Free SEO Audit Run free scan
Updated for Google's June 24–26, 2026 Spam Update

Know your site's AI Spam Score before Google finds it

Google's June 2026 spam update demoted sites built on scaled AI content — globally, in every language. PureRank crawls your site, runs 40+ forensic signals and tells you exactly how machine-made your content looks, and how to fix it.

Free · No signup · Content-first sampling: editorial pages get ~65% of the crawl budget · ~1–2 minutes

40+forensic signals
7risk dimensions
230+AI language markers
0–100calibrated AI score

Initializing…

1Discover
2Crawl
3Linguistics
4Cross-page
5Trust
6Score
Detection engine

Seven dimensions. One honest score.

The same families of signals search engines use to separate expertise from generated filler.

🔤

AI Language Markers

230+ curated stock phrases, giveaway artifacts ("as an AI language model…") and LLM-favored vocabulary, measured per 1,000 words.

📈

Stylometric Fingerprint

Sentence burstiness, rhythm uniformity, hapax legomena and type-token ratios — statistical tells that survive light paraphrasing.

🧱

Formatting Patterns

Bold-lead bullet walls, em-dash density, triadic constructions, colon-headings and canned "Conclusion" outros typical of raw model output.

🧬

Originality & Templating

Shingle-based cross-page similarity catches doorway variants and mass-templated pages before Google's deduplication does.

Publishing Velocity

Sitemap forensics: pages-per-day bursts and index bloat — the exact footprint of "scaled content abuse" named in Google's spam policy.

🛡️

Trust & E-E-A-T Surface

Authors, dates, contactability, legal pages, citations and schema.org — the context that decides whether AI-assisted content survives.

🎯

AI-Search Manipulation

Prompt-injection strings aimed at AI Overviews, hidden text blocks and self-serving "Top N" listicles — a named spam violation since May 15, 2026.

Workflow

From URL to action plan in a minute

Point

Enter any domain — yours or a competitor's. No signup, no tracking snippet.

Crawl

We fetch robots.txt, sitemaps and up to 12 representative pages across site sections.

Analyze

40+ signals across 7 dimensions are computed and weighted into a calibrated 0–100 score.

Fix

You get page-level verdicts and a prioritized remediation list mapped to Google's spam policies.

Why now

The June 2026 Spam Update changed the math

On June 24, 2026 Google began rolling out a global spam update, completing it by June 26 (per Google's Search Status Dashboard). It sharpened automated detection of spam-policy violations — with scaled content abuse (mass-produced, low-value pages, typically AI-generated) squarely in scope.

Weeks earlier, on May 15, 2026, Google rewrote its spam policies to explicitly name manipulation of AI search responses (AI Overviews / AI Mode) as spam — the first time AI manipulation was written into policy.

AI-written text is not penalized by itself. Mass-produced, unhelpful, trust-free content is. PureRank measures exactly that gap.

  • May 15, 2026Spam policy rewriteManipulating generative AI responses in Search is named a violation for the first time.
  • Jun 24, 2026Spam update beginsGlobal rollout starts ~noon ET, all languages, Search + AI Overviews + AI Mode.
  • Jun 26, 2026Rollout completesSites violating spam policies "rank lower or don't appear in results at all".
  • NowPeriodic refreshesGoogle refreshes spam systems continuously; recovery takes months. Audit before the next pass.
Free API + MCP

Give your publishing agent a spam-risk gate

Agentic pipelines ship content at a pace no human reviews — and one bad habit can sink the whole domain. PureRank is the scanner, not another generator: our open-source MCP server lets the agent score every draft before it publishes, and ship only what comes back pass.

claude mcp add purerank --env PURERANK_API_KEY=KEY -- npx -y purerank-mcp-server

Free key: 25 draft checks + 5 site scans a day — enough to gate a real cadence. No card, no crawl needed for drafts, plain REST if you don't use MCP.

Docs & examples →  ·  GitHub (MIT)  ·  npm

agent> purerank_score_draft("10 SEO Best Practices…")
← 94/100 · FAIL — Likely AI-generated
33.8 AI stock phrases /1k · uniform rhythm
agent> purerank_score_draft("Postgres war story…")
← 2/100 · PASS — Likely human
rule> publish only if gate == "pass"

Real, unedited API responses. Every signal is explained in the report — methodology is open, benchmark is published.

Pricing

Start free. Scale when it matters.

Monthly Annual · save 17%

Scout

$0 / forever
Free — no card
  • 10 scans / day
  • 12 pages per scan
  • Full 7-dimension report
  • API + MCP — 25 draft checks, 5 scans/day
  • Public share links + badge
Run a free scan
FIRST ONE −50%

Deep Audit

$49 one-time
One-off — no subscription
  • Full-coverage crawl (300 pages)
  • Detailed PDF + fix plan
  • Domain history (Wayback)
  • Private report
  • Code FIRSTAUDIT50 → $24.50

Agency

$149 / month
For teams & resellers
  • 1,000 scans / day
  • 300 pages per scan
  • 25 client sites, white-label PDF
  • Competitor benchmarking
  • API — 1,000 scans + 2,000 draft checks/day, 5 keys
FAQ

Straight answers

Does Google penalize AI content?

Not per se. Google's stance is "reward high-quality content, however it is produced". What gets demoted is scaled content abuse: mass-produced pages with no added value, expertise or trust — which in 2026 overwhelmingly means unedited AI output. The June 2026 spam update tightened exactly that detection.

How accurate is the AI Spam Score?

It is a probabilistic estimate built from 40+ public, explainable signals — not a black box and not a Google metric. No detector can prove authorship; short or heavily edited texts are genuinely ambiguous. Use the score to prioritize a human review, and read the per-signal breakdown rather than the single number.

What exactly do you crawl?

robots.txt, your sitemaps, and up to 12 representative pages sampled across site sections (content pages are prioritized). We fetch politely with an identified user-agent, ~2 requests/second max, and never execute your JavaScript.

Can I scan a competitor?

Yes — any publicly reachable site. All signals are computed from public data only: HTML, sitemaps and robots.txt. Nothing is bypassed and nothing private is accessed.

My score is high. Am I doomed?

No. The report tells you which dimension drives the risk: boilerplate phrasing, templated pages, publishing bursts or a weak trust surface. Sites recover by consolidating thin pages, adding real expertise and provenance (authors, dates, sources) — typically over one to several months, per Google's own guidance.

Do you store my content?

Only the computed report (scores, signals, URLs) is stored locally so you can share a report link. Raw page HTML is discarded after analysis.