RankEngineRankEngine Help

Generating blog posts

RankEngine's blog generator is built for the AI search era — content that's hard to detect as AI-written, free of plagiarism, and optimized for both Google and AI citation.

Pipeline overview

keyword input
   └─► research (RankEngine related keywords + SERP titles)
        └─► outline (RankEngine AI)
             └─► draft (RankEngine AI with anti-detection prompts)
                  └─► sanitize (strip AI signposting + generic adjectives)
                       └─► quality gate (3 categories, 90/100 pass threshold)
                       └─► plagiarism check (5-layer stack — see below)
                            └─► retry up to 2× if fails
                                 └─► finalized draft + image
                                      └─► save to Calendar (status=draft)

Anti-AI-detection (phases 63-67)

Blog generation enforces 4 techniques across 3 layers (prompt + sanitizer + quality gate):

  1. Burstiness — varied sentence length (CV >= 0.2). Short decisive sentences interleaved with longer ones.
  2. No signposting — bans 19+ phrases like "additionally," "consequently," "in particular."
  3. Mechanism over attribute — "the formulation chelates iron" beats "this is an effective formulation."
  4. Analytical voice — first-person hedging ("we notice…", "the data suggests…") instead of declarative third-person.

Target: ZeroGPT AI-probability < 30% on every generated post.

Plagiarism prevention (phases 68-72)

5 independent layers:

  1. Prompt — RankEngine AI is told to paraphrase any source data, never quote.
  2. Source corpus — capped at 50 strings / 5000 chars to prevent over-fitting.
  3. N-gram overlap — 7-word verbatim matches between draft and sources flagged.
  4. Semantic similarity — RankEngine embeddings; cosine > 0.85 between any sentence pair flagged.
  5. External Originality.ai (Pro plan, optional) — third-party plagiarism score, must be > 80% original.

If any layer flags an issue, the pipeline retries with targeted instructions. Max 2 retries.

Quality gate

3 new 2-pt categories on top of the standard rubric:

  • Burstiness — sentence length variation
  • Mechanism Density — concrete > abstract
  • Analytical Voice — first-person hedging

Total max score 100, pass threshold 90. Failed drafts retry; final draft below 90 saves as draft + flagged for manual review.

Usage

RankEngine shows the action before generation starts. Wallet usage is recorded only for successful generation and failed retries do not double-count the task.

Related