SEO Metadata Rules, Sitemap & GSC Operations
Page Title and H1 formulas, fitPageTitle 75-char enforcement, honest Free-to-Try copy rules, sitemap architecture, and Google Search Console operations.
Slloy enforces SEO metadata quality, multilingual hreflang / og:locale completeness, and sitemap integrity via automated build gates (npm run check:meta, npm run check:sitemap, npm run check:sitemap:xml, and npm run check:ld).
1. Title, H1 & Copywriting Rules
Title Formula & 75-Character Hard Cap (fitPageTitle)
- Structure:
`{Primary Query} — {Differentiator}`(target<= 60characters, hard ceilingTITLE_MAX = 75characters). - Why
<= 75characters is strictly enforced (LONG_TITLE_BUDGET = 0intools/check-meta.ts):- Search engines truncate overlong
<title>tags from the right, cutting off brand suffixes or leaving broken parentheses in SERPs.
- Search engines truncate overlong
- Deterministic Truncation (
frontend/site/content/page-title.ts):- If
title.length <= 75, returns unchanged. - Otherwise, scans clause separators (
":",":"," — "," – "," · "," | ",";",";","?","?","!","!",",",",") from right to left and returns the longest complete prefix<= 75chars (naturally dropping trailing| Slloy AI Blogor secondary subtitle clauses without an ellipsis). - Only if no clause boundary fits does
hardCut()drop any unclosed bracket (dropOpenBracketfor(,(,[,【,«,“), cut at a word boundary (or character boundary for CJK), and append"…".
- If
H1 & Honest "Free to Try" Copy Rule
- H1 Structure:
`{Query} {Category}`— directly states what the tool or preset produces (e.g.,"Seedream 5.0 AI Image Generator","Amazon White-Background Product Photo Prompts"). - Honest "Free to Try" Rule:
- Slloy grants
100free credits on signup and10credits on daily check-in, which allows users to try any image model and low-cost video models (wan-fast,wan-i2v-flash,wan-i2v,mini) for free. - Copy and meta descriptions must state "Free to try with 100 signup credits" (or direct action verbs like "Generate Image"), never deceptive clickbait like
"100% Free Unlimited No Login".
- Slloy grants
2. Build-Time Metadata Gate (tools/check-meta.ts)
Running npm run check:meta inspects all 1,094+ pre-rendered HTML files in dist/ (decoding HTML entities via decodeEntities before measuring lengths) and fails CI on any of the following 11 issues:
| Rule ID | Condition Checked | Purpose |
|---|---|---|
missing-title / missing-description | Every indexed page (!SKIP.has(file)) must have non-empty <title> and <meta name="description">. | Prevents blank SERP snippets. |
placeholder | Neither title nor description may contain { or }. | Catches uninterpolated i18n template variables. |
duplicate-title | No two pages within the same locale may share the exact same <title>. | Catches template variable bugs that collapse multiple pages into one title. |
long-title | title.length <= 75 (TITLE_MAX = 75, LONG_TITLE_BUDGET = 0). | Zero tolerance for SERP title truncation. |
copy-of-english | On non-English pages (locale !== "en"), title and description cannot both be character-for-character identical to the English page at the same path. | Catches untranslated pages across both Latin and non-Latin scripts. |
unlocalized | On non-Latin script locales (zh, tw, ja, ko, ru, ar, he), Latin script letters (\p{Script=Latin}) in description (> 20 chars) must be < 70% (LATIN_MAX = 70). | Catches English descriptions leaked into CJK/Cyrillic/Arabic/Hebrew pages while allowing model names (Seedream, Qwen-Image). |
missing-og-locale / bad-og-locale | Every indexed page (!noindex) must have <meta property="og:locale"> matching /^[a-z]{2,3}_[A-Z]{2}$/ (e.g., zh_CN, en_US, never BCP 47 zh-CN or bare en). | Valid OpenGraph locale format. |
missing-hreflang / missing-x-default | Every indexed page must include <link rel="alternate" hreflang="..."> tags (matched case-insensitively on hrefLang/hreflang) including hreflang="x-default". | Prevents Google from treating 14 locale versions as duplicate content. |
3. Sitemap Architecture & Indexing Posture
Split Sitemap Index & Real lastmod Dates
/sitemap.xml serves a Sitemap Index referencing 3 sub-sitemaps (core pages, blog articles, and prompt presets):
- Canonical URLs Only: Zero
.htmlextensions and zeroindex.htmlsuffixes (prettyIndex()normalizes directory indexes, verified across allcanonical,hreflang, and<loc>tags). - Authentic
lastmodTimestamps (check:sitemap:xml):- Static core pages read real modification dates from
seo/pages-lastmod.json(only populated when a page's content meaningfully changes). - Blog articles use the article's
updatedordatefield. - Prompt pages use the database record's
updated_attimestamp. - CI guard
npm run check:sitemap:xmlfails the build if all URLs share the samelastmoddate (preventing the"all-pages-today"anti-pattern that causes Google to ignorelastmod) or if any date is in the future.
- Static core pages read real modification dates from
Parameter Page Indexing Posture & robots.txt
- Allowed & Canonical-Converged:
?style=and?mode=are allowed inrobots.txtand converge via<link rel="canonical">to the clean base URL. - Invalid Filter Values Return
404: Requests to/promptswith unknown values for known filter keys (?family=invalid,?use=invalid,?platform=invalid,?season=invalid) returnHTTP 404(hasUnknownPromptFilter), and empty filter combinations output<meta name="robots" content="noindex,follow">. - Blocked for Crawl Budget:
?model=query parameters on the generator remain disallowed inrobots.txtbecause each model already has its own dedicated static landing page.
4. Google Search Console (GSC) Verification & Operations
Domain Property & Verification
- Property Type: Domain Property (
slloy.com) in Google Search Console, coveringhttps://slloy.comand all locale subpaths. - Verification Method: DNS
TXTrecord (google-site-verification=...) configured in the Cloudflareslloy.comDNS zone. - Sitemap Submission: Submit
https://slloy.com/sitemap.xmlin GSC -> Sitemaps (and import into Bing Webmaster Tools via GSC One-Click Import). - Initial Launch Baseline & Full Multilingual Expansion:
- During the initial
2026-09-30GSC onboarding, sitemap scope started with the full English site plus the 14 locale homepages (~191core URLs) to establish clean indexing withoutindex.htmlredirect noise, and subsequently expanded to the full1,904-URL multilingual baseline tracked inseo/sitemap-baseline.txt.
- During the initial
Automated Daily SEO Health Alert (seo-daily-alert.mts)
GitHub Actions workflow .github/workflows/seo-alert.yml runs daily at 09:30 Beijing Time (npm run seo:alert):
- Checks that total sitemap URL count has not drifted by
> 10%. - Samples
60live sitemap URLs and alerts if404rate exceeds5%. - Samples
60legacy.htmlURLs fromseo/legacy-redirects.csvand verifies that301redirects remain alive. - Supports offline GSC Performance CSV ingestion (
npm run seo:alert -- --gsc "<report.csv>") to detect day-over-day impression drops.
14-Locale Architecture & Edge Negotiation
Supported 14 languages (including RTL Arabic and Hebrew), URL prefix mapping, and Worker-level createLocaleRedirect priority rules.
URL Migration & 301 Redirect Policy
Engineering governance for Slloy's 882 table-driven legacy 301 redirects, 8 mandatory URL rules, 90-day shape cooldown, and post-deploy verification.