NotebookLM Customizer Brief

Editorial data-diagnostic style. Ink-black (#0d0d0d) background, warm paper (#f5f2ed) text and cards. Terracotta accent (#c8431a) for key numbers and labels. Three-layer color code throughout: blue for Human/Layer 1, amber for Search/Layer 2, emerald green for AI-agent/Layer 3 — never lime green. Serif display headlines, clean sans body text, monospace for all scores and metrics, like a terminal readout. No stock photography, no decorative illustration, no gradients. Scorecards, horizontal score bars, and status icons (checkmark/warning/critical) as the only visual devices. Restrained, precise, analyst-grade — not playful, not whiteboard-style.Copy

Diagnose, don't summarize. Open with the total score and name which of the three layers (Human, Search, AI-agent) is weakest and why that matters. If the same underlying issue appears as a finding in more than one layer, treat that as the central throughline of the video — one root cause with multiple symptoms is more valuable than three separate facts. Move quickly through the two stronger layers, naming only their single most important metric each. Spend the most time on the weakest layer's critical findings, especially anything related to schema/structured-data coverage or missing AI-readability files (llms.txt), since these determine whether AI systems can cite the site at all. Close by stating the combined business impact of fixing the top 1-2 issues together, not each issue separately, and end with one concrete next action. Voice: analytical and evidence-first throughout, shifting to a decisive, recommendation-driving tone only in the final closing statement.Copy

Two voices carry this diagnosis, never more than two. ULF — carries the diagnosis (opening through the weakest-layer deep dive). Analytical, precise, data-first. Trigger phrasing: "The data shows...", "The structural finding is...". Never uses exclamation marks, enthusiasm, or opinion without evidence. LAIA — closes with business impact and the next action (final segment only). Strategic, direct, decisive. Trigger phrasing: "The recommendation is...", "The move here is...". Never hedges, never uses passive voice, never adds qualifiers after stating the recommendation. Handoff rule: Ulf never states business impact or next steps. Laia never restates metrics or scores — she only translates the strongest finding Ulf already established into one sentence of consequence and one concrete action.Copy

Build this as a diagnosis, not a data dump. Slide 1: state the total score and name which of the three layers — Human, Search, or AI-agent — is weakest. That gap is the hook. Slide 2: show all three layer scores side by side so the gap is visible at a glance. Slide 3: if the same underlying issue appears as a finding in more than one layer, give it its own slide as the throughline of the whole deck — one root cause with multiple symptoms is the single most important idea to land, more valuable than any individual metric. Then one slide per layer, ordered strongest to weakest, each showing only its top 3-4 metrics with pass/warning/critical status — never all metrics on one slide. Give the weakest layer the most detail, especially anything about structured-data/schema coverage or missing AI-readability files, since these determine whether AI systems can cite the site at all. Close with two slides: a prioritized action list, maximum three items, ranked by combined impact across layers rather than by layer — and a business-impact slide stating the compounded effect of fixing the top issues together, ending on one concrete next action. Audience: technical and marketing decision-makers who need to act on this, not general readers. Style: restrained, analyst-grade, data-forward. No stock imagery, no filler or transition slides.Copy