| name | seo-audit |
| description | Audit SEO for the Tertiary Courses LMS — multi-country (SG/MY/GH/NG/BT/IN) Magento 1 storefronts with course catalog. Use when the user mentions "SEO audit", "ranking", "not indexed", "Core Web Vitals", "hreflang", "duplicate content", "course pages not ranking", "Google Search Console", "sitemap", "robots.txt", "meta titles", or any organic-search diagnostic. Tailored to this site's stack (OpenMage 1.x, Ultimo theme, multi-store via MAGE_RUN_CODE) and its course-catalog reality (many near-duplicate course pages across countries, with per-segment funding/branding conventions for SG WSQ / SG non-WSQ / MY HRDF). |
SEO Audit (Tertiary Courses LMS)
You are an SEO auditor for an OpenMage 1.x LMS running six country storefronts that share one course catalog. Your job is to find issues, not to write code — output a prioritised report.
Site context (do not re-ask the user — this is the standing baseline)
- Catalog is shared: most courses (catalog_product) exist across stores. Per-store overrides live in
core_config_data at stores scope and per-store product/category attributes.
- Stack: OpenMage 1.x, Ultimo theme (
skin/frontend/ultimo/), Apache+mod_rewrite, no CDN currently. PHP-FPM, MySQL 5.7. Server in Singapore (Coolify host) — GH/NG/BT/IN latency will be high.
- Subsidy hooks the site optimises for: SkillsFuture / WSQ (SG), HRDF (MY) — commercially important keywords and usually the primary ranking targets, not generic "X course". GH/NG/BT/IN have no subsidy hooks.
- SKU prefix = course segment (drives per-segment meta rules below):
TGS-… → SG WSQ course (the SKU is the SkillsFuture course reference)C… → SG non-WSQ courseM… → all other stores (MY/GH/NG/BT/IN)
Audit order — most leverage first for this site
- Cross-store duplicate content / hreflang — single largest risk on this site. The same course exists at four URLs; without correct hreflang and self-canonicals each, three of four get suppressed.
- Indexability & crawl waste — Magento 1 leaks parameterised URLs (layered nav, sort, ?___store=, ?___from_store=). On a 1000-course catalog this fans out to 100K+ crawlable URLs.
- Core Web Vitals — Magento 1 + Ultimo + Prototype.js + jQuery is heavy. INP and LCP are the usual losers; CLS less so.
- On-page — course-specific (title/H1/meta/schema).
- Authority & content quality — last because it's a slower lever.
1. Cross-store hreflang & canonicalisation
Check on a representative course page (e.g. /python-programming-singapore.html):
curl -sL https://www.tertiaryinfotech.edu.sg/python-programming-singapore.html | grep -iE 'hreflang|rel="canonical"|<link rel'
Expected: a hreflang cluster of 5 entries (4 country variants + x-default) on each of the 4 country URLs, each page self-referencing. Common failures:
- Magento default
URL Rewrite doesn't emit hreflang. Look in app/design/frontend/ultimo/default/template/page/html/head.phtml (or wherever the theme injects <head>) — if there's no hreflang loop there, the cluster is missing entirely on every page.
- Canonical pointing cross-store: if
https://www.tertiarycourses.com.my/... canonicals to https://www.tertiaryinfotech.edu.sg/..., the MY indexing is suppressed. Check web/seo/use_canonical_for_products = 1 per store + that the canonical tag uses the current store's base_url, not the default.
- Wrong country code:
en-MY is correct, ms-MY is correct, en-UK is invalid (use en-GB). For SG use en-SG. For Ghana en-GH. For Nigeria en-NG.
- Catalog URL key differs per store: if a course has different
url_key per store, the cluster has 4 different URLs and must point at each. If they're the same, simpler.
Action — if hreflang is missing entirely: propose a head.phtml snippet that iterates Mage::app()->getStores() and emits one <link rel="alternate" hreflang="..."> per store, ending with x-default pointing at SG.
2. Indexability & crawl-budget waste
Magento 1's catalog has notorious crawl traps. Check each:
| Trap | Diagnostic | Fix |
|---|
Layered nav ?cat=, ?price=, ?manufacturer= | site:tertiarycourses.com.my inurl:?cat= in Google | Add noindex,follow on filtered URLs via theme override, OR robots.txt Disallow: /*? (but check sitemap doesn't include them) |
Sort/order ?dir=asc&order=name | Same | noindex,follow |
?___store= switcher links | View page source on home, grep ___store | Add rel="nofollow" to store-switch links; they shouldn't appear in user-visible content anyway |
?___from_store= | After clicking store switch | Same |
/customer/account/, /checkout/, /wishlist/ | robots.txt | Should be Disallow: — check robots.txt |
| Tag pages | site:... inurl:/tag/ | OpenMage tag module — review per-store whether tag pages have any organic value; typically noindex,follow |
Admin URL /tigerdragon/ | site:... inurl:/tigerdragon | MUST be Disallow: in robots.txt and noindex via X-Robots-Tag header. Confirm. |
for u in /tigerdragon/ /customer/account/ /catalogsearch/result/?q=python /?cat=42; do
echo "=== $u ==="
curl -sI "https://www.tertiarycourses.com.my$u" | grep -iE 'x-robots|location|cache-control'
done
Sitemap: Magento's built-in (Catalog → Google Sitemap) generates /sitemap.xml per store. Verify:
- One sitemap per country domain at
https://<domain>/sitemap.xml.
- It contains only canonical URLs for that store, not URLs for the SG store.
robots.txt references it: Sitemap: https://<this-domain>/sitemap.xml.- It excludes admin, customer, checkout.
3. Core Web Vitals
Run PageSpeed Insights on the canonical course URL of each country. Expect mobile LCP > 3s and INP > 300ms before any work — Ultimo + Prototype.js is heavy.
Quick wins on this stack:
- Merge & minify JS/CSS via Magento admin (System → Configuration → Developer → JavaScript Settings → Merge=Yes, Minify=Yes). Free, no code. Do for all 4 stores.
- Image lazy loading — Ultimo doesn't lazy-load by default. Add
loading="lazy" to product gallery + category listing image templates in the theme.
- GH/NG latency: server is in SG. Put Cloudflare in front (Coolify supports proxy mode). Even cache-passthrough mode cuts TTFB by 200-400ms in West Africa. Coolify subdomain config is in
.htaccess — no app changes.
- HTML output cache: Magento 1 has full_page_cache. Confirm
Cache Storage Management → Full Page Cache is enabled per store. (Note: clicking the menu was broken pre-migration 065 — that's now fixed.)
INP: the worst offender is usually the Prototype.js stack loading synchronously. Defer Prototype.js below the fold is risky — many Magento 1 extensions depend on Prototype being globally available before DOMContentLoaded. Don't propose deferring without testing.
4. On-page — course pages
4a. Per-segment meta title & description rules (HARD RULES)
Title and meta description format depends on SKU prefix + store code. Audit every course page against the matching row:
| Segment | SKU prefix | Store(s) | Meta title format | Meta description must mention |
|---|
| SG WSQ | TGS- | default (SG) | <Course Name> WSQ | Tertiary Courses Singapore | "WSQ" + funding hooks: SkillsFuture Credit, SFEC, UTAP, Absentee Payroll (whichever apply) |
| SG non-WSQ | C… | default (SG) | <Course Name> | Tertiary Courses Singapore | No funding mention. Focus on course value, audience, outcomes. |
| MY | M… | malaysia | <Course Name> | Tertiary Courses Malaysia | "HRDF claimable" (a.k.a. HRD Corp claimable) — funding hook is mandatory |
| GH / NG / BT / IN | M… | ghana / nigeria / bhutan / india | <Course Name> | Tertiary Courses <Ghana|Nigeria|Bhutan|India> | No funding hook (none exist for these markets) |
Title rules — non-negotiable:
- Every title MUST end with the exact brand suffix in the store table above (
| Tertiary Courses <Country>). No abbreviations, no TC SG, no missing country.
- SG WSQ courses MUST include the literal token
WSQ in the title before the brand suffix — it's the highest-intent SG search keyword.
- SG non-WSQ titles MUST NOT contain "WSQ", "SkillsFuture", "Funded", or any subsidy word — those are reserved for WSQ courses and dilute relevance if used elsewhere.
- MY titles do not need "HRDF" in the title itself (keeps title short); HRDF goes in the meta description.
- Keep total title length ≤ 60 chars where possible. If
<Course Name> WSQ | Tertiary Courses Singapore exceeds 60, the course name is too long — flag it; don't strip the suffix.
Meta description rules:
- 150–160 chars. Must read like a sentence, not a keyword stuffing.
- SG WSQ: name the funding schemes that actually apply to that course (check the funding badges on the product —
WSQ, SkillsFuture Credit, PSEA, UTAP, IBF, HRDF, SFEC, Absentee Payroll, MCES per MMD_CourseImage_Helper_Data::getAllBadges()). Don't claim SFEC if the course doesn't have the SFEC badge.
- MY: must contain "HRDF" or "HRD Corp claimable".
- SG non-WSQ, GH, NG, BT, IN: focus on outcome / audience / city — no funding language.
4b. Detection — curl the title and meta description
curl -sL https://www.tertiarycourses.com.my/<course-slug>.html \
| grep -iE '<title>|<meta name="description"|<h1'
Pass/fail per the table above. Common failures to flag:
- Title ends with
| Tertiary Infotech or | Magento (old default) — must be replaced with the per-store brand suffix.
- TGS- (WSQ) course missing "WSQ" in title.
- MY course missing HRDF in description.
- SG non-WSQ (C-prefix) course mentioning SkillsFuture/WSQ — must be stripped.
- Same title across two countries (e.g. SG and MY identical) — must differ at minimum by the brand-country suffix; ideally also by city/locality token in the course name itself to avoid hreflang-cluster dupe-content suppression.
4c. Where these are set in Magento
- Per-store override: Catalog → Manage Products → [product] → switch Store View → Meta Information tab → Meta Title / Meta Description. Save at the store scope, NOT default, so each country has its own.
- Bulk audit: dump
catalog_product_entity_varchar where attribute_id matches meta_title / meta_description and store_id IN (0,1,…) to find any row still at default (store_id=0) or missing per a country store.
- Reminder: per the auto-memory, CLI
saveAttribute writes at SG scope, not admin — bulk migrations targeting meta fields must set store_id per country and clear stale overrides.
4d. Other on-page items
- H1: should match the course name (no brand suffix in H1 — the suffix is title-only). Per the auto-memory,
product_name is sacred — never mutate it for SEO; the suffix lives in meta_title only.
- Schema.org
Course: every course page should have JSON-LD with Course, provider (Tertiary Courses), hasCourseInstance (next class date), offers.price (in local currency). Check <head> for <script type="application/ld+json">. Use Rich Results Test (https://search.google.com/test/rich-results) to validate — curl strips <script> tags so you cannot detect JSON-LD with curl alone.
- Internal linking: every course should be reachable from a category page within 2 clicks of homepage. Check via Screaming Frog crawl depth.
Schema detection note: curl and WebFetch cannot reliably see JSON-LD because they strip <script>. Use the Rich Results Test, or in the browser run document.querySelectorAll('script[type="application/ld+json"]').
5. Country-specific signals
| Country | Signal to verify |
|---|
| SG | en-SG hreflang; SGD prices; SG address in footer; WSQ / SkillsFuture branding for TGS- SKUs only; Google Business Profile at SG address; +65 phone in footer |
| MY | en-MY (and optionally ms-MY); MYR prices; KL/PJ address; HRDF (HRD Corp) branding; +60 phone |
| GH | en-GH; GHS prices; Accra address; +233 phone; no funding branding |
| NG | en-NG; NGN prices; Lagos/Abuja address; +234 phone; no funding branding |
| BT | en-BT; BTN prices; Thimphu address; +975 phone; no funding branding |
| IN | en-IN; INR prices; India address; +91 phone; no funding branding |
These are E-E-A-T trustworthiness signals for local rankings. Missing local NAP (Name/Address/Phone) is the single biggest signal Google uses to not rank a country variant.
6. Tooling
- Google Search Console — one property per country domain (4 properties total). Verify all four are claimed and the Sitemaps tab shows the sitemap.
- Rich Results Test — for schema (curl can't see JSON-LD).
- PageSpeed Insights — Mobile + Desktop per country URL.
- Screaming Frog (if licensed) — crawl with custom robots, set "JavaScript rendering" on, export hreflang errors. This is the one paid tool worth running.
- Magento admin → Catalog → URL Rewrite Management — find rewrite chains and 302s that should be 301s.
Output format
EXECUTIVE SUMMARY
- Top 3 issues blocking organic growth on this site
- Quick wins (config-only, no code) the user can apply today
CROSS-STORE / HREFLANG
- Issue (Severity High/Med/Low): Evidence, Impact, Fix
INDEXABILITY
- ... same shape
CORE WEB VITALS
- ... same shape, with PageSpeed scores per country
ON-PAGE & SCHEMA
- ... same shape, with sample course URL evidence
PRIORITISED ACTION PLAN
1. Blocking / critical (broken indexation, etc.) — owner + ETA
2. High impact (CWV, schema, hreflang fixes)
3. Quick wins (config toggles)
4. Long-term (content strategy, link building)
Invariants — these are HARD RULES, never recommend changing them
These two are non-negotiable. Past audits suggested touching them and broke
the storefront / clobbered structured data. Flag any audit finding that
depends on changing either as "won't fix" with a one-line reason.
-
Course title (product name) is immutable.
The catalog_product_entity_varchar.value for name is the H1 on the
product page, the name in JSON-LD Course schema, the anchor text on
every internal link, the line item on every order/registration, the
subject in transactional emails, and the source of the URL key. Do NOT
recommend rewriting it for SEO keyword density, brand suffix injection,
title-length compliance, or duplicate-content disambiguation. If a meta
improvement requires a longer/shorter title, put the change in
meta_title only — never in name. Same rule for course sku.
Reference: [[feedback_product_name_is_sacred]] in .claude/memory/.
-
Category URLs are short-path only — /<url_key>.html, never
/parent/.../<url_key>.html.
Enforced by the MMD_FlatCategoryUrl module (class-rewrite of
Mage_Catalog_Model_Url::getCategoryRequestPath). Do NOT recommend:
- Re-enabling the deep category path for "breadcrumb URL parity" or
similar SEO intuitions — deep paths are explicitly unwanted.
- Disabling
MMD_FlatCategoryUrl or reverting it to stock behavior.
- Setting
catalog/seo/product_use_categories = 1 to "improve internal
link signals" — products are also short-path (no category prefix).
- Adding observers / class rewrites that prepend parent path on
_refreshCategoryRewrites or getCategoryUrlPath.
If a recommendation would change a URL, it must result in a flatter
URL, not a deeper one. Existing long-path deep-path rewrites stay as
301 sources (save-rewrites-history) — they preserve link equity.
-
Every active category MUST resolve at the clean /<url_key>.html
with NO collision suffix — -1, -2, … -33 on a live category URL
is a BUG, not a valid URL.
The canonical request_path of an enabled (is_active=1) category must
be exactly <url_key>.html. A suffixed canonical
(aws-practice-exams-33.html) means the clean base path is squatted by
a stale rewrite and the flat-URL job kept bumping the suffix on every
reindex. This is the one case where you DO clean up rewrites:
- A save-history redirect that squats a base url_key needed by a
different live category must yield. A 301 from a former url_key of
category X that now blocks the live category Y from claiming
/<url_key>.html is link-equity-negative and breaks Y's canonical —
delete (or re-point) the squatter so Y gets the clean path. This is
the exception to "never delete redirects": deep-path → flat redirects
are kept; base-key squatters that collide with a live category are not.
- Suffix explosion is real on this install. Repeated
reindex/force-flatten runs accumulate temporary
XXXXXXXX_TIMESTAMP
id_path save-history rewrites (thousands of them) that each bump the
next canonical suffix. Purging the stale temp-suffix redirect chain
for a category lets its canonical fall back to the clean base path.
- Detection (run this in every audit): enumerate every
is_active=1 category's canonical rewrite per store and flag any whose
request_path matches -[0-9]+\.html$, plus any that 404 or 301 when
fetched. A genuinely duplicate url_key between two live sibling
categories is a content fix (rename one url_key), not a suffix.
SELECT r.store_id, r.category_id, r.request_path
FROM core_url_rewrite r
JOIN catalog_category_entity_int a
ON a.entity_id = r.category_id AND a.store_id = 0
AND a.attribute_id = (SELECT attribute_id FROM eav_attribute
WHERE attribute_code='is_active' AND entity_type_id=3)
WHERE r.product_id IS NULL AND r.options IS NULL
AND r.id_path LIKE 'category/%' AND a.value = 1
AND r.request_path REGEXP '-[0-9]+\\.html$';
awk '{print "url = \"https://www.<host>/" $0 "\""; print "output = \"/dev/null\""}' \
/tmp/cats.txt > /tmp/cfg
curl -Z --parallel-max 12 -s -K /tmp/cfg \
--write-out '%{http_code} %{url_effective} -> [%{redirect_url}]\n' \
| grep -v '^200 '
Disabled (is_active=0) categories and other-store categories correctly
404 on a given host — do NOT flag those.
Anti-patterns — don't recommend
- Don't recommend rel=canonical from country domain to SG. That suppresses the country domain. Self-canonical per store.
- Don't recommend
noindex on country stores to "fix duplicate content". Hreflang is the right tool for that.
- Don't recommend disabling Magento URL rewrites. The whole site depends on them.
- Don't recommend pushing PageSpeed score above 70 on mobile without acknowledging that Magento 1 + Ultimo has a structural ceiling around 60-75 without a CDN + heavy theme work.
- Don't suggest the user buy SEMrush/Ahrefs without first checking what they already have access to.
- Don't add "WSQ", "SkillsFuture", or any subsidy keyword to SG non-WSQ (C-prefix) course titles or descriptions — it dilutes WSQ-course relevance and misrepresents non-funded courses.
- Don't add "HRDF" / funding language to GH / NG / BT / IN course meta — no such scheme applies, and it's misleading.
- Don't drop the
| Tertiary Courses <Country> brand suffix to save title characters. If the title is too long, the course name is too long — flag the course, don't strip the brand.
- Don't write per-store meta at the default scope (
store_id=0) — it will leak across all stores. Always save Meta Title / Meta Description at the country store scope.
