Diagnosis before treatment
Six symptoms, six different diagnoses
Do not prescribe more copy until the failure is named. Inspect the final URL and rendered HTML, then repair the narrowest cause and verify it independently.
Google sees a nearly empty page
- Likely cause
- Critical content depends on client-side code, blocked resources or a failed fetch
- Repair
- Render indexable content in the initial or server-generated response and handle data failures visibly
- Verify
- Inspect the rendered HTML and URL Inspection result, not only the browser screenshot
Several URLs compete for the same query
- Likely cause
- Generated routes differ only by wording, parameters or trailing variants
- Repair
- Keep one useful canonical page, redirect true duplicates and consolidate internal links
- Verify
- Check the declared and Google-selected canonical after recrawl
Pages are discovered but rarely indexed
- Likely cause
- The content adds little beyond a repeated template or targets no distinct decision
- Repair
- Merge overlaps and add original examples, evidence or tools that justify a separate URL
- Verify
- Review each page beside its nearest sibling and explain the unique user outcome
Important pages stay undiscovered
- Likely cause
- Navigation is visual but not built from crawlable links, or pages are isolated
- Repair
- Use real anchor links from hubs and related content with descriptive context
- Verify
- Crawl from the home page and confirm the route is reachable without a search box
Search snippets are confusing
- Likely cause
- Titles and descriptions are duplicated, generic or disconnected from visible content
- Repair
- Write a unique promise that the page immediately fulfils and keep headings aligned
- Verify
- Compare metadata, H1, introduction and actual search intent side by side
Traffic arrives but does not use the product
- Likely cause
- The article answers a keyword while hiding the next useful action
- Repair
- Connect the explanation to a relevant scan, checklist or adjacent decision without interrupting the answer
- Verify
- Track the path from landing page to meaningful product action, not just page views
Content quality is part of technical SEO
Indexable is not the same as worth indexing
A generator can produce dozens of grammatically clean pages from a spreadsheet. If they repeat the same framework and differ mainly by keywords, they create no new reason for a searcher—or a search engine—to prefer them. Technical correctness cannot manufacture information gain.
Give every page a distinct decision, evidence set and working format. A comparison, threat model, clinic and test protocol should not share the same body with nouns swapped. Merge pages when the reader would take the same action after either one.
- One canonical page per primary decision
- Original examples over abstract volume
- Visible authorship and evidence boundaries
- Updates only when the material actually changes
Decision matrix
A compact pre-indexing review
This is a release check for a page, not a promise of ranking. The final column gives the artifact a reviewer should be able to inspect.
| Layer | Question | Common failure | Evidence |
|---|---|---|---|
| Discovery | Can a crawler reach the URL through a normal link? | Orphan page or click handler without an anchor | Crawl path and sitemap entry |
| Rendering | Is the main answer present and stable? | Loading shell, blocked fetch or client-only failure | Rendered HTML and failure state |
| Consolidation | Is one URL clearly canonical? | Parameters and duplicate routes split signals | Canonical, redirects and consistent links |
| Intent | Does the page solve a distinct task? | Keyword variant of an existing article | One-sentence user outcome and comparison |
| Evidence | Are factual claims supported and bounded? | Confident synthesis without primary sources | Source notes and visible limitations |
| Experience | Can mobile and assistive users consume it? | Overflow, blocked zoom or poor semantics | Responsive and accessibility review |
Applied example
Clinic case: 100 pages, almost no impressions
A site publishes a page for every variation of ‘AI website checker’, ‘AI website detector’ and ‘vibe-code checker’. The introductions and cards are nearly identical, and all pages point to the same scanner.
- The routes are technically crawlable and included in the sitemap.
- The pages do not answer materially different decisions or contribute original evidence.
- Internal links distribute attention across competing URLs instead of establishing one strong canonical resource.
- Keeping one transactional checker and a smaller set of distinct editorial articles creates clearer intent coverage.
Plain answers
SEO clinic questions
Do AI-written pages automatically rank worse?+
Search systems evaluate usefulness and policy compliance, not a simple production label. Scaled low-value pages remain a risk whether written manually or generated.
Should every article be server-rendered?+
Critical indexable content should be available reliably to crawlers. The right rendering method depends on the framework and content, but a loading shell that requires fragile client execution is avoidable risk.
Will submitting a sitemap index every page?+
No. A sitemap helps discovery; it does not guarantee indexing or ranking. Page quality, canonicalization and crawl accessibility still matter.