$ cat The 2024 Google API Leak
Google Content Warehouse documentation: authenticity, evidence provenance, leaked fields, researcher interpretations and limits on ranking conclusions.
Analysts published accounts of internal Google Content Warehouse documentation in May 2024. Google subsequently acknowledged the documents, while warning against decontextualized conclusions. This chapter distinguishes the reported fields from analyst interpretation and from separate public court evidence. The selected field audit below now pins a public package, checksum, module and line context. It does not claim an exhaustive module/attribute census. [1][2][3]
1 · What the leak is
The documents are automatically-generated reference material for an internal Google API called the Content Warehouse. They list modules, features, and attributes (field names, types, and short descriptions) that Google's systems use to represent documents, links, users, and search interactions. They were briefly exposed in a public code repository, discovered, and shared with SEO analyst Mike King (iPullRank) and marketer Rand Fishkin (SparkToro), who published them in late May 2024. leak-indicated[1][2]
This is the crucial point that makes the leak citable. Google confirmed to journalists that the documentation is authentic, while cautioning against "making inaccurate assumptions" because the material may be out of context, outdated, or incomplete. confirmed[3] Authenticity of the documents does not independently verify this archive’s transcription of every field or its current use in ranking.
2 · What the documents actually contain
This first table retains the analysts’ broader inventory. The versioned audit that follows independently traces seven selected names or strings; other entries remain secondary-source descriptions. A field name alone establishes neither production deployment nor a ranking weight, threshold or causal mechanism.
| Area | Representative attributes described | Note |
|---|---|---|
| Click / engagement (NavBoost) | goodClicks, badClicks, lastLongestClicks, squashed vs unsquashed clicks, impressions | Click-related names in generated documentation; production use needs separate evidence. |
| Chrome data | attributes referencing Chrome view counts (e.g. chrome_trans_clicks) | A Chrome-related string; collection and current ranking use are unresolved. |
| Site-level authority | siteAuthority | A reported field name; it is not equivalent to Moz Domain Authority or Ahrefs Domain Rating. |
| Links | tiered link indexes (low/medium/high), sourceType, multiple PageRank fields (rawPagerank, firstCoveragePageRank), anchor data | The selected sourceType comment describes tiers; traffic gating remains an inference. |
| Freshness / age | hostAge, dates on first coverage | Age-related fields; their names alone do not prove a universal new-site suppression period. |
| Content / quality | fields referencing an OriginalContentScore, title-match, and Panda-like signals | Analyst-reported score names; not independently traced in the selected audit. |
| Authors | isAuthor, author fields | Analyst-reported authorship fields; current storage and use are not established here. |
Versioned field audit · completed September 17, 2026
Artifact: google_api_content_warehouse 0.4.0, registry release dated March 28, 2024. The downloaded package matches the SHA-256 returned by the release API: c091e5e2a315a163295a9b1de95619d4bc254f8c57f094fa86dd72975391e0be. This binds the audit to a retrievable version; agreement with the registry is not an independent cryptographic signature.
| Selected name or string | Unmodified module download | What its context permits |
|---|---|---|
goodClicks (line 29), badClicks (line 24), lastLongestClicks (line 32) | QualityNavboostCrapsCrapsData | Names and documented types are present. This does not identify a current click classifier or purchasable ranking effect. |
siteAuthority (line 35) | CompressedQualitySignals | The comment mentions conversion from SiteAuthority and Qstar. It does not define third-party domain-authority scores or supply current weights. |
sourceType (line 30) | AnchorsAnchor | The comment distinguishes source quality tiers and says the relationship with index tier is not identical. It does not establish a universal traffic threshold. |
hostAge (line 113) | PerDocData | The comment refers to earliest first-seen dates and fresh spam. It is not a universal new-site waiting-period specification. |
chrome_trans_clicks (line 35) | QualitySitemapTargetGroup | The string appears in the topUrl description, not as a field declared chrome_trans_clicks in this module. Its collection and present ranking use are not established. |
Reproduce the trace using the file hashes and line manifest, captured release response and provenance notes. The five supplied files retain their copyright headers and Apache 2.0 license. No Google endpoint was queried, and no package code was executed. The court record, leaked schema, Google’s acknowledgment and analysts’ conclusions remain distinct evidence layers.
3 · What researchers infer (and what is disputed)
Everything in this section is interpretation. It is the analysts' reading of the fields above, not Google policy, and some of it is contested even within the SEO community.[4]
- Clicks influence ranking via NavBoost. Public 2023 antitrust testimony separately describes NavBoost’s use of click information. The meaning and current production use of each leaked field remain a different question. [5]
- Links are gated into quality tiers by traffic. King reads the tiered link index as meaning links on pages with no clicks are shunted to a low-quality tier and effectively ignored. This is the interpretation the PBN chapter leans on, always as inference. interpretation
- A site-wide authority score exists. The presence of
siteAuthorityis read as contradicting Google's public "we don't have domain authority" messaging, though its weight in ranking is unknown. interpretation - New sites are dampened ("sandbox"). Age fields are read as supporting a long-suspected new-site suppression period. interpretation
- Anchor text is less central than assumed. A widely-cited but debated reading. disputed
A stored attribute is not a confirmed ranking factor. The docs do not include weights, do not say which fields are live, and Google explicitly warned they may be outdated or lack context. confirmed[3] A ranking-use claim needs context and corroboration beyond the existence of a stored attribute.
4 · Why it matters for black hat SEO
The documents suggest research questions about click information, links and site-level signals. They do not demonstrate that purchased clicks, trafficless links or any other tactic succeeds. To evaluate a hypothesis, preserve the exact schema/module/version, quote its relevant context, identify the analyst’s inference, and compare it with independently measured outcomes. This archive supplies a versioned field trace, secondary analyses and proposed studies. It still lacks current production-use evidence or a causal validation of those ranking hypotheses.
5 · References
- An Anonymous Source Shared Thousands of Leaked Google Search API Documents with Me. Rand Fishkin, SparkToro, 2024. secondary reporting / analyst interpretation
- Secrets from the Algorithm: Google Search's Internal Engineering Documentation Has Leaked. Mike King, iPullRank, 2024. interpretation
- Google confirms the leaked Search documents are real. The Verge, May 29, 2024; reporting Google’s statement. This replaces a May 28 article that predated confirmation.
- 2024 Google Search documentation leak. Wikipedia (overview, context, and disputes).
- US v. Google — public September 2023 trial testimony discussing NavBoost and clicks. primary
Other chapters cite this page whenever they lean on the leak. Any claim traced here is, at strongest, leak-indicated (the field exists) or interpretation (what it means), never a confirmed Google ranking rule.