The backlinks agency, evolved.

You hired an agency to move rankings.We engineer the signals that actually move them.

The authority infrastructure that determines who Google ranks and who AI recommends. Off-page and on-page, for both surfaces.

Co-citation graph · live sample 127 core / ~3,000 cited
127 In the authority bubble
~3,000 Domains cited overall
10-20% Citations that recur

AI cites the most connected brand, not the loudest. Hover a node to inspect its cluster.

We trained teams from

  • Accor Hotels
  • Pepsi
  • Australian Broadcasting Corporation
  • Audi
  • BMW
  • Coca-Cola

01. The problem

Google and AI is a data science problem. Most agencies are not data science companies. We are.

The link building model everyone is still selling stopped working in 2021. The 2024 Google API leak confirmed it.

Most agencies sell you a Domain Rating number and call it strategy. They are working in two dimensions. Google works in three.

We engineer 3D vector authority and harmonic centrality. Keyword-relevant links with real traffic, ranked pages, and topical coherence. This is what moves rankings now.

This is what WLDM builds.

02. The evidence

The signals Google actually uses

In May 2024, Google's Content Warehouse API was publicly leaked and independently documented, including by Rand Fishkin and Mike King. It exposed the internal field names from Google's production ranking pipeline. These are not theory or SEO folklore. They are the actual variables the system checks.

Content Warehouse API · topic embeddings

  • siteEmbeddingSite topic centroid
  • pageEmbeddingOne vector per page
  • siteRadiusDrift from the centroid
  • siteFocusScoreSticks to one subject

Topical focus

siteFocusScore

How closely a linking site sticks to a single subject. A site publishing across 40 unrelated topics passes less authority than one that has stayed in its lane.

Weight in WLDM's scoring model0.30

Vector semantics

siteEmbedding / pageEmbedding

Vector representations of the linking site and page. Semantic signals, not keyword matches.

Weight in WLDM's scoring model0.30

Highest weight

context2

Google does not just read your anchor text. It reads the paragraph around the link, hashed into a kind of fingerprint. The paragraph context outweighs the anchor text.

Weight in WLDM's scoring model0.50

Quality tier

sourceType

Google sorts every linking page into one of three buckets: HIGH, MEDIUM or LOW. Domain Rating is not one of them.

Tiers in the leaked fieldHIGH · MEDIUM · LOW

Penalty

anchorMismatchDemotion

If your anchor text says one thing and the page it links to is about another, Google penalises that link.

Anchor weight in WLDM's scoring model0.20

Topic drift

siteRadius

How far individual pages drift from a site's topic centroid. High drift means lower authority weight.

Effect of high driftLower authority weight

The leak confirms these fields exist and are active in the pipeline. It does not publish Google's weights. What we have built at WLDM is a scoring model against the fields that are actually documented, not a guess dressed up as a formula.

How WLDM engineers against them

score = 0.50 × cos(context_paragraph, target)      ← context2
      + 0.30 × cos(referring_page_body, target)    ← siteFocusScore / siteEmbedding
      + 0.20 × cos(anchor_text, target)            ← anchor match

We score every potential link from 0 to 1 with the Cosine Scorer.
A link at 0.6 passes 3× the authority of one at 0.2, from the exact same domain. Below 0.2 we reject the placement, no exceptions.

We also check every linking page for real, human traffic before we ever reach out. That part is not guesswork either. It is grounded in the Reasonable Surfer Patent (US 7,716,225 B1), Google's own model for weighting a link by how likely a real person is to click it. Page-level traffic. Not domain traffic. Not DR.

Don't take our word for it. The Cosine Scorer runs entirely in your browser, on the same model behind our audits.

Score your real links

03. Proof

From below rank 30 to #1 for “Online Casino.”
10 months. ML-driven. No shortcuts.

Ranking movement · “Online Casino”

#1 #10 #20 #30+ MONTH 0 · BELOW RANK 30 MONTH 10 · #1

The keyword

#1

Final position, from below rank 30, in 10 months.

Monthly searches85,000
Keyword difficulty95
Keyword traffic value$549,000 / month
Referring domains to compete in the top 10~1,400
Non-brand organic growth2x in 90 days

“Online Casino” pulls 85,000 monthly searches, a keyword difficulty of 95, and $549,000 in monthly traffic value. Competing for a top 10 spot takes roughly 1,400 referring domains.

This is not a keyword you stumble into by publishing more blog posts. This is a keyword you engineer your way to.

“Within 90 days of engagement, WLDM's strategic link acquisition efforts helped us secure #1 positions for several ultra-competitive keywords and doubled our non-brand organic traffic.”

Peter Macinkovic, SEO Lead, Stake.com

04. What we build

Six services. One authority infrastructure.

Google ranking and AI citation probability are different problems, even though most agencies still treat them as one. WLDM builds for both, separately, from the content on your page to the links pointing at it.

Open research

Research

We do not ask you to trust the claims on this page. Every one of them comes out of a programme in the lab, and we publish the corpus, the method, the figures and the limits, in enough detail that a technical SEO can pull it apart line by line.

Read the research

05. How we work

We start with your data. Not a pitch deck.

01

Book a strategy session

One session. Your authority gaps mapped. No pitch deck.

02

Get a strategy built on your data

Built for your competitive position. No generic recommendations.

03

Start with a 3-month project

Not a retainer. Defined outcomes. Stake.com doubled traffic in 90 days.

06. Research

Built on research no other agency has

After 11 million citations, one finding stood out: AI cites the most connected brand, not the loudest.

Only 10-20% of AI citations form a stable, recurring cluster. The rest cycle in and out. In one client's blackjack query cluster, the same 127 domains out of approximately 3,000 cited showed up in nearly every answer we tracked. Brie named this the authority bubble.

Google's results and ChatGPT's citations barely overlap. Our own research puts the overlap at approximately 12%. Ranking number one on Google tells you almost nothing about whether you exist inside AI answers.

AI citations research

Over 2,000 hours. August to November 2024, with DataForSEO, across ChatGPT and Gemini, run separately.

Link-building database

600,000 domains tracked. Roughly half a million vectorised for semantic scoring.

AI citation coverage

7-8% hit rate on a random sample of real AI citations, against an expected baseline of around 1%.

ML backlink classifiers

Five models. 18,000 training domains curated by hand. 99%+ accuracy. Three of five must agree.

Proprietary toolkit

Nine tools. Two are free and run in your browser. Seven stay internal.

AI citations outreach

As far as we know, nobody else does this. Scored against the co-citation graph and Harmonic Centrality, not the link graph.

ClassifierAccuracy
Logistic Regression98.97%
Random Forest99.08%
SVM98.74%
MLP Neural Network99.31%
XGBoost99.31%

No placement proceeds unless at least three of the five models agree.

Read the methodology in full

The people who bought the methodology

2× exit in two years
“We ended up exiting the business in two years for twice what we'd paid, and WLDM was a big part of our success.”
Lucjan Suski Lucjan SuskiCo-founder, Surfer SEO
#1 for “Online Casino”
“Within 90 days of engagement, WLDM's strategic link acquisition efforts helped us secure #1 positions for several ultra-competitive keywords and doubled our non-brand organic traffic.”
Peter Macinkovic Peter MacinkovicSEO Lead, Stake.com
Enterprise-grade requirements met
“We are satisfied with the services and the quality, as the agency was able to meet our very high and exclusive requirements.”
Dominykas Damalakas Dominykas DamalakasNordVPN
AI training data coverage
“WLDM is one of the few teams building a real methodology around AI training data coverage.”
Stephen Burns Stephen BurnsCommon Crawl

08. Public tools

We built these to measure what most agencies can't see.

Free, browser-local, and running the same models as our internal pipeline. Two of nine. The other seven stay in the lab.

Public tool 01

AI Citations Audit

Enter your domain and we force live ChatGPT searches across your target queries. No cached results, no stale data dressed up as current.

  • Reveals the citation clusters your competitors are already inside, and you are not
  • Outputs your authority tier and a prioritised placement plan
Run Your AI Citations Audit

Public tool 02

Backlink Relevance Cosine Scorer

Paste two URLs. We score how closely they actually match, on the same three signals Google's own pipeline checks.

  • Flags the weak links in your profile before they cost you rankings
  • Runs entirely in your browser. Nothing about your links leaves your device.
Score Your Backlinks

Brie Moreau · Founder

Brie Moreau

If you've read this far,
you probably saw me speak.

Good. That means you already know the methodology is real.

I've spent nearly twenty years in SEO and data science. I built WLDM because I got tired of watching good companies get sold reports instead of results. The research lab, the ML classifiers, the 11 million citation study. That is how we actually work.

If something from the talk connected with what you're dealing with, let's talk it through. No pitch deck. Just your data.

WLDM works with a limited number of enterprise partners. If your goals are serious, this is the conversation to start.

Search authority is not a strategy. It is infrastructure. WLDM builds it.

Start with the free audit. It scores your backlink profile and your AI citation position, and it costs you a domain name.