The Intelligence Report: A Grok-Style Editorial Layer for Every Audit
Raw scores tell you what happened. The Intelligence Report tells you what it means. A six-section editorial narrative — generated by AI, grounded in data — that reads like a senior analyst's assessment of your site's AI retrieval posture.
A dashboard full of scores is not a decision surface. Twelve numbers — each meaningful in isolation — do not tell you what your site's overall AI retrieval posture is, where the compound risk lives, or what to do first. The Intelligence Report exists to close that gap. It is a Grok-style editorial audit narrative: six named sections, each with a subscale rating and analytical prose, synthesized from all twelve module scores and the top audit issues into a single, readable assessment.
What the Intelligence Report Contains
The report opens with a Signal Summary — a paragraph-length orienting statement about the site's overall AI retrieval position, written in plain language without jargon. This is the answer to "what is the single thing I need to understand about this audit?" It names the dominant theme: whether the site is technically strong but semantically weak, entity-rich but trust-deficient, well-structured but absent from recommendation surfaces.
- Signal Summary — the dominant finding in plain language, one paragraph
- SEO Posture — technical foundation health, scored X.X/10, with three named strengths and two critical gaps
- AI Visibility — extractability, entity clarity, and retrieval readiness assessment
- Machine Trust — trust formation and decay analysis, contradiction detection summary
- Entity Health — primary entity clarity, consistency, external validation depth
- Strategic Verdict — composite score, prioritized recommendation, and competitive outlook
How It Is Generated
The Intelligence Report is generated at the end of each audit via the executive-summary API route. It assembles a structured context object from all module scores, the top twelve issues sorted by severity, entity data, retrieval simulation results, Machine Trust sub-scores, and Recommendation Surface probabilities. This context is passed to the AI model as a richly formatted prompt via the executiveSummaryPrompt template — a 2000-token generation call that produces six named sections, each with an X.X/10 score and 150–250 words of analytical prose.
●The Intelligence Report is cached in Redis for 24 hours. If you improve your site and run a new audit, the report is regenerated fresh from the new data. The cache is keyed by audit ID and model version — a new audit always produces a new report.
The Token Problem We Solved
Generating a structured JSON response with six sections, subscores, and 1200+ words of prose requires significantly more than 1024 tokens. Our initial AI client hardcoded a 1024-token limit — a setting appropriate for focused scoring tasks but fatal for comprehensive narratives. The response would be truncated mid-JSON, causing a parse error and a 503 response to the user. We resolved this by making the maxTokens parameter explicit in callAI, with the executive summary route passing 3000 tokens. The limit is now caller-specific, not global.
Reading the Strategic Verdict
The Strategic Verdict section is the highest-value output. It contains the composite Machine Trust Intelligence Score (a weighted composite of all twelve module scores), a plain-language statement of the single most important action, and a brief competitive outlook — where the site stands relative to typical domains in its category and what the trajectory looks like if nothing changes. The verdict is deliberately opinionated: it names a single priority, not a list. If you need the full ranked list, the Fix Plan and Decision Roadmap expand on it.
Run a full audit and read the Intelligence Report for your domain — 12 scores synthesized into one editorial assessment.
Get your Intelligence Report