Report Style Showcase
Evidence-based report components for AI-search, SEO/GEO, and operational audit exports.
By Marcus Quinn.
Evidence-based report components for AI-search, SEO/GEO, and operational audit exports.
By Marcus Quinn.
Internal toolkit · May 2026 · v4
Component stress-test for DESIGN.md-backed report styles. Render this same Markdown through different templates to compare typography, palette, spacing, borders, cards, tables, badges, code blocks, and print profiles.
34
Renderer templates.
16:9
Slide PDF profile.
A4
Document PDF profile.
MD
Canonical source.
This showcase demonstrates the report components expected in the LLM Visibility Toolbox examples: a non-numbered executive summary, numbered H2 chapters, plain H3 subheadings, sources grouped by type, badge keys, checklists, code/block templates, callouts, accordions, and a version summary.
Peer-Review Peer-reviewed paper or controlled comparison with documented methodology.
Strong Large primary-data analysis from an independent source.
Vendor Vendor-published study with methodology, but commercial conflict of interest.
Practitioner Practitioner report or anecdote, often without a control group.
Hygiene Technical baseline that is not experimentally measured but follows from how the bots work.
Schema was downgraded after controlled evidence; engine divergence is now structural; source sections are grouped by evidence type; reports keep Mermaid and LaTeX as readable fallbacks unless local renderers pre-render them.
Action: review the numbered chapters below and compare the generated HTML against the source Toolbox patterns.
Reference this report action: review the numbered chapters below and compare the generated HTML against the source Toolbox patterns.
Guide me through the tools, resources, accounts, permissions, source material, and access needed to take this action. Break the work into numbered steps, call out any missing inputs before execution, include safe handling for credentials or confidential data, and finish with verification evidence I can capture.Use anchor links, appendix links, numbered steps, accordions, coloured panels, and source cards in the same canonical Markdown.
Evidence:Verified Evidence:Partial Evidence:Inferred Evidence:Missing
Evidence values should read as plain Evidence: text followed by a colour-coded mini badge for the value only.
Use plain narrative and bullets when that is clearer than a panel. Panels are reserved for warnings, action blocks, source cards, or high-emphasis evidence.
Six converging studies point to third-party mentions as a stronger AI visibility signal than isolated owned-page edits.
Action: coordinate one trade article, one community thread, one video transcript, and one partner citation within the same quarter.
Reference this report action: coordinate one trade article, one community thread, one video transcript, and one partner citation within the same quarter.
Guide me through the tools, resources, accounts, permissions, source material, and access needed to take this action. Break the work into numbered steps, call out any missing inputs before execution, include safe handling for credentials or confidential data, and finish with verification evidence I can capture.A plain bullet section should remain plain:
| Component | Stress condition | Expected result |
|---|---|---|
| Evidence badge | Long table cell with badge Evidence:Verified | Badge stays readable and does not split words. |
| Facts table | Multiple columns with prose | Table remains usable in HTML and constrained in print. |
| Source card | Evidence note near claim | Card is visually distinct from normal paragraphs. |
| Sidebar | Many headings | Sticky TOC remains secondary to content and active link updates. |
Use info panels for caveats, assumptions, and reading guidance that should not become recommendations.
Why this changes the plan: a high-impact finding changes sequencing, budget, owner, or acceptance criteria.
Next action: turn the finding into a concrete implementation step with owner, proof path, and rerun command.
Reference this report action: turn the finding into a concrete implementation step with owner, proof path, and rerun command.
Guide me through the tools, resources, accounts, permissions, source material, and access needed to take this action. Break the work into numbered steps, call out any missing inputs before execution, include safe handling for credentials or confidential data, and finish with verification evidence I can capture.Crawlable claims, source IDs, direct answers, and page-type weighting.
Unsupported claims, hidden content, generic tactics, and missing verification.
Every page needs the same GEO checklist.
Page type determines which tactics are useful, conditional, or noise.
Every tactic carries a badge. Use RCT/academic for controlled research, strong primary data for large independent data, vendor study where methodology exists but incentives are commercial, practitioner for field evidence, and hygiene for baseline technical work.
A visible uplift in one engine is not proof of universal AI visibility. Keep AIO, Gemini, ChatGPT, AI Mode, and Perplexity separate until the closing synthesis.
Quotes highlight expert evidence, user language, or a decision constraint without turning it into a recommendation.
Written by Dr. Jane Doe, PhD
Principal Data Scientist, ExampleCo
Jane led the data platform team from 2019 to 2024 and now researches LLM retrieval.
LinkedIn | Google Scholar | Personal site.agents/scripts/report-render-helper.sh render report.md --template lottiefiles --pdf-profile slides-16-9-2 --output report.htmlflowchart LR
SourceLedger --> Finding
Finding --> Recommendation
Recommendation --> Verification\text{LLM visibility} = \alpha citations + \beta mentions + \gamma retrieval - \delta decayInline LaTeX fallback: {{latex:\text{visibility} = \alpha citations + \beta mentions + \gamma retrieval}}.
Authority work belongs outside the site as much as on it: profile parity, trusted third-party mentions, practitioner credentials, and community proof all support retrieval and citation decisions.
Claims that circulate widely but cannot be traced to a primary source, or are contradicted by controlled evidence, should be called out explicitly.
“Schema markup alone creates citation uplift.” Contradicted by controlled or near-controlled studies; ship schema as hygiene.
“Longer content always gets cited more.” Engine-dependent; content depth helps only when it improves answer density and source usefulness.
Review spacing, table width, no-wrap badges, active TOC highlighting, print CSS, and component contrast.
dateModified in schema.AI is now a discovery layer, but engines disagree on sources. Tracking only mentions or only citations misses the retrieval gap. Keep the final synthesis short, source-backed, and tied to the next action.
Use the same Markdown-first report structure for one-off audits, monthly retainers, lead magnets, and routine handoffs. Keep deterministic evidence collection separate from interpretation.
Result: monthly AI referral traffic grew from near-zero to a measurable assisted-conversion channel.
Tactics applied: direct-answer page restructure, original technical benchmarks, schema hygiene, and trade-publication mentions.
Result: citations appeared across Google AIO, ChatGPT, and Gemini after entity facts and expert review were made visible.
Tactics applied: YMYL author bylines, source-backed comparison tables, third-party profile parity, and prompt reruns.
Primary evidence
Prompt capture, crawl export, and source ledger row.
Third-party corroboration and profile parity note.
Supplementary evidence
Appendix file, screenshot reference, or companion report.
Peer-reviewed papers and academic studies
The first peer-reviewed baseline. Tactics tested on 10k queries.
Top source concentration and citation dynamics.
Q2 2026 field evidence
1,885 vs 4,000 controls, difference-in-differences. Source used to downgrade schema from growth lever to hygiene.
Only a small share of cited URLs overlap across engines; engine-specific reporting is required.
B2B buyers increasingly start with answer engines, so reports separate discovery, shortlist, and conversion evidence.
V4 · compiled May 2026 from 70+ primary sources · internal toolkit