UI/UX, SEO, AEO, and GEO Improvement Report

Status

Current implementation report for IARPG-OPS-2 2.0.7-wip. The repository changes and local validation are complete; deployment-host, native-browser, assistive-technology, live-crawler, and field-performance verification remain required before promotion beyond work-in-progress status.

Purpose

Document the deep interface, navigation, search-discovery, answer-engine, and generative-discovery improvement round applied to iarpg.com while preserving the active .uai/ startup memory, international fairness rules, browser-local operation, and the rule that only the centered menu strip may remain sticky.

Scope

This round covers the shared PHP page shell, centered navigation, responsive interaction behavior, accessibility states, structured data, canonical routing, answer-first content, report and standard citation surfaces, crawler controls, machine-discovery artifacts, image delivery, conditional JavaScript loading, release metadata, memory routing, and offline validation. It does not claim deployment-host search indexing, Core Web Vitals field data, native Firefox behavior, native assistive-technology speech output, or stable-release readiness.

Executive Summary

The centered .menu-strip is now the publication's only sticky surface. It stays at the top of the viewport across desktop and mobile layouts, receives a visible scrolled state, respects safe-area insets, preserves anchor offsets, and contains a more understandable responsive navigation system. The former unstructured overflow menu is organized into functional groups for exploration, building and review, assurance and publication, and learning and governance.

The shared page shell now emits clean canonical URLs, consistent title and description metadata, Open Graph and Twitter metadata, page-aware article metadata, BreadcrumbList markup, and a schema graph connecting the organization, website, page, breadcrumbs, and page-specific entities. Representative standards, reports, regional methodology pages, collections, and the answer center now expose answer-first summaries and citation routes without turning machine hints into a second source of truth.

Discovery was broadened with explicit crawler policy, an OpenSearch description, an Atom update feed, a machine-readable answer catalog, and a non-authoritative llms.txt citation guide. Raw Markdown, UAI, checksum, and JSON companions remain fetchable but are marked noindex at the server-header layer so canonical HTML readers remain the preferred search result. Static assets are versioned and cacheable, large partner images have WebP alternatives, and feature-specific JavaScript bundles are no longer loaded on pages that do not use them.

Evidence Reviewed

Repository evidence reviewed during implementation included:

• AGENTS.md and the documented .uai/ startup order.
• Shared layout, navigation, theme, search, content, and footer code in app/layout.php, app/content.php, assets/css/site.css, and assets/js/site.js.
• Page-specific renderers for the home page, FAQ, standards, individual standards, report archive, individual reports, global atlas, and international methodology.
• Apache/LiteSpeed configuration in .htaccess, public discovery records, sitemap, search index, manifests, release gates, and public UAI mirrors.
• Current durable implementation and memory architecture reports, especially the split-memory migration report and international fairness method.
• Primary public guidance reviewed for this round: Google Search Central documentation on people-first content, title links, snippets, structured data, mobile-first indexing, breadcrumbs, and AI search features; OpenAI documentation for OAI-SearchBot, GPTBot, and ChatGPT-User controls; and Bing Webmaster guidance on IndexNow and crawl discovery.

Findings

Sticky centered navigation

• .menu-strip is explicitly fixed as the sole sticky selector through the final CSS cascade.
• Previously sticky sidebars, filter bars, workbench panels, publication rails, and section headings are normalized to static positioning.
• The menu uses top: 0, a high stacking context, safe-area padding, a stable height variable, reduced-transparency fallback, and a scrolled-state shadow.
• Document anchor targets use scroll-padding-top and scroll-margin-top so the sticky menu does not obscure headings.
• Mobile navigation is height-limited to the dynamic viewport and scrolls internally rather than extending beyond the screen.

Navigation and interaction design

• The overflow menu is grouped by user intent instead of presenting one undifferentiated list.
• Navigation supports outside-click dismissal, Escape dismissal, deterministic desktop reset on resize, and accurate open/close accessible labels.
• Search, theme, and navigation controls have larger targets and clearer hover and focus states.
• The command palette remains keyboard-accessible and its purpose is described in plain language.
• Print output removes navigation and other non-content chrome.

Search engine optimization

• Canonical paths normalize direct PHP entry points and dynamic records to their clean public routes.
• Every page receives a unique title, description, canonical URL, language alternatives, social metadata, and a connected schema graph.
• Error responses emit noindex,follow,noarchive rather than advertising missing pages for indexing.
• Breadcrumb markup is present in both visible navigation and JSON-LD.
• Individual standards and reports expose article/report entities with names, descriptions, categories, review context, citations, and stable URLs.
• Report readers now keep exactly one page-level H1 even when a preserved Markdown source contains several top-level headings. Later source H1 headings are demoted visually to H2 while their stable anchors remain intact.
• Canonical report readers with duplicate source headings receive distinct, source-qualified page titles, H1s, metadata descriptions, and Report names without rewriting the preserved Markdown body.
• Collection pages expose ItemList markup for their visible records.
• The sitemap is regenerated from current routes and canonical report records while excluding four compatibility-only redirect aliases.

Answer engine optimization

• The FAQ is expanded into direct questions that match actual user intent, including what IARPG is, what is currently available, international fairness, AI boundaries, privacy, research provenance, and the difference between the reference build and a shipped MMO.
• The visible FAQ and FAQPage structured data use the same question and answer source.
• Reusable answer briefs place a concise answer before long-form supporting material on priority pages.
• Citation cards distinguish canonical standards, implementation facts, and exploratory durable research.
• A machine-readable answer catalog mirrors the visible public answers and links each answer to canonical evidence routes.

Generative engine optimization

• llms.txt provides a compact discovery and citation route while explicitly stating that it is not authoritative and does not override visible pages, standards, implementation, or evidence records.
• Crawler policy distinguishes OAI-SearchBot, ChatGPT-User, GPTBot, Googlebot, Bingbot, and generic crawlers without exposing private .uai/, application code, or API paths.
• Public UAI mirrors remain discoverable while hidden active startup memory remains server-denied.
• Answer, report, standard, organization, website, and breadcrumb entities use stable identifiers and canonical routes that can be cited consistently.
• International-fairness summaries appear in answer-first copy and link to the full method, reducing the risk that an extracted answer omits the site's equal-evidence, rights, oversight, uncertainty, endonym, and anti-typecasting constraints.

Performance and crawl efficiency

• CSS and JavaScript URLs include the release version and can receive long immutable caching.
• Large partner advertisements and the internal hero logo use WebP with PNG fallbacks and explicit dimensions.
• Tool-specific JavaScript bundles load only on pages with matching functionality; ordinary content pages load only the shared site script.
• Brotli and deflate compression declarations cover text assets when the host modules are available.
• Machine and source companions receive short revalidation caching and X-Robots-Tag: noindex.

Decisions or Recommendations

• Keep .menu-strip as the only sticky component and reject future sticky sidebars unless the repository constraint is explicitly superseded.
• Treat canonical HTML pages as the primary public answer and citation surface. Machine-readable discovery files are routing aids, not independent authority.
• Keep direct answers concise, then link to standards, implementation evidence, and durable research instead of copying long report bodies into page chrome.
• Preserve the same international fairness summary wherever country, institution, authority, or regional content is condensed for search or AI extraction.
• Do not create or publish an IndexNow key from the build environment. Configure IndexNow only on the actual host, with host-owned credentials and submission logs.
• The single best next action is to deploy this WIP package to a staging Apache or LiteSpeed host and collect live crawl, structured-data, performance, Firefox, and assistive-technology evidence.

Risks and Limitations

• Search engines and answer systems decide whether and how to index, quote, or cite public content; metadata and discovery files do not guarantee inclusion or ranking.
• llms.txt is an optional convention and is intentionally treated as supplementary rather than a source of truth.
• Structured data must continue to match visible content; future content edits can create drift if validators are not run.
• Long immutable caching is safe only because CSS, JavaScript, and image URLs are versioned or stable immutable assets. A deployment that replaces bytes at the same unversioned image URL must purge caches or change the URL.
• No field Core Web Vitals data, search-console crawl data, Bing live inspection, IndexNow host submission, or production server-header capture is available in the local build.
• Native Firefox, native assistive-technology speech output, and actual target-host behavior remain unverified promotion blockers.
• The research archive contains source-derived reports with explicit truth boundaries; improving their discoverability does not independently verify every external claim.

Validation Performed

Final validation evidence for this round is recorded in the versioned root validation files and release-promotion gates. The completion suite includes 88 PHP lint checks, 5 JavaScript syntax checks, 267 JSON parses, 3 XML parses, CSS parsing, 64 schema-declared record validations, visible/structured/machine parity for 26 answers, and a full local crawl of all 334 sitemap routes. Every crawled canonical route returned HTTP 200, contained exactly one H1, exposed a matching canonical URL, and had a unique title and meta description.

The suite also verifies 43 representative HTTP cases (37 current routes, two intentional 404 responses, and four compatibility redirects), 24 Chromium fixture interaction checks, 2,438 local documentation links, 2,061 heading anchors, 194 pointer records and checksums, 95 byte-identical UAI mirrors, 255 preserved source files, 676 documentation checksums, source and partner-artwork byte integrity, zero nested archives, flat archive shape, ZIP integrity, and the external ZIP SHA-256 sidecar.

Promotion-sensitive checks that require a real host, native Firefox, native assistive-technology speech output, or field search/performance data remain explicitly not tested rather than being converted into passing claims.

Memory References

• Current progress
• UI shell
• Accessibility
• Performance audit
• Publication
• Search
• Test plan
• Pointer ledger

Related Durable Documents

• [UAI Memory Organization Report](uai-memory-organization-report.md#executive-summary)
• Split-memory architecture
• [International institutional fairness, rights, and evidence audit](equal-standards-across-unequal-records-international-institutional-fairness-rights-and-evidence-audit.md#mandatory-analytical-distinctions-and-rights-safeguards)
• [Global comparative intelligence methodology](global-methodology-for-comparative-intelligence-and-security-institutions.md#5-methodology-source-hierarchy-and-confidence-framework)
• Long-term memory index

Supersession Status

Current for IARPG-OPS-2 2.0.7-wip. A later implementation report may supersede this report only by naming stable ID ltm-report-ui-ux-seo-aeo-geo-improvement, preserving its audit history, and updating the pointer ledger.