the target keyword plus other words (e.g. "this backlink monitoring tool we tested").
Branded anchor: contains your brand name (e.g. "LinkGuard", "LinkGuard.ai").
Naked URL: the literal URL as the link text (e.g. linkguard.ai). Strong editorial-linking signal.
Donor cluster: a group of inbound link sources treated together by ownership pattern, IP, or content fingerprint. The operational test: if five or more donors share at least two of (same registrant, same /24 IP range, identical CMS/theme footprint, publication cadence sync inside ±48 hours, near-identical footer blocks), treat them as one cluster.

One more orientation note before pre-flight: this isn't a Google Search problem if your traffic loss is on AI-search citations. For that, see our GEO vs SEO guide — anchor distribution doesn't move the AI-citation needle, content and brand mentions do.

What success looks like

For most readers: an honest snapshot of the current distribution, a documented benchmark range you'll defend over the next 12 months, and the discipline to stop reacting to single-month ratio shifts. For the narrow cohort with real concentration footprints: a triage map of which donor clusters are likely to get neutralized (so you stop spending budget on links that already don't count), plus an outreach plan for the salvageable ones.

Related reading: anchor health and link attributes go together — see dofollow vs nofollow links: what they are and when each matters.

0 / 21 · 0%

Critical Important Nice to have

Show only undone

Four concept checks. The 2018 playbook will mislead you in three of these four. Get this right or every item downstream is solving the wrong problem.

Critical Drop the "<5% exact match or you die" rule — it's 2018 folk wisdom, not 2026 data Ahrefs published their own anchor distribution against ranking position. Median exact-match across all positions is zero. Correlation between exact-match share and ranking is 0.1436 (weak). Top-5 ranking pages routinely run 6-13% exact-match anchors. Operators who treat 5% as a hard ceiling spend budget fixing portfolios that are already in the ranking band.

How to do this

Before doing anything, get the current exact-match percentage for the property in front of you. If it sits between 4-15% and the donor cluster looks legitimate (see pf-02), it's well within the band Ahrefs' own data shows as ranking-compatible in 2026. The 2018 rule was specific to Penguin 1-3 (2012-2014), which Google retired with Penguin 4.0.
Critical Internalize the 2026 mental model: SpamBrain looks like cluster-level devaluation, not target punishment SpamBrain's internal architecture isn't publicly documented at the mechanism level. But the published case studies are consistent with cluster-level devaluation: when a spammy donor cluster stops counting, the target site doesn't get punished — the cluster just stops contributing ranking signal. Joy Hawkins's cleanest case study (August 2025): "Referring domains going up, organic traffic going down" because the donor cluster was quietly devalued. Looks like a penalty, doesn't seem to be one. Disavow doesn't fix it; better acquisitions do.

How to do this

What this means for audit work: stop asking "is this anchor risky"; start asking "is this donor cluster about to get neutralized". Operators who've internalized the cluster model stop bringing anchor-ratio panics to their team leads — they bring donor-cluster diagnoses. If your portfolio shows a sudden "referring domains up, traffic flat or down" pattern in the last 6 months, that's the Hawkins signature. Action: jump to a-04 and triage your donor clusters.
Critical Rule out Core Update and Helpful Content drops before touching anchors Most "my anchors are bad" complaints are actually Core Update drops or Helpful Content losses (algorithmic content suppression). Anchor work won't fix those. Misdiagnose this and you waste weeks on the wrong audit.

How to do this

Cross-reference the traffic-drop date with Google's announced ranking-system updates. If the drop aligns with a Core Update or Helpful Content update, head to a content audit instead. If the drop aligns with an anchor-related update (Joy Hawkins flags June 2024 and August 2025 as the link-relevant ones in this cycle), the audit below is the right work.
Nice Pick benchmark targets by site stage, not by universal rule A 6-month-old SaaS site and a 10-year-old publisher have different defensible distributions. New sites that show 15% exact-match commercial anchors in month 4 look manufactured. Established sites in the same band look like they earned competitive rankings. Match the benchmark to the stage.
How to do this
Starting heuristics we use internally as a first cut — treat them as ranges to calibrate against the property's own competitive set (a-03), not as thresholds:
New site, 0-6 months: branded + naked URL ≥70%, exact match ≤3%
Growing site, 6-24 months: branded floor 40-50%, exact match 3-8%
Established site, 2+ years: branded floor 30%, exact match 6-13% acceptable IF donor quality holds
These are starting ranges, not thresholds. The donor cluster quality (pf-02) is the load-bearing variable.

Five steps that produce the snapshot worth defending. Most agencies skip the categorization step (a-02) and make decisions on Ahrefs' raw "Anchors" report, which lumps everything into one ranked list.

Critical Pull the full anchor list from Search Console plus a backlink tool Search Console gives you Google's view of your inbound anchors (limited to ~1000 top occurrences but authoritative). Ahrefs or Semrush gives you depth + frequency. Both are needed: GSC tells you what Google sees; the third-party tool fills the long tail.

How to do this

GSC: Links → External links → Top linking text. Export.
Ahrefs: Site Explorer → Backlinks → Anchors. Sort by referring domains (not by total links) so you don't get one spammy footprint inflating a count. Semrush: Backlink Analytics → Anchors tab.
For client-side categorization without uploading the link profile to a third party, use our anchor distribution analyzer — paste the anchor list, get the five-bucket breakdown.
Audit anchor distribution free
Critical Categorize every anchor into five buckets: branded / naked URL / partial / exact-match commercial / junk Ahrefs' raw report lists anchors by frequency. It doesn't tell you whether "backlink monitoring software" is an exact-match commercial anchor for your money page or a partial-match for a blog post. You have to categorize. Without categorization, every downstream decision is guessing.
How to do this
The five buckets:
Branded: contains your brand name (LinkGuard, LinkGuard.ai).
Naked URL: the bare URL as the anchor (linkguard.ai). Strong editorial-link signal.
Partial match: contains the target keyword plus other words.
Exact match commercial: exactly the target money-keyword phrase.
Junk / irrelevant: "click here", "read more", or unrelated phrases. Mostly harmless; mostly dilution.
Tag each anchor in a spreadsheet column. Now you can sum percentages by bucket and compare to pf-04 benchmarks.
Audit anchor distribution free
Important Benchmark against 3-5 ranking peers in the same niche, not against universal rules Universal anchor ratios are misleading because competitive verticals have different natural distributions. Finance and law sites run much higher branded-anchor floors than open-source dev tools. Pick 3-5 currently-ranking competitors in your top SERPs and measure THEIR distribution; that's your defensible competitive band.

How to do this

Pick 3-5 sites that rank top-10 for your primary commercial query. For each, pull their anchor distribution (Ahrefs Site Explorer works). Average the bucket percentages. Your site's distribution should sit within ±5 percentage points of the competitive median on each bucket. Anything outside that band is worth investigating (not necessarily fixing).
Important Flag anchor-concentration footprints: the same exact phrase from 30+ unrelated donors Natural editorial linking produces a long tail of variation. When you see the same exact commercial phrase across 30+ unrelated donors inside 60 days, that's a footprint — SpamBrain's pattern detector looks for exactly this. Single high-frequency anchors aren't usually the problem; single high-frequency anchors from a concentrated donor cluster are.
How to do this
In your categorized list (a-02), sort exact-match commercial anchors by referring-domain count. Flag any anchor phrase with ≥30 distinct referring domains where the donors don't appear in any of your branded-anchor lists.
Velocity context: 30 in 60 days on a site organically acquiring 2 referring domains/month is screaming. 30 in 60 days on a site that gets 200/month is noise. Score against the property's baseline acquisition rate.
Shared-ownership tells (2026 set):
IP / network — dig +short each donor; group by /24 IP range. Flag groups of ≥5.
CMS / theme fingerprint — modern PBN clusters share identical WordPress theme + plugin combinations across supposedly-unrelated donors. Check page source for matching wp-content paths and generator meta tags.
Publication cadence sync — posts within ±48 hours across the cluster is a coordination signal.
Footer / sidebar link blocks — near-identical outbound link sets. Open 5-10 donors, diff their footers manually or with Screaming Frog's Custom → Extraction.
WHOIS / registrar — still useful when not privacy-shielded, but modern PBNs run Cloudflare-proxied WHOIS, so this signal degrades over time.
Nice Identify the "weird tail" — junk anchors that dilute but rarely hurt Most portfolios carry a 5-15% tail of "click here", "read more", "this article", non-Latin character strings, and obvious scraper junk. These rarely trigger anything algorithmic; they just dilute. Knowing the tail's size stops you from panic-attributing concentration patterns to it.

How to do this

Mark every uncategorizable anchor as junk. Calculate the junk-tail percentage. Anything under ~20% is normal long-tail noise; anything over 30% is worth a spot-check for scraper farms (which usually show as concentrated donor clusters — see a-04).

Five rules for when to act and when to leave the portfolio alone. Most operators over-act here, because acting feels productive and inaction doesn't.

Critical Triage by donor-cluster quality, not by anchor text alone A 10% exact-match commercial anchor from high-DR editorial donors is the footprint that helped pages rank for a decade. The same 10% from a concentrated PBN cluster is what SpamBrain neutralizes. Don't treat them the same in the audit. The anchor phrase is the symptom; the donor cluster is the underlying signal.
How to do this
For each flagged anchor (a-04), score the donor cluster on four signals:
Referring-domain diversity — unique registrants? unique /24 IP ranges? unique WHOIS contacts? Pull with dig +short for IPs, whois for registrant data.
Editorial context — is the link inside real content, or in a footer / sidebar / directory list?
Donor topical relevance — do the donors publish about your niche, or are they hosting unrelated thin pages?
Donor's own outbound anchor profile — if the donor's own outbound links to other sites are 40%+ exact-match commercial, that donor is a link-seller regardless of how editorial their content reads. Single best predictive signal.
Scoring buckets:
4 yes-answers: leave alone, those links are working.
2-3 yes-answers: monitor; don't acquire more from the cluster, but don't disavow either.
0-1 yes-answers: the cluster is at risk of neutralization. Treat as already discounted. Don't add more.
Important For legitimate donors with bad anchor text, send polite anchor-edit requests Most operators never try this. Legitimate editorial donors who used an anchor you don't want will often agree to a swap if you ask plainly — especially if you bring a value-add ("we noticed our citation; we'd love the anchor to read X instead, here's an updated screenshot of the data"). In our experience anchor-edit requests on recent editorial citations reply at far higher rates than removal outreach against PBN-residue donors, but we haven't seen a published benchmark we'd defend — track yours, don't quote ours.
How to do this
Outreach template that lands. Use this on recent (≤12 month-old) editorial citations with a discoverable editor email; cold outreach to a 3-year-old citation through a generic info@ contact form barely outperforms removal requests.
Subject: Quick note on the [Your Brand] citation Hi [Name], Thanks for citing us in your [post title]. Two things: 1) The link currently reads "[current anchor]". If you'd be open to swapping it for "[preferred anchor]", that would help us reflect the post more accurately. 2) Attached is an updated screenshot of the data you cited, in case you want to refresh the figure. Either way, thanks for the mention. [Name]
If you're not the data source, swap step 2 for an alternative value-add: a typo fix, a more recent stat, a broken-link replacement on the same post. Track responses in the same outreach log format used for our disavow outreach.
Critical Don't reach for disavow over anchor patterns alone Disavow is for unnatural-links manual actions or self-confessed paid-link history. Anchor-pattern paranoia without a manual action is not a disavow trigger in 2026 — SpamBrain is already doing the work and disavowing changes nothing. Mueller has called the toxic-link-cleanup industry "a billable waste of time".

How to do this

If you're genuinely considering disavow, run the 4-gate pre-flight in our disavow checklist first. If you don't pass any of the four pre-flight gates there, anchor patterns are not the actionable problem.
Important Rebalance gradually through future acquisitions, not through one-shot fixes If your distribution skews outside the defensible band, the durable fix is acquiring branded + editorial + naked-URL links over the next 6-12 months to shift ratios. A one-shot "anchor cleanup" (mass disavow + mass outreach) looks like a footprint of its own and rarely improves rankings.

How to do this

Target ratio: if you're at 18% exact-match commercial and want to land at 8%, calculate how many new branded / editorial links you need to acquire to dilute the ratio mathematically. Plan against that number, not against the disavow tool.
Nice Document every anchor decision in a single one-page memo The next agency reviewing your portfolio needs the WHY for each call you made. Without a memo, six months from now someone will reverse your audit by accident. The disavow checklist's "decision memo" pattern applies here too.

How to do this

Memo fields: date, property, distribution snapshot (5 bucket percentages), competitive benchmark median, decisions taken (outreach? ignore? rebalance?), expected outcome, recheck date. One page maximum.

Four rules for new link acquisitions so you don't have to clean up a 2018-style footprint two years from now.

Critical Set a target ratio range per bucket, not a single number Single-number targets create rigid plans that miss editorial reality. Ranges absorb the natural volatility of inbound linking. Use pf-04 benchmarks as starting bands; review against the competitive set every quarter.

How to do this

Document the band on each bucket in your link-building brief: e.g. "target band: branded 35-50%, naked URL 15-25%, partial 20-30%, exact match 5-10%, junk ≤10%". Reject acquisition proposals that push any bucket outside its band.
Important Match exact-match anchors to high-authority donors only An exact-match commercial anchor from a high-DR editorial donor in your niche is almost always the kind of link that survives algorithmic devaluation. The same anchor from low-quality donors is what SpamBrain neutralizes. If you're going to use exact match, use it where it's defensible.

How to do this

Rule of thumb: exact-match commercial anchors only on donors with DR ≥40, topical relevance to your niche, and real editorial content (not directory or sponsored listings). Everything else gets branded, naked URL, or partial.
Important Treat naked URL anchors as underrated — they're the editorial-linking signature Naked URLs ("linkguard.ai" as the anchor) are how journalists, bloggers, and editors link in the real world. A portfolio with 15-25% naked URLs reads as editorially earned, not bought. The 2018 playbook treats them as filler; the 2026 mental model treats them as evidence of organic linking.

How to do this

When briefing outreach, explicitly allow naked URL as an anchor option. Many operators force exact-match by default and miss easy diversity wins.
Nice On new sites under 6 months, run a branded-anchor floor of 60-70% New sites that show 15% exact-match commercial anchors in month 4 look manufactured. The honest starting distribution for an organically-launched site is branded + naked URL ≥70% — that's what real editorial coverage produces. Force this in early outreach and you skip the "why does our new site rank for nothing" cycle.

How to do this

For the first 6 months: brief writers, journalists, and outreach partners to default to the brand name or bare URL as the anchor. Reserve exact-match commercial anchors for editorially-earned placements from high-DR donors only.

Three anti-patterns that show up in the post-mortem of nearly every anchor-cleanup mistake.

Critical Don't run a one-month "ratio fix" sprint — it looks like a footprint A sudden burst of outreach removing or swapping 200 anchors in a month is itself an unusual pattern. SpamBrain looks for unusual patterns. A gradual rebalance (ac-04) is durable; a one-month sprint rarely is.

How to do this

If the gap between current and target distribution is large, plan the rebalance across 6-12 months. Track progress quarterly. Avoid touching more than ~15% of the portfolio in any single month.
Important Don't trust vendor toxicity scores on individual anchors Same advice as the disavow checklist: Semrush's Toxicity Score and similar vendor heuristics are internal models, not Google's judgment. They flag anchors that look unusual without knowing if they're unusual in a bad way. Use them as candidate-pattern detection, never as the action trigger.

How to do this

For the full version of this argument with named-source backing (Mueller, Illyes, Shepard, Gabe), see our disavow file checklist — the anti-pattern is identical.
Nice Don't change live anchors on your own pages without a backup record Internal-anchor changes (across your own navigation, sidebar links, internal cross-references) get hit-and-miss treatment in the rush of an audit. Without a backup, you lose the ability to revert if rankings drop on the page you edited.

How to do this

Before changing any internal anchor in bulk, export the current internal-link graph (Screaming Frog or Sitebulb crawl). Keep it for 90 days minimum. If a rebalance drops a specific page in Search Console, you can diff the changes and revert.

About the Author

Andrei

SEO and digital marketing professional with 13+ years of experience. Started as a website administrator in 2011, transitioned to SEO, and achieved top-3 rankings for competitive keywords. Co-founded a consulting firm specializing in marketing audits for companies in Ukraine and internationally. Built LinkGuard to solve the problem he experienced firsthand: most SEO teams purchase links but never monitor their survival. Based in Kyiv, Ukraine.

Anchor text portfolio health checklist (what 2026 data actually shows)

Andrei

Vocabulary, before we start

What success looks like

Pre-flight — get the 2026 mental model right

Audit — see what your portfolio looks like

Action — clean what's worth cleaning

Build — future anchor strategy

What NOT to do

Tags

About the Author

Andrei

Vocabulary, before we start

What success looks like

Tags

About the Author

Andrei

Related Articles

Quarterly backlink audit checklist for agencies: 40 items for 2026

On-page SEO optimization checklist: 35 items for the 2026 audit

Site migration SEO checklist: 37 items for surviving a 2026 domain move