How to Improve On-Page SEO

Whitehat SEO • Updated 20 March 2026  |  Explore our digital marketing FAQs

Google shipped three confirmed core updates in 2025 plus a spam update and its first-ever Discover-specific core update in February 2026. Each reinforced the same direction: genuine expertise beats content volume, and mediocre content faces increasingly harsh penalties.

On-Page SEO in 2026: What's Changed and What Actually Matters Now

Google's December 2025 core update broadened E-E-A-T requirements beyond YMYL topics to all competitive categories. AI Overviews now reduce organic clicks by 58% and appear on roughly 30% of US desktop queries. The traditional playbook still works as a foundation, but the rise of AI answer engines has created a parallel optimisation challenge: getting cited by ChatGPT, Perplexity, and Google's AI features matters as much as ranking in position one.

1. Text Relevance Is Now The Dominant Signal (0.47 Correlation)

2025 data shows text relevance (query-to-content match) now has the strongest correlation with rankings at 0.47—beating out traditional factors like domain authority, backlinks, and page speed. This has been building for three years, but it's now the decisive factor.

What changed: In 2023, text relevance ranked third. In 2024, it climbed to first. Now it's not just the top factor—it's far ahead of the second-ranked signal. Pages ranked #1 have a mean semantic similarity score of 0.68 compared to #2-#5 pages at 0.52.

Practical implication: Target a single search intent per page. Stop the practice of "covering everything" on one page. A page that perfectly matches "best CRM for SME marketing teams" will rank higher than a page that broadly covers "best CRM software" and merely mentions SMEs in passing.

Implementation: Write your opening paragraph as a direct answer to the query. Not metadata, not marketing copy—the actual, literal answer. Then support that answer with evidence. Format it as a statement, not a question.

2. E-E-A-T Now Applies To All Competitive Topics (Not Just YMYL)

Google's December 2025 core update officially broadened E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) from YMYL-only topics to all competitive categories. This was previewed in the helpful content guidance update but now it's enforced algorithmically.

What's new: E-E-A-T is no longer a YMYL-only requirement. It's a universal competitive signal. Technical content, product reviews, SEO guides, marketing resources—all now face E-E-A-T scrutiny.

The critical element: Experience. The other three—Expertise, Authority, Trust—can be faked. Expertise can be claimed. Authority can be bought. Trust can be designed. But Experience cannot. You cannot fake having actually used a product, conducted an experiment, or built a solution.

Practical implication: If you're writing product reviews, use the product. If you're writing SEO advice, show your own ranking results. If you're writing software tutorials, provide original screenshots and walkthroughs. If you're publishing financial data, cite primary sources you've verified. This is why Reddit, X, and product forums rank so high now—they're full of raw, unfiltered user experience.

Implementation: Add an author bio that includes relevant experience. Link it to publicly verifiable credentials (your Twitter, portfolio, LinkedIn). Include original photos, proprietary research, or firsthand testing results where possible. At minimum, disclose your methodology and testing approach.

3. AI Overviews Reduce Organic Clicks By 58%—But Create New Opportunities

Google's AI Overview feature (launched as SGE, now rolling out broadly) reduces organic click-through rates by 58% when present. But here's the critical insight: brands cited within AI Overviews earn 35% more organic clicks than brands not cited.

The shift: Getting to position #1 no longer guarantees the most clicks. Instead, getting cited by the AI Overview now creates a 35% organic CTR uplift compared to non-cited sites, even if you're in positions #3-#5.

This fundamentally changes content strategy. You're now optimizing for AI citation, not just traditional SERPs.

How to get cited by AI Overviews:

• Front-load key information in the first 30% of your content. AI models scan the opening section first when generating summaries.
• Use tables and structured formats. Tables are cited 2.5x more often than paragraph text. If you have comparative data or lists, format them as tables.
• Include question-based headers. Phrase your headers as questions ("What is E-E-A-T?") rather than statements. This matches the query pattern and makes it easier for AI to extract relevant sections.
• Provide 30-50 word direct answers at the start of each major section. AI models prefer concise, answerable content. A 50-word direct answer beats a 500-word paragraph.
• Include source citations and data attribution. AI models credit sources within their responses. Being cited increases brand visibility in the overview.

Example of AI-optimized formatting:

AI Overviews don't replace traditional search results—they supplement them. Sites cited in the overview still get significant organic traffic. But non-cited sites see a 58% traffic drop on that query.

4. Core Web Vitals: Still A Tiebreaker, Now Less Forgiving

Core Web Vitals (LCP, INP, CLS) haven't changed in 2026—the thresholds remain the same: LCP ≤2.5 seconds, INP ≤200 ms, CLS ≤0.1. But their role has shifted. CWV is no longer a ranking boost; it's a tiebreaker when two pages have equal relevance.

Current state: Only 48% of mobile websites pass all three Core Web Vitals thresholds. Desktop performance is better (around 65%) but still problematic at scale.

What this means: If you write perfectly relevant content and lose to a competitor with mediocre content, the first thing to check is CWV. If your competitor's site is significantly faster and yours is sluggish, CWV is amplifying their advantage even though it's not a primary ranking factor.

Priority order for 2026:

1. Text relevance (do this first): Rewrite for exact query match. This moves the needle the most.
2. E-E-A-T signals: Add original research, author credentials, firsthand testing, proprietary data.
3. AI Overviews optimization: Reformat for AI citation (tables, question headers, 30-50 word answers).
4. Core Web Vitals: Fix only if you're losing to a competitor and both pages have similar relevance.

Most sites are still optimizing in the reverse order. They're squeezing milliseconds out of LCP when they should be rewriting for text relevance. That's why the impact is so often disappointing.

5. Link Quality Over Link Quantity (Backlinks Remain Stable)

Backlink counts have plateaued as a ranking factor. Google's algorithm is now far more selective about which links matter. A single .edu link pointing directly to your content is worth more than 50 blog directory links.

What changed: Link velocity (how fast new links appear) is now more important than total link count. A site receiving 10 highly relevant links over 2 weeks will rank higher than a site that acquired 100 links over 2 years, assuming similar relevance and authority.

Practical implication: Stop chasing bulk backlinks. Focus on getting cited by relevant, authoritative sources. For B2B companies, that means industry publications, analyst reports, and relevant SaaS directories—not directory listing services.

Implementation: Create original research, data, or tools that journalists and industry analysts want to cite. Then make it easy to link to by providing a quotable summary and direct media contact information.

6. User Experience Signals (RankBrain Still Active, Behavioral Metrics Declining)

Google's RankBrain (AI-based query understanding) remains active and important, but traditional behavioral metrics (CTR, dwell time, bounce rate) have become less reliable ranking signals. Why? Because they're easier to manipulate and mobile user behavior is too variable to be trustworthy.

What matters: Whether users who click your result convert (sign up, download, purchase, subscribe). Google measures this through behavioral aggregation from Chrome and Search Console data. A page that gets clicked 100 times and results in 20 conversions ranks higher than a page clicked 100 times with 5 conversions.

Practical implication: Don't optimize for CTR anymore. Optimize for intent match. A lower CTR with higher conversion intent is more valuable than a higher CTR with mismatched traffic.

7. Topic Authority & Semantic Clustering

Topic Authority (Google's ability to assess whether your entire site—not just one page—is an authority on a subject) has become more important. A site with 50 pages of mediocre financial content doesn't benefit from topic authority. But a site with 10 high-quality financial pages that interlink semantically does.

What this means: Internal linking strategy has changed. You're no longer linking for PageRank flow. You're linking for semantic clustering and topic architecture. If you have a pillar page on "SEO", sub-pages on "on-page SEO", "technical SEO", and "link building" should all link back to the pillar using specific anchor text.

Implementation: Audit your internal linking. Does your linking structure reflect your topic clusters? Are sub-topic pages explicitly linked to their parent pillar? This semantic architecture signals topical expertise to Google's algorithm.

8. What's No Longer Important (Or Lost Its Power)

• Exact match keywords: They help with relevance, but "blue leather shoes" and "leather blue shoes" now rank equivalently if the intent is the same.
• Word count: Longer content doesn't automatically rank higher. A 500-word page that perfectly matches intent will rank above a 5,000-word page that doesn't.
• Header tags (H1, H2, H3): They're semantic markers, not ranking signals. Using the right tag structure helps, but having multiple H1s or skipping H2s won't tank your rankings.
• Meta keywords: Google confirmed they're completely ignored.
• Bolding and italics: Formatting helps readability and can help users find answers faster, but it's not a ranking factor.
• Exact match domains: "cheap-flights.com" doesn't rank higher than "kayak.com" for the query "cheap flights".

A/B Testing Framework For On-Page SEO Changes

Most sites have no idea whether their on-page changes actually move rankings. Here's a framework to measure what works:

Measurement metrics:

• Primary: Average ranking position (check via Search Console and Ahrefs), organic traffic (GA4)
• Secondary: Conversion rate, average session duration, pages per session
• Do NOT use: Bounce rate (mobile bounce is highly variable), CTR (depends on snippet quality, not just rankings)

On-Page SEO Checklist For 2026

Frequently Asked Questions

Should I rewrite all my old content?

No. Prioritize pages that: (1) target high-value keywords, (2) rank in positions 6-15 (most likely to move), or (3) are losing traffic compared to competitors. Pages already ranking #1 rarely need rewrites.

How important is keyword density?

Not important. Google understands semantic meaning, not keyword counts. Using the exact phrase "on-page SEO" 15 times on one page will actually hurt your rankings because it reads as unnatural. Vary your language while maintaining intent clarity.

Can I rank without backlinks?

Yes, increasingly so. New sites in competitive niches will still struggle without backlinks, but on-page relevance is strong enough that highly relevant content from newer domains can rank without external links. The barrier is higher, but it's possible.

What about AI-written content?

AI-written content can rank if it matches intent and includes original elements (data, experience, firsthand testing). Pure AI-generated content with no original research will struggle because it lacks the "Experience" component of E-E-A-T. Use AI to draft, then add original insights and proprietary data.

How often should I update my content?

Google's algorithm doesn't reward arbitrary updates. Update only when: (1) the information is outdated, (2) you have new data or research to add, or (3) your rankings are declining. A cosmetic update without content changes won't help.

Is SEO dead?

No, SEO is more specialized. Generalist tactics (keyword stuffing, buying links, publishing fast) no longer work. But expertise-driven, intent-matched, well-researched content ranks better than ever. The bar for success has risen, but the opportunity is bigger for sites willing to invest in quality.

Final Thoughts

The 2026 SEO landscape rewards depth over speed, expertise over volume, and relevance over tricks. The sites winning now are the ones writing for actual humans with actual expertise, not for algorithms.

Text relevance matters most because it directly measures whether your content actually answers the question. E-E-A-T matters because it ensures that answer comes from someone credible. AI Overviews matter because getting cited is the new competitive moat.

If you implement one thing from this guide, make it this: Rewrite your opening paragraph to be a direct, unambiguous answer to the search query. Not a teaser. Not marketing copy. A literal answer. Pair it with original research or firsthand experience, and you'll outrank most of your competitors regardless of their backlink profile.