On-page SEO in 2026 looks different than it did in 2023. Google now rewrites roughly 76% of title tags and 63% of meta descriptions, AI Overviews are cutting organic click-through by up to 61% on queries where they appear, and the content depth signal has fully replaced raw word count. The pages that win are the ones that match search intent precisely, ship clean schema, lead with the answer, and link internally to the right next step.
What on-page SEO actually means in 2026
On-page SEO is everything you control on the page itself — titles, meta descriptions, headings, body copy, images, internal links, schema markup. It is the layer where you communicate to Google, Bing, and AI assistants what a single URL is about and why it deserves to rank.
It sits between technical SEO (can the bots reach and render the page) and off-page SEO (does the rest of the web vouch for it). If technical is plumbing and off-page is reputation, on-page is the actual product. You can have brilliant plumbing and great reputation and still lose if the page itself does not match the searcher’s question.
Why on-page SEO has changed in 2026
Three things shifted the rules in the last 18 months. AI Overviews now appear on a meaningful share of informational queries and intercept clicks. Google’s title and meta rewrite rates climbed to the point where you cannot assume your tags will show up as written. And the September 2023 Helpful Content Update, folded into the March 2024 core update, raised the floor on what counts as on-page quality.
The implications are concrete. You cannot write to satisfy your title-length checker any more — you have to write to survive a rewrite. You cannot rely on a featured snippet position any more — you have to either land inside the AI Overview or aim for the queries that do not trigger one. And you cannot pad word count any more — depth and intent match are the signal, length is just a side effect.
The six pillars of on-page SEO in 2026
Title & meta
Written to survive Google’s rewrite, optimized for CTR within pixel limits.
Heading hierarchy
H1 + nested H2/H3 that mirrors how people actually search.
Content depth & intent
Topical coverage and intent match, not raw word count.
Internal linking
Descriptive anchors that pass authority to money pages.
Image SEO
Modern formats, alt text, LCP-aware loading.
Schema markup
Article, Organization, Person, Product — and entity graphs for LLMs.
Pillar 1: Titles and meta descriptions when Google rewrites most of them
Zyppy’s Q1 2025 study put the title rewrite rate at around 76%, up from 61.6% in their original 2021 study and the figure widely cited until last year. Ahrefs’s 953,276-page analysis found Google is 57% more likely to rewrite titles that are too long, and when it does ignore the title, it uses the H1 50.76% of the time. Seven point four percent of top-ranking pages have no title tag at all and Google still ranks them.
What you should actually do:
- Write the H1 with the same care as the title. If Google rewrites the title, your H1 is the fallback half the time
- Keep titles in the 40–60 character range. Backlinko’s analysis of 4M search results found titles in that range get 8.9% higher CTR than titles outside it
- Front-load the keyword. Titles with the target term earlier are less often rewritten
- Stay under ~600 pixels. Roughly 525–535 px on desktop is where truncation starts
- Write multiple variants for high-value pages. Test them via Search Console’s average position and CTR over a 28-day window
For meta descriptions the rewrite picture is similar. Ahrefs found 62.78% of meta descriptions get rewritten; Portent’s analysis put it at 71% on mobile and 68% on desktop. The brutal Ahrefs finding: staying within character limits does not reduce the rewrite rate. Properly-sized descriptions are rewritten almost as often (63.69%) as oversized ones (61.46%).
The practical conclusion is that meta descriptions are no longer about controlling SERP copy. They are about giving Google reasonable starting material and influencing CTR when they do happen to be used. Write them anyway, keep them under ~155 characters on desktop and ~120 on mobile, but stop spending hours on them.
Pillar 2: Heading hierarchy
John Mueller has said it bluntly: “Your site is going to rank perfectly fine with no H1 tags or with five H1 tags.” Google does not penalize multiple H1s. Headings are primarily an accessibility and content-structure signal. They still matter, but not the way the early-2010s playbook suggested.
What headings actually do in 2026:
- Map your subtopics for both humans and LLMs. A well-structured H2/H3 hierarchy is one of the strongest signals LLMs use to extract sections for citation
- Win featured snippets and AI Overview placements by aligning question-style H2s with the actual queries searchers use
- Improve dwell time by making the page scannable, which indirectly helps rankings via behavioral signals
One H1 per page (for clarity), question-style H2s where they make sense (“How does INP affect SEO?”), and short H3s for sub-claims. That is the entire heading playbook in 2026.
Pillar 3: Content depth and search intent
Backlinko’s 11.8M-result study found that the average first-page result is around 1,447 words. That is correlation, not causation. The pages that rank are typically thorough, and thorough usually means longer — but length itself is not the signal. Depth is.
Depth means topical coverage: do you address every subtopic and follow-up question a searcher might have, including the obvious objections and edge cases? Tools like Clearscope, Surfer, and Frase quantify this. The single biggest ranking blocker I see on otherwise-good content is intent mismatch — writing a how-to guide when the SERP is full of comparisons, or a comparison when the SERP is full of definitions.
The four canonical search intents both Ahrefs and Semrush classify against:
| Intent | What the searcher wants | SERP signals | Right content format |
|---|---|---|---|
| Informational | To learn or understand | People Also Ask, AI Overview, featured snippets | Guide, definition, how-to |
| Navigational | To reach a specific site | Knowledge panel, brand sitelinks | Brand pages, login pages |
| Commercial | To compare options | Listicles, “best of” pages, review carousels | Comparison, “best X for Y” |
| Transactional | To buy or sign up | Shopping ads, product carousels, pricing pages | Product, pricing, demo request |
The fastest on-page win on any underperforming page is to check the top 10 results, identify the dominant intent, and reformat your page to match. If five of the top ten are listicles and your page is a definition guide, length and depth will not save you.
Pillar 4: Internal linking
Internal links are the most under-used on-page lever I see in practice. Ahrefs’s analysis of 1M SERPs found that external links correlate more strongly with rankings than internal links, but internal links matter disproportionately for distributing authority to the pages you most want to rank.
The rules are simple:
- Use descriptive anchor text. “Click here” and “learn more” pass minimal context to Google. “Technical SEO audit” passes a lot
- Link from contextually relevant pages. An internal link from a related article carries more weight than one from a global navigation menu
- Every important page should have at least 3–5 internal links pointing at it from other relevant pages
- Audit anchor text diversity. Too many identical anchors look unnatural; vary the phrasing while staying topically tight
- Fix orphan pages. Indexable URLs with zero internal links are signaling to Google that you yourself do not think they matter
If you are doing only one piece of on-page work this quarter, audit your internal linking. The ROI is consistently underrated.
Pillar 5: Image SEO
The 2024 HTTP Archive Web Almanac showed WebP at 12% of web images, AVIF at 1.0% (up 386% since 2022), and JPEG falling from 40% to 32%. The same study noted that a “huge number” of <img> elements still have missing or meaningless alt text — one of the largest accessibility-and-SEO gaps on the web.
Web image format share, 2024
Share of all images served on the web — HTTP Archive Web Almanac
The image SEO checklist that actually moves the needle:
- Serve WebP or AVIF with JPEG fallback for older browsers. WebP cuts file size by 25–35% versus JPEG; AVIF by 50% or more
- Write descriptive alt text — describe the image content, include keywords only if they are genuinely relevant
- Lazy-load below the fold only. Lazy-loading your LCP image is a known anti-pattern that kills Largest Contentful Paint
- Use
fetchpriority="high"on the LCP image to signal to the browser that this is the priority resource - Specify width and height on every image to prevent CLS
- Compress aggressively. The visual difference between 85 and 65 quality is usually invisible; the byte difference is meaningful
Pillar 6: Schema markup for on-page
JSON-LD schema adoption hit 41% of pages in 2024 and continues climbing. The 2026 story is that schema is no longer just for Google rich results — Microsoft’s Fabrice Canel publicly confirmed at SMX Munich in March 2025 that Bing Copilot uses schema for LLM understanding, and Google followed with similar language in April 2025.
Note that Google deprecated FAQ rich results in May 2026 and HowTo rich results in 2023, but it still ingests the structured data. The visible SERP feature is gone; the understanding signal remains.
The highest-leverage on-page schema in 2026:
- Article with author, datePublished, dateModified, image — feeds both freshness signals and LLM source verification
- Organization with sameAs links to LinkedIn, Crunchbase, Wikipedia. This is how AI assistants disambiguate your brand
- Person for author bylines — credentials, affiliations, professional profile links. E-E-A-T at the markup level
- Product with offers, reviews, brand — the only path to product rich results in 2026
- Breadcrumb — small, but consistently respected by Google
- FAQPage and HowTo — no longer rendered as rich results, but still useful for AI understanding
The technique that distinguishes good schema from token schema: connect entities via @id references in a single JSON-LD graph. Article → author → Organization, all linked. Disconnected snippets are tolerated; a connected entity graph is rewarded.
The AI Overview problem and what to do about it
Seer Interactive’s September 2025 analysis of 25 million impressions found that organic CTR on AI-Overview queries fell from 1.76% to 0.61% — a 61% drop. Paid CTR fell 68%. Ahrefs’s December 2025 update of their tracking study put the CTR drop for the top-ranking organic result at 58% when an AI Overview is present, up from 34.5% in April 2025.
The defense is twofold. First, target queries that do not trigger AI Overviews — commercial and transactional intent queries are still mostly AIO-free. Second, optimize the queries that do trigger AIOs to win the citation inside the overview itself. The mechanics are different from classical SEO and that is where GEO / AI search optimization picks up.
The on-page mechanics that increase your odds of being cited in AI Overviews:
- Bottom Line Up Front (BLUF) formatting. Lead with the answer in the first paragraph. LLMs disproportionately extract from the top 30% of an article
- Defined terms. Explain key terms close to where they are first used. Avoid vague pronouns and undefined acronyms
- Short paragraphs and tight lists that are easy to quote
- Citation-friendly facts — specific numbers, named sources, dates. “Recent studies” gets ignored; “Ahrefs’s 953K-page study found 62.78%” gets quoted
Want me to look at your on-page setup?
Most sites I audit have 10–30 quick on-page wins that combined move rankings within weeks. I find them, prioritize them by effort, and tell you straight which are worth shipping.
Book a free 30-min consultationThe on-page audit framework I use
When I do an on-page audit as part of a full SEO audit or as a standalone SEO consultation, the order is:
- Intent match check on each priority URL. Does the page format match the SERP format for its target query?
- Title and H1 audit against length, keyword position, and CTR signals from Search Console
- Content depth gap analysis versus top-ranking competitors — what subtopics are they covering that you are not?
- Internal link audit — orphan pages, weak anchor text, missing links from contextually relevant content
- Schema validation and entity-graph review
- Image SEO scan — formats, alt text coverage, LCP image handling
- AI Overview presence check for target queries — am I cited or am I losing clicks?
Tools I actually use
Common mistakes I see
- Writing for the title-length checker. A title that fits is useless if it does not match search intent. Google will rewrite it anyway
- Padding word count. A 3,000-word version of a 1,200-word answer ranks worse than the tight original. Depth, not length
- Identical anchor text everywhere. Using “SEO services” as the internal link anchor 40 times across the site looks unnatural. Vary the phrasing while staying topical
- Generic alt text. “Image1.jpg” or “logo” misses the entire point. Describe the image content
- Schema spam. Adding every possible schema type hoping something sticks. Google ignores irrelevant markup
- Ignoring AI Overviews. If your money queries trigger AIOs and you are not measuring citation rate, you are flying blind on a meaningful chunk of lost traffic
How on-page connects to everything else
On-page is the layer where everything else either pays off or does not. The pages I work on get the most lift when:
- Technical SEO is clean — crawl, render, schema all working
- Off-page authority is meaningful — backlinks and brand mentions support the page
- GEO formatting is applied — BLUF, defined terms, citation-friendly facts
FAQ
Does word count still matter in 2026?
Indirectly. The pages that rank tend to be thorough, and thorough usually means more words. But length is a side effect, not a target. Optimize for depth — every subtopic and follow-up question a searcher might have — and the right word count usually emerges. Padding a 1,200-word answer to 3,000 words almost always hurts rankings.
Should I worry that Google rewrites 76% of titles?
Yes and no. You cannot rely on your title showing up as written, so the H1 has to be just as strong — Google uses it as the fallback half the time. Keep titles in the 40–60 character range with the target term front-loaded, write strong H1s, and stop spending hours perfecting copy that often does not display. Focus the saved time on intent and content depth.
How do I optimize for AI Overviews?
Two moves. First, use Bottom Line Up Front (BLUF) formatting — lead with the answer in the first paragraph, since LLMs extract disproportionately from the top of an article. Second, ship citation-friendly facts: specific numbers, named sources, dates. Vague claims get ignored; specific ones get quoted. My GEO guide covers this in detail.
What is the single highest-leverage on-page change?
For most sites I audit, it is intent match. Check the top 10 results for your target query — if the dominant format is a comparison and your page is a how-to guide, reformat. Intent match is the lowest-effort, highest-impact on-page change you can make and it is consistently the cheapest win in any audit.
How long until on-page changes show up in rankings?
Four to twelve weeks for most changes to fully reflect, with first signals visible in Search Console within two to four weeks. Title and meta updates show up fastest because they affect CTR immediately. Content depth changes take longer because Google needs to re-crawl, re-rank, and observe new user-behavior signals. Plan in quarters, not weeks.
Do I still need FAQ schema if it does not show in SERPs?
Yes — but for different reasons than in 2022. Google deprecated FAQ rich results in May 2026 and HowTo rich results in 2023. They no longer display as visible snippets in SERPs. But Google and Bing still ingest the markup for content understanding, and Bing Copilot specifically confirmed using it for LLM citations. The SERP feature is gone; the understanding signal stays.
Sources: Zyppy title rewrite studies (2021, 2025); Ahrefs Title Tags Study (953K pages); Ahrefs Meta Description Study; Portent meta description analysis; Backlinko “We Analyzed 4M Google Search Results”; Backlinko 11.8M-result ranking study; HTTP Archive Web Almanac 2024 (Media); Seer Interactive AIO Impact Study (September 2025); Ahrefs AI Overviews CTR Study (December 2025); Google Search Central documentation and blog (2024–2026); web.dev LCP and image guidance; Search Engine Land coverage of John Mueller and Martin Splitt; Search Engine Journal coverage of the Helpful Content Update.
