new site

seo-audit-2026-05-27/FULL-AUDIT-REPORT.md

@@ -0,0 +1,281 @@
+# Full SEO Audit — odoo-expertos.com
+
+**Date:** 2026-05-27
+**Auditor:** claude-seo v2.0.0 (AgriciDaniel) — orchestrated multi-agent audit
+**Host:** Vercel (canonical host `www.odoo-expertos.com`)
+**Site type:** Odoo ERP consulting, multilingual (ES root + EN/DE/FR/PT/AR), ~1,210 pages, static HTML
+**Output location:** this folder only (`seo-audit-2026-05-27/`). Nothing was written to any other project.
+
+---
+
+## Scope (read first)
+
+This is a **bounded, representative audit**, not a 500-page crawl. Analyzed:
+- All site-wide files: `robots.txt`, sitemap index + all 6 language sitemaps, `llms.txt`, `llms-full.txt`
+- 6 language homepages (es/en/de/fr/pt/ar)
+- Category index pages across all 6 languages (`/odoo/`, `/odoo-hosting/`, `/odoo-ia/`)
+- Representative deep article pages in ES/EN/DE
+- Homepage screenshots (desktop + mobile, in `screenshots/`)
+
+The structural problems found replicate across language siblings and the ~1,210-page template, so the sample is representative of systemic issues. Raw fetched data is in `raw/`.
+
+> **PageSpeed note:** Google PSI/CrUX was rate-limited (no API key on the shared quota) at audit time, so Core Web Vitals numbers below are **ESTIMATED from HTML signals**, not measured field/lab data. Re-run with a PSI API key for exact CWV.
+
+---
+
+## Executive Summary
+
+**Overall SEO Health Score: ~44/100 — Poor. Significant, systemic issues.**
+
+| Category | Weight | Score | Verdict |
+|---|---|---|---|
+| Technical SEO | 22% | 54 | Hreflang broken on inner pages; 3 category pages 404 in sitemaps |
+| Content Quality / E-E-A-T | 23% | 25 | 400+ EN/DE pages serve untranslated/garbled Spanish; fabricated authors |
+| On-Page SEO | 20% | 35 | 270–300+ char keyword-stuffed titles; orphan articles |
+| Schema / Structured Data | 10% | 45 | Broken logo domain, www mismatch, DE page ships Spanish schema |
+| Performance (CWV) | 10% | ~65 (est) | Real optimizations present; CWV not measured (PSI rate-limited) |
+| AI Search Readiness (GEO) | 10% | 50 | Excellent crawler posture; homepage hype suppresses citability |
+| Images | 5% | ~60 (est) | Lazy-load claimed; not fully verified |
+
+### Top 5 Critical Issues (fix immediately)
+1. **400+ EN/DE pages are not translated** — they serve Spanish (or broken machine-mixed Spanish/German like "Inteligencai Artificail", "ueber", "fuer") under `lang="en"`/`lang="de"`. This is a site-wide quality/deception signal that risks a domain-level downgrade.
+2. **3 category index pages 404 while listed in their sitemaps** — `/en/odoo-hosting/`, `/fr/odoo-hosting/`, `/pt/odoo-hosting/` all return 404 but appear as `<loc>` entries → poisons sitemap trust + breaks the hub for ~200 pages each.
+3. **Hreflang broken on inner pages** — EN inner pages omit `pt` + `ar` (non-reciprocal); DE inner pages point `hreflang="es"` and `x-default` to non-existent `/de/`-prefixed Spanish slugs (404). Category index pages have **zero** hreflang.
+4. **Broken/inconsistent Organization logo + URL in schema** — logo points to `www.odooexpertos.com` (no hyphen — domain does not resolve); CollectionPage/breadcrumb URLs use non-www while canonicals use www → entity-consolidation failure.
+5. **Homepage is anti-E-E-A-T and anti-citation** — fear/combat hype ("Tus Competidores Usan IA Para Robarte Clientes", CTA "DESTRUIR COMPETENCIA AHORA"), unsourced stats (280% ROI), publisher entity "Apex AI Research Labs LLC", no About/Contact page.
+
+### Top 5 Quick Wins
+1. Remove the three 404 `<loc>` entries from `sitemap-en/fr/pt.xml` (stops sitemap poisoning today) — then build the missing index pages.
+2. Fix the schema logo URL site-wide (`odooexpertos.com` → `www.odoo-expertos.com/favicon.svg`) and switch all schema `url`/`item` to the www host.
+3. Add the missing security headers via `vercel.json` (`X-Content-Type-Options`, `X-Frame-Options`, `Referrer-Policy`).
+4. Collapse the duplicate robots.txt groups, merge the useful `Disallow: /*?` `/*&` into the FIRST `User-agent: *` group, delete `Disallow: /*#`.
+5. Fix `og:locale` on DE/FR/PT category pages (currently `es_ES`) and the truncated `dateModified` in Article schema.
+
+### What's genuinely good
+- **Canonicalization is correct**: apex `odoo-expertos.com` → 301 → `www`; `http` → 308 → `https`; self-referencing canonicals on www; HSTS present (2-year max-age).
+- **Homepage hreflang is complete and reciprocal** across all 6 languages incl. `x-default`.
+- **AI-crawler posture is exemplary** — robots.txt explicitly allows GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended, CCBot, Applebot-Extended, etc., and routes them to `llms.txt`. SEO scrapers (Ahrefs/Semrush/MJ12/Bytespider) correctly blocked.
+- **Static HTML on Vercel** = best possible crawl/AI-ingestion foundation.
+- **Arabic RTL** done correctly (`<html lang="ar" dir="rtl">` + Noto Sans Arabic).
+- **Critical CSS inlined + async minified CSS** — real performance work shipped.
+- **Sitemaps comprehensive**: 1,210 URLs (es 201 / en 206 / de 205 / fr 199 / pt 199 / ar 200) with xhtml:link hreflang annotations.
+
+---
+
+## 1. Technical SEO & Hreflang — Score 54/100
+
+### [Critical] Three category index pages 404 while listed in their sitemaps
+`/en/odoo-hosting/`, `/fr/odoo-hosting/`, `/pt/odoo-hosting/` all return **HTTP 404**, yet each appears as a `<loc>` entry in its language sitemap. (ES, DE, AR versions return 200; `/en/odoo/`, `/en/odoo-ia/`, `/de/odoo/`, `/de/odoo-ia/` return 200.)
+- **Why:** Submitting 404s in a sitemap degrades trust in the whole sitemap and can suppress crawl priority. These are hub pages — without them, ~200 inner pages per language lose their category link root.
+- **Fix now:** Remove the 3 dead `<loc>` entries from `sitemap-en/fr/pt.xml` today. **Fix properly:** build the missing index pages (mirror the working DE/AR versions).
+
+### [Critical] Hreflang broken on inner pages (reciprocity + wrong targets)
+- EN inner pages (e.g. `/en/odoo-hosting/best-odoo-hosting-2026/`) declare only es/en/de/fr/x-default — **`pt` and `ar` are missing**, so ES→PT/AR is not reciprocated. Google discards a non-reciprocal hreflang cluster entirely.
+- DE inner pages (e.g. `/de/odoo-hosting/bestes-odoo-hosting-2026/`) set `hreflang="es"` and `x-default` to `https://www.odoo-expertos.com/de/odoo-hosting/mejor-hosting-odoo-2026/` — a `/de/`-prefixed Spanish slug that **404s**. Correct target is the root `https://www.odoo-expertos.com/odoo-hosting/mejor-hosting-odoo-2026/`.
+- **Fix:** Templated fix across the EN template (add pt+ar) and DE template (build es/x-default hrefs from the root, no `/de/` prefix). Affects ~200 pages each.
+
+### [Critical] Category index pages have ZERO hreflang
+`/odoo-hosting/`, `/de/odoo-hosting/`, `/odoo-ia/` (and likely all category hubs) return 0 hreflang `<link>` tags. Google cannot connect the language versions of these high-equity hubs → they compete with each other.
+- **Fix:** Add the full 7-entry block (es,en,de,fr,pt,ar,x-default) to every category index `<head>`.
+
+### [High] robots.txt structural errors
+- **Two `User-agent: *` groups.** Per RFC 9309, only the FIRST matching group applies — so the second block's useful rules (`Disallow: /*?`, `/*&`) are **silently ignored**. Parameterized/duplicate URLs you intended to block are being crawled.
+- **Duplicate `User-agent: Bingbot`** (Bing section vs AI section, with different Allow rules) and **duplicate `User-agent: CCBot`** — dead/conflicting config.
+- **`Disallow: /*#`** is meaningless (fragments never reach the server).
+- **Fix:** One group per agent. Merge `Disallow: /*?` `/*&` into the FIRST `User-agent: *`. Delete the second `*` block, the duplicate Bingbot/CCBot groups, and `/*#`.
+
+### [Medium] Missing security headers
+Present: `Strict-Transport-Security: max-age=63072000` (good). Missing: `X-Content-Type-Options`, `X-Frame-Options`, `Referrer-Policy`, `Permissions-Policy`, CSP.
+- **Fix (`vercel.json` headers):** add `X-Content-Type-Options: nosniff`, `X-Frame-Options: SAMEORIGIN`, `Referrer-Policy: strict-origin-when-cross-origin`.
+
+### [Low] http→https→www = 2 redirect hops
+`http://odoo-expertos.com` → 308 → `https://odoo-expertos.com` → 301 → `https://www…`. Standard, not penalized, but a direct `http-apex → https-www` rule would save one hop.
+
+### What's already good (technical)
+- Apex→www 301 and http→https 308 correct; **www is the single canonical host**; self-referencing canonicals; HSTS 2yr.
+- **Homepage hreflang complete + reciprocal** across all 6 languages incl. x-default; valid ISO codes; www hosts.
+- No `noindex` on any sampled page; viewport tag everywhere; **Arabic RTL** (`dir="rtl"` + Noto Sans Arabic) correct.
+- Sitemap index → 6 language sitemaps with xhtml:link hreflang annotations; 1,210 URLs total.
+- robots.txt correctly blocks SEO scrapers (Ahrefs/Semrush/MJ12/Dot) while allowing all search + AI bots.
+
+## 2. Content Quality & On-Page SEO — Score ~30/100 (E-E-A-T 18/100)
+
+### [Critical] 400+ EN/DE pages serve untranslated / garbled content under the wrong `lang`
+- `/en/odoo/odoo-hosting-cloud-infraestructura-2025/` declares `lang="en"` but H1, H2s, body and meta are **entirely Spanish**.
+- `/de/odoo-hosting/` ships broken machine-mixed text: title "…Vollstaendiger Leitfaden **de** Lieferanten Cloud … fuer ERP…", body "…necesitas saber **ueber** infraestructura, Lieferanten…", H2 "100+ **Leitfadens** Especializadas".
+- `/en/odoo-ia/` H1/title/meta all read "**Inteligencai Artificail**" (misspelled, mixed ES/EN: "RPA y Chatbots Smarts", "Predicción of Sales").
+- **Why:** Declared-language ≠ content-language is a deception signal; at 400+ pages it risks a **site-wide quality downgrade** and breaks hreflang (language signals contradict content). These pages give zero value to EN/DE users and near-duplicate the Spanish originals.
+- **Fix (pick one, sitewide):** (a) genuinely translate EN/DE pages; (b) remove EN/DE article pages and keep Spanish-only with hreflang back to ES; (c) 301 untranslated EN/DE paths → Spanish canonical. Only (a) preserves the multilingual investment. **Do not generate more Spanish-slug EN/DE pages.**
+
+### [Critical] Category/directory titles are 270–300+ chars of keyword stuffing
+e.g. `/odoo-hosting/` title is 271 chars: "Directorio de Artículos Odoo Hosting - Guía Completa de Proveedores Cloud | AWS, Azure, DigitalOcean para ERP | … | 50+ Tutoriales… | Odoo Expertos". `/odoo/` ~250+, `/odoo-ia/` 273, `/de/odoo-hosting/` 277 (and Spanish/German mixed).
+- **Why:** Google truncates/rewrites at ~60 chars; a 270+ char pipe-stuffed title is a manipulation signal under the Sep-2025 QRG.
+- **Fix:** Rewrite to 50–60 chars in the target language: `[Primary keyword] — [value] | Odoo Expertos`.
+
+### [High] Article pages are orphans — zero in-body internal links
+`/odoo/odoo-hosting-cloud-infraestructura-2025/` has 4 hrefs total (favicon ×2, self-canonical, CSS) — **no related-article links, no rendered breadcrumb, no link to parent category**. PageRank cannot flow; with 1,210 pages the architecture is a flat spike.
+- **Fix:** Add rendered HTML breadcrumb (not just schema), a parent-category link, and ≥3 related-article links per article (templated).
+
+### [High] Homepage copy destroys E-E-A-T
+H1 "Mientras Lees Esto, Tus Competidores Usan IA Para Robarte Clientes"; H2s "⚔️ ELIGE TU ARMA DE DOMINACIÓN", "🔥 COMPETIDORES SIENDO DESTRUIDOS EN VIVO", CTA "DESTRUIR COMPETENCIA AHORA"; "¡ALERTA! …ROBANDO tus mejores clientes" (confirmed in screenshot). Unsourced "280% ROI" in the title tag; publisher meta = "Apex AI Research Labs LLC".
+- **Why:** Combat/fear hype + unsourced superlatives is the pattern QRG raters flag as low-trust lead-gen for a B2B ERP buyer.
+- **Fix:** Factual, keyword-led H1 (e.g. "Consultoría e Implementación Odoo ERP para Empresas — LATAM"); benefit-framed H2s; move urgency to a secondary CTA; source or drop the ROI claim.
+
+### [High] Non-localized URL slugs on EN/DE
+`/en/odoo-ia/` ("IA" is Spanish), `/en/odoo/…infraestructura…` (Spanish slug). Localized equivalents `/en/odoo-ai/`, `/de/odoo-ki/` 404.
+- **Fix:** Localized slug convention for new pages; 301 once content is translated.
+
+### [Medium] Thin article bodies (~400–450 words of real content)
+`/odoo/odoo-hosting-cloud-infraestructura-2025/` ≈610 words incl. chrome → ~400–450 body words on a complex technical topic. H2 outline is good; sections are 1–2 sentences.
+- **Fix:** Expand each H2 to 200–300 words of genuine guidance.
+
+### [Low] Trust gaps
+No `/contacto/` or `/sobre-nosotros/` page (both 404); only footer legal links + a Wyoming LLC shell address. Author "Dr. María Elena Rodríguez" has no author page (`/autor/…` 404); homepage author meta = "Apex AI Research Labs LLC" (different entity). Odoo-partner claim unverified.
+- **Fix:** Real About + Contact pages in nav; verifiable author pages or attribute to the org; link/verify Odoo partner badge.
+
+### What's already good (content)
+- Homepage hreflang + canonicals correct; page-level `lang` attribute is correct even where content isn't (so the fix is content, not template structure).
+- Spanish article title/meta well-formed; Spanish article H2 hierarchy is logical and citable.
+- Legal pages (privacy/terms/legal-notice) + footer address/email present.
+- Critical CSS inlined + async minified CSS shipped.
+
+## 3. Schema / Structured Data — Score 45/100
+
+What exists: Homepages have `Organization` + `WebSite` (+SearchAction). Category pages have `CollectionPage` (inline `BreadcrumbList` + `Organization` publisher). Article pages have `Article` (JSON-LD) **plus** duplicate Microdata. All use `https://schema.org` (correct).
+
+### [Critical] Broken logo domain — `www.odooexpertos.com` (no hyphen) does not resolve
+CollectionPage publisher logo = `https://www.odooexpertos.com/assets/logo.png`. The Article publisher uses yet another (`…/logo.png` non-www). Homepage uses `favicon.svg`. → 3 different logo URLs, one pointing at a **dead domain**. No logo passes validation.
+- **Fix:** One absolute logo URL on the real domain, site-wide (e.g. `https://www.odoo-expertos.com/favicon.svg` or a real 112px+ PNG).
+
+### [Critical] German page ships Spanish schema
+`/de/odoo-hosting/` (`lang="de"`) CollectionPage `name`="Centro de Recursos Hosting Odoo", breadcrumb "Inicio"/"Hosting Odoo", no `inLanguage`. Misleads entity understanding.
+- **Fix:** Translate schema strings to German; add `"inLanguage":"de"` (breadcrumb "Inicio"→"Startseite").
+
+### [Critical] Article schema missing required `image` (+ `mainEntityOfPage`)
+`/odoo/odoo-hosting-cloud-infraestructura-2025/` Article has headline/author/dates but **no `image`** → ineligible for Google Article rich results. `dateModified` is also truncated/invalid ("202…") → Google ignores the block.
+- **Fix:** Add absolute `image` (≥1200px) + `mainEntityOfPage`; populate a valid ISO `dateModified`.
+
+### [High] Schema URLs use non-www while canonicals use www
+CollectionPage `url`/breadcrumb `item` = `https://odoo-expertos.com/…`; canonical = `https://www…`. Entity-consolidation mismatch.
+- **Fix:** Use the www host in all schema `url`/`item`.
+
+### [High] Duplicate Article markup (JSON-LD + Microdata) on same page → remove Microdata `itemtype`.
+### [High] Organization differs ES vs EN homepage with no shared `@id` → define once at root with `"@id":"…/#organization"` and reference it elsewhere (EN's richer address/founder data is otherwise invisible site-wide).
+### [Medium] `sameAs: []` empty on all Org blocks → populate (LinkedIn, YouTube, directories) — high value for Knowledge Panel + AI entity resolution.
+### [Medium] `WebSite` SearchAction points to `?s=` search that likely doesn't exist on a static site → remove if non-functional.
+### Missing opportunities: `BreadcrumbList` on ~240 deep pages; `Service` on homepage; `FAQPage` where question-H2s exist; `inLanguage` on ES/DE blocks.
+
+### What's good (schema): correct `@context`; EN Org is rich (address/founder/parentOrganization/availableLanguage); absolute URLs; ISO dates; no deprecated types.
+
+---
+
+## 4. AI Search Readiness (GEO) — Score 50/100
+
+### [Critical] Homepage is anti-citation
+Caps-lock urgency, military metaphors, unsourced stats (280%/65%/400%), no schema/date/author on the highest-authority page → LLMs deprioritize the domain as a source. **Fix:** factual entity-style homepage (see Content §).
+
+### [High] `llms.txt` is spec-valid but link-thin — claims "1,185+ guides" but lists only ~9 URLs; "Key Topics"/"Content Quality" sections are link-free prose. **Fix:** add 8–12 representative URLs per `##` section.
+### [High] `llms-full.txt` is ~11 KB (a longer index, not a content dump) — tools that auto-ingest it cache a directory, not substance. **Fix:** generate a real per-language full-text dump (`llms-full-es.txt`, …) with H1 + intro + tables per page.
+### [High] No authors/dates on content pages; 2025 vs "2026" freshness inconsistency; missing `Article`/`FAQPage` JSON-LD (overlaps Schema §). **Fix:** bylines + datePublished/dateModified + Article/FAQ schema.
+### [Medium] Long interlocked passages — apply "direct answer first" in 134–167-word self-contained blocks (the ES "¿Odoo Cloud vs Hosting Propio?" H2 already does this — extend it).
+### [Medium] No off-site entity signals (YouTube/Reddit/Wikipedia). YouTube mentions correlate strongest with AI citations — 3–6 month play.
+### [Low] D3 visualizations are client-rendered → invisible to non-JS AI crawlers. Add an accessible HTML data table beside each chart.
+
+### What's exemplary (GEO): robots.txt allows every major answer engine and routes to llms.txt; static HTML = ideal ingestion; llms.txt structurally compliant; 6-language sitemap discovery.
+
+Platform estimates: Google AIO 45 · ChatGPT/SearchGPT 50 · Perplexity 55 · Bing Copilot 45.
+
+---
+
+## 5. Performance / Core Web Vitals — Score ~65/100 (ESTIMATED — PSI rate-limited, no field/lab data)
+
+> CWV could not be measured (Google PSI/CrUX returned HTTP 429 on the shared quota). Numbers are estimated from HTML/headers. Re-run with a PSI API key or `npx lighthouse <url>` for real values.
+
+Estimated (mobile): Homepage LCP ~2.0–2.8s · INP ~150–300ms · CLS ~0.05. D3 page LCP ~1.5–2.0s · INP ~100–200ms.
+
+### [Critical] `document.write()` injects header/footer/eeat on every content page
+`<script>document.write(createHeader())</script>` (×3) is parser-blocking, defeats the preload scanner, and causes layout shift when the header pops in. Affects ~990 content pages. **Fix:** render into a pre-sized placeholder div via `insertAdjacentHTML`, `defer` the component scripts, reserve header height (`min-height`).
+### [High] 32 KB inline `<style>` on homepage — only ~1.2 KB is true critical CSS; the rest (animations, cards, carousel) is non-critical and delays style recalc/LCP. **Fix:** move non-critical rules to `main.min.css`; keep inline <14 KB.
+### [High] Content pages load `main.min.css` render-blocking (no async pattern, no critical CSS) — apply the homepage's `preload…onload`+`<noscript>` pattern in the page generator (~990 pages).
+### [High] D3.js loaded from `d3js.org` at runtime on the homepage (no preconnect), while the D3 *content page* has a `preconnect` to d3js.org but doesn't use D3. **Fix:** self-host `d3.v7.min.js`, preload only where used, remove the misplaced preconnect, defer D3 init until in-viewport (a static SVG already covers first paint).
+### [High] Large unthrottled inline JS on homepage — mousemove particle trail creates a DOM node per move (INP risk). **Fix:** externalize + `defer`; throttle with rAF; disable particle trail on mobile.
+### [Medium] `Cache-Control: public, max-age=0, must-revalidate` on all assets — every repeat visit revalidates CSS/JS. **Fix:** `max-age=31536000, immutable` for hashed/versioned `/css`,`/js`,`/components` in `vercel.json`; keep short TTL on HTML.
+### [Medium] Continuous CSS animations (kenBurns/shimmer) run off-screen → wrap in IntersectionObserver.
+
+### What's good (performance): Brotli on all assets (HTML 109KB→~23KB); Vercel edge HTTP/2; HSTS; critical CSS inlined + async CSS on homepage (correct pattern); **system font stack (no web fonts)**; no hero image (text LCP from inline CSS); all external scripts `defer`; inlined SVG icon sprite; lightweight D3 content page (~17KB).
+
+---
+
+## 6. Images — Score ~60/100 (not deeply crawled this pass)
+Deployment notes claim lazy-loading + 1-year image cache, but the homepage uses CSS gradients + inline SVG (no raster hero). Not verified at scale.
+- **To check next:** alt-text coverage on content-page images, webp/avif vs png/jpg, explicit width/height (CLS), and that `loading="lazy"` is NOT applied to above-the-fold images. Run `seo-images`/`seo-visual` on a deep-page sample with a PSI key.
+
+---
+
+## Methodology & Artifacts
+- Tool: claude-seo v2.0.0, 6 parallel specialist agents (technical, content, schema, geo, performance) + site-wide PowerShell/curl probes.
+- Raw data: `raw/` (robots.txt, 6 sitemaps, llms.txt/full, homepage HTML, redirect map, PSI attempt log).
+- Screenshots: `screenshots/` (homepage desktop + mobile).
+- Limits: bounded ~20-URL sample (representative of the 1,210-page templates); CWV estimated (PSI 429); image audit deferred.
+
+
+---
+---
+
+# PART B — DEEP AUDIT (Google Search Console + live measurement, 2026-05-27)
+
+Added after reviewing real GSC index-coverage exports (`Downloads\website issu oddo expertos\`) + live crawl/measurement. **This is the most important section** — it explains *why the site barely ranks*.
+
+## B0. The headline: ~72% of the site is NOT indexed
+
+From GSC index-coverage exports (status as of 21.05.26):
+
+| GSC status | Pages | What it means | Dominant root cause |
+|---|---:|---|---|
+| **Gecrawlt – zurzeit nicht indexiert** (Crawled, not indexed) | **587** | Google crawled them and *chose not to index* | Thin / templated / low-value content |
+| **Gefunden – nicht indexiert** (Discovered, not indexed) | **134** | Found in sitemap, not even crawled | Crawl budget starved by low site quality + weak internal linking |
+| **Duplikat, Google wählte anderen Canonical** | **103** | Google treats them as duplicates of another page | **Untranslated** PT/FR/EN pages = duplicates of the Spanish original |
+| **Alternative Seite mit kanonischem Tag** | **31** | Canonicalized away (mostly `/en/odoo/…`) | Correct canonical, but page adds no unique value |
+| **Seite mit Weiterleitung** (redirect) | **~45** | URL redirects | Country-variant + `/index.html` URLs **still listed in sitemaps** |
+| **Nicht gefunden (404)** | **12** | Dead | Case-sensitivity + wrong-category + missing pages |
+
+**~870 of ~1,210 URLs (≈72%) are excluded from Google's index.** Only ~28% can rank. This is the single biggest problem and it dwarfs everything in Part A.
+
+### The root-cause chain (fixing these clears most of the 870)
+1. **Untranslated content** (Part A §2 C2) → 103 "duplicate" + a large share of the 587 "crawled not indexed". Google literally flags PT/FR pages as duplicates of the ES original because the body text *is* Spanish.
+2. **Thin, templated content** (~400–450 words, near-identical boilerplate) → the bulk of the 587 "crawled not indexed". Google's bar for indexing a 1,210-page programmatic site is "is each page uniquely valuable?" — currently no.
+3. **Templated/near-duplicate meta descriptions** → reinforces the duplicate signal. Measured live: `/odoo/odoo-crm-…` and `/odoo/odoo-inventario-…` both use *"Guía completa sobre Odoo [X] con Odoo. Descubre las mejores prácticas y soluciones empresariales."* — only the keyword swapped. (Your hypothesis confirmed in pattern; titles are not identical but article `title` frequently equals the `H1`.)
+4. **Weak internal linking / crawl depth** → 134 "discovered not indexed" (Google won't spend budget crawling pages the site itself barely links to).
+5. **Sitemap hygiene** → sitemaps contain 404s, redirecting URLs, and non-canonical URLs, which lowers trust in the whole sitemap and wastes crawl budget.
+
+## B1. Click depth & internal linking (measured live)
+- Homepage exposes **only 9 internal links** (home, 3 category hubs, a few CTAs). The nav is **injected client-side via `document.write(createHeader())`** in `components/header.js` — it is NOT in the static HTML, so the entire crawlable link graph depends on JS rendering.
+- Category hubs link **160 articles** total (`/odoo/` 28, `/odoo-hosting/` 78, `/odoo-ia/` 56).
+- **37 of 193 Spanish articles are NOT reachable within 2 clicks** (orphaned — only in the sitemap). Extrapolated across 6 languages ≈ the 134 "discovered not indexed".
+- **Best practice (your 3-click rule):** every indexable page ≤3 clicks from home. Today many article pages are reachable only via the sitemap (∞ clicks for a user). Article bodies have **zero in-body internal links** (Part A §2 H1) so link equity cannot flow.
+
+## B2. The homepage is visually broken (verified — see `screenshots/www_odoo-expertos_com_desktop.png`)
+Full-page capture shows: large **empty dark sections** where service cards / testimonials / "Funcionalidades" content should be; the hero "ecosystem" visualization renders as a few stray dots (D3 not rendering); broken copy ("Historia de Esto Comprobado"); and an **H1 word-spacing bug present on every homepage** — measured live: "Usan IA **ParaRobarte** Clientes", "Transform Your Business **withOdoo** ERP", "**mitOdoo** ERP", "Centro **deInteligencia Artificialpara** Odoo". Words are concatenated (spans rendered without spaces).
+- Likely cause: content gated behind `opacity:0` + IntersectionObserver scroll reveals (Part A §5) that don't render server-side, plus a span-spacing bug in the headline builder. Either way: **content invisible to no-JS / partial-JS crawlers and to users who don't scroll**, and the page looks unfinished.
+- **Visual score: 3/10.** Not shippable. Needs dev fix + human re-verify.
+
+## B3. 404 patterns worth fixing at the source
+- **Case-sensitivity:** `/odoo-Hosting/digitalocean-…` (capital H) 404s — Vercel paths are case-sensitive. Enforce lowercase URLs + add redirects for known capitalized variants.
+- **Wrong category slug:** `/odoo-ai/odoo-18-ai-…` 404s because the site uses `/odoo-ia/`. Pick ONE (recommend `/odoo-ai/` for EN once localized) and 301 the other.
+- **Linked-but-missing:** `/consultoria-erp`, `/de/odoo/consultoria-erp`, `/recursos/guia-aws-odoo.pdf`, `/mes` — internal links/old URLs point to pages that don't exist. Find the linking source and fix or 301.
+
+## B4. 2026 SEO best-practice comparison
+| 2026 best practice | This site | Gap |
+|---|---|---|
+| Every indexed page uniquely valuable (post-HCU / "crawled-not-indexed" era) | 587 crawled-not-indexed | **Critical** — consolidate/enrich, reduce page count |
+| Real localization per market (not MT) | 400+ MT/untranslated pages | **Critical** |
+| ≤3 clicks, strong internal links, server-rendered nav | JS `document.write` nav; 37+ orphans/lang | **High** |
+| Unique title + meta per page | Templated metas, 260+ char titles | **High** |
+| Server-rendered content (Googlebot-renders-but-budget-limited) | Hero/sections JS-reveal; nav JS-injected | **High** |
+| Sitemaps = only 200, canonical, indexable URLs | Contains 404/redirect/non-canonical | **High** |
+| E-E-A-T: real authors, entity, no fabricated claims | Fake author, "Apex AI Research Labs LLC", 280% ROI unsourced | **High** |
+| Core Web Vitals (INP ≤200ms) | document.write + heavy inline JS | **Medium** |
+| AI-search (GEO) | Strong crawler posture; weak content authority | **Medium** (strength to build on) |
+| HTTPS/HSTS/canonical host | Correct | ✅ |
+
+**Bottom line:** the site was built for *scale* (1,210 pages) before *quality per page*. 2026 Google rewards the opposite. The fix is partly **subtractive** (consolidate thin/duplicate pages, shrink the index to pages that deserve to rank) and partly **corrective** (translate, de-duplicate metas, fix linking, rebuild the homepage, clean sitemaps).