Chapter 3: On-Page SEO

Pre-publish checklist for a page you actually want to rank

Work top to bottom; each line has a reason, not just a rule.

1. Title tag — 50–60 characters, primary phrase near the front, no repetition of the brand in the middle. If the page targets a year or a number, put it in the title only when the content genuinely reflects it.
2. Meta description — written as an invitation, not a summary. Include the phrase, the problem and what the reader gets. It is a conversion asset, not a ranking factor.
3. H1 — one per page, states the promise in the reader’s words. Similar to the title but not a copy; the title is written for the SERP, the H1 for the page.
4. URL slug — three to five words, lowercase, hyphenated, stable. Changing slugs later costs redirects, so choose once.
5. Opening answer — the first 40–60 words answer the query directly, before any preamble. This is what gets pulled into featured snippets and AI answers.
6. Structure — H2s that map to the sub-questions a reader would ask, each answering one thing. Depth beats length: match the top results’ coverage, then add the part they all missed.
7. Images — descriptive alt text, keyword-relevant filenames, compressed and served in a modern format, explicit width and height to avoid layout shift.
8. Internal links — three to five contextual links to related pages, with anchors that describe the destination. Add one or two outbound links to primary sources where a claim needs support.
9. Schema — the markup that matches the page type (Article, Product, FAQ, HowTo). Validate it before shipping; invalid markup is worse than none.
10. Mobile check — test on a real phone, not just a resized window: tap targets, no horizontal scroll, readable without zooming.
11. Speed — check field data if you have it, lab data otherwise. LCP under 2.5s and CLS under 0.1 are the thresholds to aim at, not the score someone screenshots.
12. Last look — can a stranger landing here from a search result find what the title promised within ten seconds? If not, the page is not finished.

Interview questions and answer frameworks

Q1: How would you take a page stuck at positions 11–20 into the top three?

Framework: Diagnose in four passes before changing anything. One, match the SERP format — are the top three comparison tables, tools, videos, or long guides, and is your page the same shape? A ranking problem is often a format problem. Two, close the coverage gap — list the sub-questions the top three answer and check which ones you skip. Three, technical and freshness — load speed, schema, last-updated signals, and whether your facts are visibly older than the competition’s. Four, click appeal — at 11–20 you are often seen and skipped, so the title and snippet matter more than they will at position three. If those four come back clean and the content is genuinely equal, the remaining levers are links: a few relevant external links and stronger internal links from your highest-authority pages.

Q2: What is keyword cannibalisation and how do you resolve it?

Framework: Two or more URLs on the same site competing for one query, so the engine alternates between them and neither consolidates. Find it in Search Console: filter the query, open the Pages tab, and look for multiple URLs trading impressions with unstable positions. Resolve by intent: same intent → keep the strongest page, merge what is worth keeping, 301 the rest. Different intent → differentiate properly with distinct titles, headings and angles, then link from the supporting page to the primary one so authority flows in one direction instead of splitting.

Q3: How do you get a development team to prioritise Core Web Vitals?

Framework: Speak in risk and money, not in Lighthouse scores. First, quantify exposure: how many URLs are in the “poor” bucket, and what share of organic sessions and revenue they carry. Second, translate the metric into a user consequence — a slow LCP means people tap back before the page paints, which shows up as lost conversions, not as a red number in a report. If you cite famous latency studies, cite them as directional evidence and say so; the only number that persuades your own CFO is your own before-and-after. Third, make the ask small and specific: this template, this image format, this preload, this sprint. Fourth, agree on the measurement before the work starts, so the follow-up conversation is about the metric and not about whether it worked.

Q4: When is a page not worth fixing?

Framework: When it has no query, no link equity worth preserving, and no audience. Concretely: zero clicks for twelve months, no internal links except a sitemap entry, no backlinks, and no strategic reason to own the topic. In that case merging it into a stronger page or removing it is the higher-value action, because a site full of pages nobody wants dilutes the pages people do want. Say the trade-off out loud — deletion loses any historical signals, so it is a decision for the owner of the site, not a unilateral cleanup by the SEO.