On-page is where three readers meet: the human deciding whether to click, the human deciding whether to stay, and the machines deciding what the page is about. Google rewrites 61% of your titles and answers more than half of all searches without a click — this module is how to win the part of the SERP you still control.
Why on-page optimization matters more than it gets credit for
On-page optimization sits in an awkward category in modern SEO. It's less glamorous than link building, less measurable than technical SEO, less strategic-sounding than "content strategy." The result: most teams treat it as a one-time exercise at page launch and never revisit. That's the mistake. On-page is a continuous discipline of matching page to intent, page to query, page to user need. Done well, it lifts traffic on existing pages 30–200% without writing new content or earning new links.
By the numbers The page is half yours, half Google’s
What happens to your on-page work in the wild
61%
of title tags get rewritten by Google in the SERP (Zyppy, 80,959 titles) — 76% by a Q1 2025 follow-up.
27.6%
average CTR for position #1 (Backlinko). Position 10: 2.4%. Page one is not the goal; top three is.
58.5%
of US Google searches end without a click (SparkToro/Datos, 2024). Your snippet competes with Google itself.
2×
two readers per page: humans deciding to click and stay, machines deciding what the page is about.
The most important on-page concept is intent matching. Google's job is to give users what they want, which means Google rewards pages that satisfy the intent behind a query — not pages that merely contain the query terms.
The four intent types
Intent
Examples
Best-matching content
Informational
"how to bake bread", "what is encryption"
Educational articles, tutorials, explainers
Navigational
"facebook login", "nike store"
Brand homepages, login pages, specific destinations
Commercial investigation
"best CRM software", "iphone vs samsung"
Comparison guides, reviews, listicles
Transactional
"buy running shoes size 10", "hire web designer"
Product pages, service pages, lead-gen pages
How to determine intent
Look at what currently ranks. If the top 10 are listicles, the intent is investigational/listicle — serving a product page won't rank.
Look at SERP features. People Also Ask, knowledge panels, video carousels signal intent.
Look at modifiers in the query. "Best," "cheapest," "near me," "how to" each anchor intent.
Look at query length. Long-tail queries usually indicate more specific intent; short head queries indicate broader intent.
Intent mismatch is the #1 reason pages don't rank
You can't out-optimize an intent mismatch. A product page can't rank for "best running shoes" against established review listicles — the page format doesn't match what users want. Either change the page format, or target a query whose intent your page actually serves.
Decoder Read the SERP like a spec sheet — tap an intent
The live SERP tells you which intent Google believes — before you write a word
“how to…”, “what is…” — the SERP shows guides, PAA boxes, featured snippets, video
Google is rewarding teachers here. A product page cannot rank on this SERP no matter how optimized — the result types are the verdict, not your content quality.
THE MOVE · Match the dominant format: if the top five are listicles, your 4,000-word essay is fighting the spec. Build the guide; link it to your money page.
The searcher is shortlisting. Google shows third-party evaluators — which is why your own category page rarely ranks for “best [category]”.
THE MOVE · Publish the honest comparison including competitors, or earn placement in the roundups that already rank. Fighting for this SERP with a product page wastes quarters.
Money intent. The SERP fills with product schema, prices, availability — machine-readable commerce signals beat prose here.
THE MOVE · Product/Offer structured data complete, price and stock accurate, category pages crawlable. This is where technical and on-page meet.
Brand names, “login”, specific sites — the SERP is one dominant result plus sitelinks
The user already chose a destination. The only on-page job: make sure YOUR brand SERP is yours — title, sitelinks, knowledge panel.
THE MOVE · Audit your brand SERP quarterly. If a review site or a competitor’s comparison page sits at #2 for your brand, that is an on-page brief, not a shrug.
Title tags and meta descriptions
Title tags
Length: 50–60 characters typically renders without truncation. Pixel width matters more than character count.
Primary keyword early; brand name at end (or omitted for non-brand pages).
Compelling: titles affect CTR, and CTR affects rankings indirectly. "The 17 Best CRM Software Tools (Compared by an Independent Reviewer)" outperforms "Best CRM Software" even if both rank the same.
Unique: every page should have a unique title. Duplicate titles indicate duplicate content issues.
Match user query language, not internal jargon. If users search "CRM," don't write "Customer Relationship Management Platform" in titles.
Google may rewrite your title
Since 2021, Google rewrites about 60% of title tags it displays in SERPs. It pulls from H1, anchor text, structured data, or generates its own. To minimize rewrites: keep title and H1 consistent, keep titles within 60 characters, don't keyword-stuff, match user-facing language.
Meta descriptions
Length: 150–160 characters usually fits.
Not a direct ranking factor, but a major CTR factor.
Include the primary keyword (Google bolds it in SERP), a benefit, and a call to action.
Don't auto-generate from page content — write custom descriptions for top pages.
Google rewrites these even more often than titles — up to 70% of the time. Still worth writing well for the times Google uses your version.
Live simulator Write a snippet, watch it truncate — the pixel truth
Your title competes in pixels, not characters
Title: 0px of ~580px · 0 chars
Description: 0 of ~158 chars
realgrowthmatters.com › learn › on-page-seo
Limits are approximate — Google truncates titles around 580px on desktop and descriptions around 155–160 characters, and rewrites snippets per query. The simulator measures real pixel width of your words, which character counters cannot do.
RGM EXPERT TRICK
Write the title for the rewrite, not against it
Google rewrites most titles, so we stopped fighting and started front-loading: the entity and the intent live in the first ~50 characters, because that is the part Google keeps when it truncates or recomposes.
Brand goes last, separated with a pipe — the first thing Google drops, the last thing we miss. Boilerplate (“Home”, category chains) never enters the tag; that is what triggers rewrites in the first place.
Then we audit survivals quarterly: GSC by template, comparing our tag to what the SERP actually showed. Templates with high rewrite rates get rewritten by us first.
WHY IT’S RARE · Most teams treat the rewrite as weather. Auditing WHICH titles Google rewrites turns it into feedback — Google is telling you your pattern is weak.
Case studies · SearchPilot SEO split tests · title tags move real traffic
+15%brand moved to front of titles+11%filler words removed-16%numbers removed from titles
SearchPilot runs controlled SEO A/B tests on large sites — half the template gets the change, half does not, and the delta is measured against forecast. Their published title-tag tests cut both ways: fronting a well-known brand lifted organic traffic 15%+, deleting filler words lifted 11%, while stripping numbers from article titles cost an estimated 16%. The often-cited lesson: title rules are hypotheses, not laws — the same edit that lifts one site sinks another, which is exactly why we test per template. (SearchPilot case studies, 10%+ test roundup)
Heading hierarchy
One H1 per page, ideally matching or close to the title tag.
H2s structure major sections; H3s nest within H2s; rarely need to go deeper than H4.
Headings should describe section content accurately — both for users and for search engines parsing topical structure.
Include semantic keyword variations across H2/H3 to cover topical breadth (e.g., for "running shoes" topic: H2 on cushioning, H2 on stability, H2 on neutral vs supportive, H2 on women's vs men's).
Headings affect featured snippet eligibility — well-structured H2/H3 questions get pulled into PAA boxes and snippets.
X-ray Two pages, same content — one outline machines can read
Heading hierarchy is the document’s API
✓ Parseable outline
H1 · Standing desk buyer’s guide
H2 · How we tested 14 desks
H2 · Best overall: the shortlist
H3 · Best under $400
H3 · Best for small spaces
H2 · Common buying mistakes
H2 · FAQ
✗ Styling, not structure
H1 · Welcome!
H1 · Standing desks (second H1)
H3 · Our story (skipped H2)
H2 · BUY NOW AND SAVE
H4→H2→H4 · sizes chosen for font size
H2 · (empty — spacer heading)
The left page hands every machine reader — Googlebot, screen readers, AI crawlers — a clean tree: one H1, descending levels, headings that summarize their sections. The right page uses headings as a font-size picker. Same words, unreadable structure.
Content depth and information gain
The 2010s wisdom was "long content ranks better." That's now misleading. Length is correlated with rankings but not causal — comprehensive content tends to be longer, but length without comprehensiveness is just word count.
Information gain
Google patents and the Search Quality Rater Guidelines reference "information gain" — how much new useful information your page adds beyond what already ranks. If you write the same advice as the top 10 results, you're commodity. If you bring new data, new perspective, new case studies, or original research, you have information gain.
How to add information gain
Primary research (surveys, data analysis, original case studies).
Updated data (refresh statistics, replace outdated references).
Counterintuitive findings (point out where conventional wisdom is wrong, with evidence).
Visual originality (diagrams, infographics, screenshots, video walkthroughs).
Calculator tools, interactive elements, or templates.
Comprehensiveness without padding
Cover the topic well, but don't pad. Pages that ramble lose engagement signals and rank lower than pages that are focused, deep, and respectful of the reader's time. Better to write 1,200 excellent words than 4,000 mediocre ones.
Field data What position is actually worth
Average organic CTR by position — the cliff is real
Position 1
27.6%
Position 2
15.8%
Position 3
11.0%
Position 5
6.3%
Position 10
2.4%
Source: Backlinko organic CTR study. Moving 10→3 is worth ~4.5× the clicks; 3→1 is worth 2.5× again. On-page work compounds because it moves positions AND raises CTR within a position.
Entity optimization and semantic search
Modern search has moved past pure keyword matching to entity-based understanding. Google identifies entities (people, places, things, concepts) in queries and pages, and looks for content that covers the relationships between those entities.
Entity optimization tactics
Use natural language that mentions related entities. For a page about "Apple iPhone," naturally include "App Store," "iOS," "Tim Cook," "Apple Pay," etc.
Link to Wikipedia / Wikidata pages for entities you reference (sparingly — build authority through citation, not over-linking).
Use structured data (schema.org) to disambiguate entities — especially for people, places, organizations.
Build content clusters around entities and their relationships, not isolated keywords.
Images, video, and multimedia
Image SEO
Descriptive filenames: running-shoe-cushioning-test.jpg not IMG_4271.jpg.
Alt text: descriptive, useful for screen readers, includes relevant keywords naturally (don't stuff).
Format: WebP or AVIF for smaller file sizes. Provide JPEG fallback.
Compression: target <100KB for content images, <30KB for thumbnails.
Width/height attributes set to prevent layout shift.
Lazy loading below the fold; loading="lazy" attribute.
Image sitemap for image-heavy sites.
Structured data (ImageObject in product schema, recipe schema, etc.) for rich results.
Video SEO
Host on a CDN-backed video platform (YouTube, Vimeo, Mux). Self-hosting MP4s without infrastructure causes performance issues.
Video schema markup (VideoObject) for rich results.
Transcripts on-page; search engines and AI search increasingly value text accessibility of video content.
Title, description, tags for YouTube videos that drive cross-domain visibility.
Embed key video on the page where relevant; video engagement boosts dwell time.
RGM EXPERT TRICK
The SERP is the brief: we spec pages from what already ranks
Before any page is written, we pull the top ten and inventory them: result types, content formats, median word count, what every result covers, and — more useful — what none of them cover.
That last list is the information-gain opportunity: the questions the ranking pages all skip. Our brief leads with those, because matching the SERP earns entry and exceeding it earns the position.
The discipline: we never brief from a keyword tool alone. The tool says what people type; the SERP says what Google decided they meant.
WHY IT’S RARE · Everyone says “analyze the competition.” Almost nobody systematically briefs from the gaps — the absence inventory is the entire edge.
Internal linking at the page level
Internal links pass topical relevance signals and PageRank-equivalent authority. Page-level best practices:
3–10 internal links per content page; more for hub pages.
Descriptive anchor text matching the destination page's primary topic.
Link to deeper pages from authoritative pages.
Link contextually within content, not just in nav and footer.
Audit orphan pages and link to them from related content.
Avoid "link wells" — long lists of internal links at page bottom with no contextual value.
RGM EXPERT TRICK
One intent, one page — and let GSC referee
Cannibalization is rarely two pages with the same keyword; it is two pages with the same intent. Our referee is Search Console: we export queries per URL and flag every query where two URLs split impressions across a quarter.
The split is the tell. Two pages alternating for one query means Google cannot decide — so both rank worse than either would alone.
The fix is editorial, not technical: merge and 301 the weaker page, or sharpen each page’s intent until the queries separate. We re-run the export 6 weeks later and the flagged list should shrink.
WHY IT’S RARE · Most cannibalization audits run on keyword lists and gut feel. The GSC query-overlap export turns it into a deterministic, repeatable report.
Google's Search Quality Rater Guidelines (a 175+ page document Google publishes for human evaluators of search quality) emphasize E-E-A-T as a core evaluation framework. While E-E-A-T is not a direct ranking factor, the signals raters look for correlate with what rankings ultimately favor.
E-E-A-T signals on-page
Author bylines with credentials and links to author bio pages.
Author bio pages with full credentials, publications, and external links to author social or work.
Date published and date updated — honestly maintained.
Source citations to authoritative external sources.
About page that explains who runs the site, what credentials they have, and how to contact them.
Editorial process disclosure — how content is researched, fact-checked, reviewed.
For YMYL (Your Money Your Life) topics — health, finance, legal — first-party expertise is heavily weighted.
Google’s automated ranking systems are designed to present helpful, reliable information that’s primarily created to benefit people, not to gain search engine rankings.
Scorecard Score a page’s E-E-A-T signals — tick what the page can prove
E-E-A-T is not a meta tag — it is evidence on the page
Experience
Expertise
Authoritativeness
Trust
0 / 12 SIGNALS
Tick what this page can actually prove — not what the brand wishes were true. Trust is the family the others report to: a page can survive thin authority, but never thin trust.
SERP feature targeting. Featured snippets get 8–35% of clicks for the queries that have them. Identify which target queries have snippets, study what gets selected, structure your content to match (Q&A sections, definition paragraphs, structured lists, comparison tables).
Topic clusters. Build a pillar page on a broad topic, surround it with detailed subtopic pages linking back. The cluster signals topical authority and improves rankings on competitive head terms.
Content refresh discipline. Top-performing pages should be refreshed every 6–12 months. Update statistics, add new examples, address new questions surfaced in PAA. Refresh often outperforms new content creation for SEO ROI.
People Also Ask mining. The PAA box surfaces real user questions. Answer each PAA question in a structured H3 section on your page; you become eligible for both featured snippet and PAA inclusion.
Author authority building. For YMYL content, invest in author authority — publish on authoritative external sites, get cited by Wikipedia or peer-reviewed sources, link author bio to social profiles with consistent NAP (name, profile).
Information gain auditing. Before publishing, compare your draft against the top 10 ranking pages. Identify gaps and overlaps. Add what they're missing; trim what they all say better.
Internal link audits at scale. Use crawler tools (Screaming Frog, Sitebulb) to identify orphan pages, weak internal linking patterns, and important pages with too few inbound internal links.
Anchor text diversity. Don't link to a page with the same anchor text from every source. Mix exact, partial, branded, and natural-language anchors.
Page-level CWV monitoring. Track Core Web Vitals per page template; identify regressions before they cascade.
Visual content investment. Original diagrams, charts, screenshots, video walkthroughs all increase engagement and likelihood of citation/linking from other sites.
Step by step The RGM rebuild of an underperforming page
Position 8 to position 3 is an on-page job — run it in this order
Pull the query set from GSC.Every query the page gets impressions for, sorted by impressions. The head query tells you the intent to win; the long tail tells you the subtopics to keep.
Re-read the live SERP for the head query.Result types, formats, freshness. If the SERP changed since the page was written — and after a core update it often has — the page is answering last year’s question.
Rewrite title + H1 for intent and pixels.Entity and intent in the first 50 characters, tested in the simulator above. H1 agrees with the title without duplicating it verbatim.
Close the information gaps.Add what the SERP’s winners cover that you don’t, then what nobody covers. Cut sections the query set never touches — addition without subtraction bloats.
Stamp the evidence.Experience signals: original images, data, named author, honest dates. One pass through the E-E-A-T scorecard above.
Rewire internal links.5-10 contextual links from topically-related pages with descriptive anchors. The page also links out to its cluster — on-page includes the page’s neighborhood.
Measure at 4 and 8 weeks.Position, CTR, and query count in GSC. A rebuilt page that gains queries but not position needs authority (module 5); one that gains position but not CTR needs a better snippet — loop to step 3.
Common mistakes
Optimizing pages without first identifying intent — the page may not match what users want at all.
Keyword stuffing titles, headings, and body copy.
Writing content to a word-count target instead of a comprehensiveness target.
Generic anchor text ("click here," "read more") for internal links.
Missing or generic alt text on images.
Treating on-page as a one-time launch task; never refreshing.
No author bylines or credentials on YMYL content.
Duplicate or near-duplicate title tags across the site.
Meta descriptions auto-generated from first 160 characters of body content.
Top-ranking competitors are listicles but you serve a product page targeting the same query.
Internal links only in nav and footer, never contextually within content.
Stuffing keywords into FAQ schema sections that don't match visible content.
Ten questions, CASE method (Context · Analysis · Strategy · Execution). Pass at 90% to unlock this module’s completion passcode — retake as many times as you like.