On-Page SEO: Pages That Earn Rankings

Ranking doesn’t start with backlinks. It starts with the page itself. Before Google can rank a page, it has to crawl it, parse it, and decide what it’s about. On-page SEO is the discipline of making that decision easy and favorable.

How Google handles a page

The pipeline, in order:

1. Crawl — the crawler fetches your HTML.
2. Parse — it extracts text, links, structure, and metadata.
3. Index — the page enters the index with its extracted terms.
4. Link processing — link signals are computed.
5. Keyword recall — for a given query, candidate pages are pulled from the index.
6. Ranking — candidates are ordered, and the top ~10 win.

Every on-page technique below exists to serve one of these stages: help the crawler, clarify the topic, or feed the ranking step clean signals.

TDH: Title, Description, H1

The old formula was TDK (Title, Description, Keywords). The meta keywords tag is dead — Google ignores it. The modern version is TDH:

• Title — the single strongest on-page relevance signal. Put your target keyword near the front, keep it readable for humans.
• Description — doesn’t directly rank, but it decides the click-through rate from search results. Write it like ad copy: what the page does and why it’s worth opening.
• H1 — one per page, echoing the title’s topic. Subheads (H2, H3) then lay out the page’s structure.

Every page gets its own TDH, targeting its own keyword. Duplicated titles across pages tell Google your pages are interchangeable — they aren’t, and shouldn’t be.

The tool above, the SEO content below

For AI tool sites, the page layout that works is a split:

• Above the fold: the tool itself. Input box, button, result area. The visitor does the thing immediately.
• Below the fold: substantial text content about the topic — what the tool does, how it works, related concepts, FAQ.

The tool earns the visit and the engagement; the content gives the crawler its 500+ words of indexable text and gives the page something to rank for. They reinforce each other.

And make the trial free without login. A login wall before the first try kills dwell time, which kills engagement signals. Let visitors use it first; ask for the account when they want to save, export, or do more.

Build a keyword tree, not a pile of pages

Structure your site’s keywords as a tree:

• Homepage → the main keyword (the broadest, hardest term)
• Subdirectories / section pages → level-2 keywords
• Deep pages → level-3, long-tail keywords

So for an AI writing site: homepage targets “AI writing assistant,” the /blog/ section targets “AI blog writing,” and individual posts target “how to write a product description with AI.” Each level reinforces the one above it through internal links, and the whole tree tells Google this site owns the topic.

Internal links carry as much weight as you think backlinks do

Internal linking is the most underrated lever on a new site. Backlinks take months to earn; internal links you can build today, and they pass relevance and authority between your own pages.

The practical rules:

• The homepage should link to new pages. Add a “latest pages” section on the homepage, and keep listing new pages there for at least five days after publication. New pages need internal link support to be discovered and to accumulate weight quickly.
• Link contextually within content. When an article mentions a concept another page covers, link it with descriptive anchor text.
• Keep the tree shallow. Every important page should be reachable within a few clicks of the homepage.

Canonical discipline

Every page’s canonical tag must point to itself — its own URL. A common beginner mistake is pointing every page’s canonical at the homepage, which effectively tells Google “all these pages are duplicates of the homepage; ignore them.” One page, one self-referencing canonical.

i18n done right (not machine-translated)

If you want traffic in multiple languages, each language needs its own URLs — typically subdirectories like example.com/es/ or example.com/ja/. Then:

• Add hreflang tags so Google knows these are language variants of the same content, not duplicates.
• Keep per-language robots.txt rules if some languages shouldn’t be crawled yet.
• Localize the keywords. Don’t machine-translate your English keyword list. Research what people actually search in each language — the terms differ, the competition differs, and a direct translation often targets a phrase nobody types.

Machine translation of the page body can be a starting point for scale, but the keyword targeting must be researched per language.

Programmatic pages: only after the site matures

Programmatic SEO — generating hundreds or thousands of pages from templates — is powerful, but premature scaling is how sites get penalized. GeFei’s rule of thumb: publish a first batch of about 10 pages by hand, observe how they perform, and only then decide what to scale. If the first 10 can’t get indexed and ranked, 1,000 template pages won’t either. Scale what works; fix what doesn’t.

The pattern across all of this: make each page unambiguous about its topic, connect pages deliberately, and give the crawler everything it needs in the HTML. On-page SEO is unglamorous work, but it’s the foundation everything else in this series stands on.

---

Series: Build Overseas Websites with GeFei | Previous: Build and Launch: Getting Your Site Live | Next: Getting Indexed by Google

Based on GeFei’s methodology (哥飞): https://github.com/qiayue/opengefei