How Bloggent Works
Bloggent uses a deterministic core for the parts that should stay stable and auditable, with language models helping at the edges where explanation and synthesis are useful.
Deterministic core, LLM at the edges
We prefer rules, evidence, and versioned workflow state for scoring, gating, and stored decisions. We use LLMs to explain outputs, draft structured prose, and help summarize complex evidence — not to silently replace the underlying mechanics.
Model cards
These cards describe what each capability is for, which signals influence it, and the limitations users should understand before acting on the output.
Opportunity Scoring
- What it does
- Ranks keywords by relative upside so teams know what to prioritize first.
- Signals it uses
- Keyword metrics, SERP composition, intent clarity, topic fit, and competitive evidence.
- What the AI does
- Helps summarize why a score looks attractive or risky in plain language.
- What the AI does not do
- It does not override the deterministic score or invent missing evidence.
- Honest limitations
- A strong score is a prioritization hint, not a guarantee of rankings or traffic growth.
Intent Classification
- What it does
- Assigns a likely search intent so content types and calls to action match the query.
- Signals it uses
- SERP features, ranking page formats, query wording, and supporting evidence from collected results.
- What the AI does
- Helps explain borderline cases and describe why the SERP looks mixed.
- What the AI does not do
- It does not classify intent without a collected SERP or silently change the saved label.
- Honest limitations
- Mixed-intent SERPs can remain ambiguous, especially when the query supports several valid outcomes.
Topic Clustering
- What it does
- Groups related keywords into page-level targets and broader hub structures.
- Signals it uses
- Shared SERP overlap, semantic similarity, intent compatibility, and structural fit within a strategy.
- What the AI does
- Helps draft human-readable names and rationale for clusters once grouping is complete.
- What the AI does not do
- It does not merge clusters simply because two phrases look similar on the surface.
- Honest limitations
- Clusters improve planning, but editorial judgment still matters for edge cases and brand positioning.
Entity Coverage
- What it does
- Highlights the entities, concepts, and competitor themes repeatedly associated with a topic.
- Signals it uses
- SERP evidence, extracted entities, recurring sources, and coverage patterns across top results.
- What the AI does
- Helps summarize coverage gaps and explain why certain entities matter for the brief.
- What the AI does not do
- It does not claim entity relevance without supporting evidence or treat frequency as truth.
- Honest limitations
- Entity patterns can lag behind market changes and should be interpreted alongside fresh SERP data.
Content Briefs
- What it does
- Turns research and strategy into a structured brief for writers and editors.
- Signals it uses
- Opportunity priorities, intent, cluster targets, entities, competitor evidence, and workflow context.
- What the AI does
- Helps phrase the brief clearly, organize sections, and surface guidance a writer can act on quickly.
- What the AI does not do
- It does not invent facts, final claims, or hidden requirements absent from the evidence set.
- Honest limitations
- A brief can accelerate production, but publishing quality still depends on editorial review and verification.
Never publish
We will not publish exact weights, thresholds, prompts, or calibration data. Those are sensitive implementation details, not user-facing methodology.
We don't train on your data
Your workspace data is used to run your requested workflows and store outputs. We do not use your private content, keywords, or project structure to train foundation models.
Version history
v1.0 — July 2026: Initial methodology publication.