A Comprehensive On-Page SEO Checklist for 2026

In this article, you’ll get a 17-point on-page SEO checklist that covers what you need to rank in Google and get cited by AI search engines like ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. Each item explains what to do, why it matters, and how to verify it works once your page is live.

✓ Short, descriptive URL

A URL like domain.com/abc/8472a/ tells nobody anything. A URL like domain.com/blog/on-page-seo-checklist/ tells a reader, a crawler, and an LLM exactly what the page is about before they open it.

Use your target keyword as the slug 90% of the time. A few rules to follow:

• Skip dates in the slug. /2024/seo-tips/ looks stale by January.

• Skip word repetition. If the page sits in /mens/, domain.com/mens/shirts/ is cleaner than domain.com/mens/mens-shirts/.

• Skip numbers that lock you in. /20-best-tools/ becomes a problem when you add the 21st.

URL slugs also show up in citation lists across AI engines. When ChatGPT or Perplexity cites you, the URL is part of the source attribution shown to the user. A clear slug is a small trust signal that compounds across thousands of AI responses.

✓ Compelling title tag and meta description

The title tag is what searchers see in Google. It is also what most LLM-generated answer interfaces display when they cite your page. A weak title costs you clicks in both places.

A simple formula that holds up across most blog formats:

Combined: On-Page SEO Checklist for 2026 (17-Point).

For meta descriptions, expand on the title and tell the reader what they will actually walk away with. Keep it under 155 characters so it does not get truncated.

Two smaller details to sweat. Keep your title under 60 characters so it renders in full on mobile. And do not stuff the title with secondary keywords. Modern search engines and LLMs are both bad at parsing keyword salad and good at parsing one clear promise.

✓ Title wrapped in H1

You should have exactly one H1 per page, and it should be your title.

Most CMSs handle this automatically. If you are not sure, view source on the published page and search for <h1>. If your title is sitting inside a <div> or an <h2>, ask your developer to fix the template.

Heading structure matters more than it used to. LLMs parse pages by chunking them along heading boundaries to extract answers. A page with one clear H1 and properly nested H2s and H3s is much easier for an answer engine to break apart and quote.

✓ Lead with the answer

Readers came to your page because your title made them a promise. The first sentence should pay it off.

If the query is “what is X,” define X in the first paragraph. If the query is “how to do X,” show the first step. If the query is “best X,” name your top pick.

This is the inverted pyramid newsrooms have used for a century, and it is the single most important formatting change you can make for AI search. LLMs are biased toward extracting the first concrete answer they find on the page. Burying yours under 300 words of context hands the citation to a competitor.

A useful test. Read your first paragraph in isolation. If a reader could screenshot only that paragraph and walk away satisfied, you have the right opening.

✓ Descriptive subheadings

Headings serve two readers. The human skimming on a phone, and the machine parsing your page for an answer. Both want the same thing, which is a heading that describes its section in plain language.

“Stage three” is bad. “How to launch your first campaign” is good.

Three rules for subheadings that hold up in 2026:

• One idea per heading. If you write “Why X matters and how to do it,” split it into two H2s.

• Use the words a real person would use. If your audience says “AI search,” do not write “generative answer engines.”

• Maintain proper nesting. H2 for major sections, H3 for sub-sections. Skipping levels confuses screen readers and LLM parsers alike.

If a heading feels muddled, run it past a colleague. If they cannot guess what is in the section from the heading alone, rewrite it.

✓ Aligns with search intent and prompt intent

Search intent is the reason behind a query. Prompt intent is the reason behind a prompt sent to an AI model. Both have to match what your page delivers, or you will not rank or get cited.

The classical method still works. Search your target keyword in Google, look at the top 10 results, and reverse-engineer what they have in common. Are they listicles or guides? Product pages or blog posts? Recent or evergreen? Your page should match the dominant pattern, or have a strong reason to break it.

For AI search, the question shifts. People do not search “best CRM for startups” in ChatGPT. They ask “What CRM should a 5-person SaaS startup use if we already use Slack and Notion?” The intent is the same, but the prompt is longer, more contextual, and has more qualifiers.

You can surface these prompts inside Analyze AI. The Suggested Prompts view inside Prompts pulls real prompts that are already triggering brand mentions in your industry, so you can see the actual wording your buyers use with LLMs.

When a pattern emerges across suggested prompts, that is your AI search intent. Restructure your page to answer those exact questions, ideally as H2s or H3s in the body. For one-off prompts you want to research without committing to tracking, the Ad Hoc Prompt Searches feature lets you run a prompt through ChatGPT, Google AI, and Perplexity to see who shows up and what they say.

Aligning with both kinds of intent is the same job done in two channels. We cover the overlap in our GEO vs SEO guide and our 4 Pillars of an Effective SEO Strategy for AI Search.

✓ Covers the topic in full

Aligning with intent is the floor. Covering the topic in full is the ceiling.

A page that fully covers its topic ranks for hundreds of related long-tail keywords, gets pulled into more featured snippets, and is more likely to be cited by LLMs as an authoritative source on the subject.

Three ways to find subtopics worth including:

1. Look at competing pages. Open the top 5 results and list every H2 they use. Subtopics that show up in 3 or more pages are table stakes.

2. Mine People Also Ask. Google’s PAA box is a free intent map for any keyword. Our People Also Ask guide walks through how to scrape and prioritize them.

3. Mine LLM follow-ups. Ask ChatGPT or Claude a question on your topic and watch the follow-up suggestions. Those are the questions LLMs already think are related, and your page should answer them too.

Coverage does not mean padding. A 2,000-word page that answers everything beats a 6,000-word page that buries the answer.

✓ Brings real information gain

Information gain is the unique value your page brings to a topic that no other page brings.

Google described a mechanism for measuring this in a 2022 patent, and it has only become more important since LLMs entered the picture. AI engines reward sources that contribute new data, novel framing, or first-party insight, because their job is to synthesize across many sources and pick the differentiated ones.

Four formats that consistently produce information gain:

• Original data. Run a survey, analyze your product data, or scrape a public dataset.

• Original frameworks. Name a process or a model you use internally.

• Expert interviews. Three quotes from practitioners is more valuable than 1,000 words of summary.

• Case studies with numbers. “We tried X for 90 days and got Y” beats “X is a great strategy.”

If you cannot add information gain to a page, ask whether the page should exist at all.

✓ Showcases experience, expertise, authority, and trust

Google’s E-E-A-T framework is not a ranking factor in the literal sense, but it describes the signals their algorithms look for to evaluate quality. LLMs lean on the same signals when deciding which sources to cite.

Three on-page moves that send strong E-E-A-T signals:

• Surface the author and their credentials in the byline. A name and a title is the minimum. A linked author bio with relevant experience is better.

• Cite primary sources. When you reference a stat, link to the report it came from, not to a blog that referenced the report.

• Show your work. If you tested a tool, share screenshots. If you ran a campaign, share the numbers.

To verify whether AI engines are picking up your trust signals, the Sources view in Analyze AI shows which content types and which domains AI models cite most often in your category. If review sites and original research dominate the citation mix, that tells you what kind of content you need to produce to compete.

Sources view in Analyze AI showing content type breakdown and top cited domains

✓ Reads cleanly

If your prose is hard to read, no human will share it and no LLM will quote it.

A few rules that hold up in any genre:

• Short sentences over long ones.

• Short paragraphs over long ones.

• Concrete words over abstract ones.

• Active voice over passive voice.

• One idea per sentence.

Free tools like Hemingway will flag every sentence that is too dense. Aim for a reading grade of 7 or 8 unless your audience is genuinely technical.

There is also an AI-search reason to keep your prose tight. Answer engines extract sentences and short passages, not paragraphs. The cleaner your sentences are, the easier it is for a model to lift one and use it as a citation.

✓ Passes Core Web Vitals

Core Web Vitals are Google’s metrics for real user experience. In 2025, Google increased their weight as a ranking signal, and they continue to matter in 2026.

The three metrics to care about:

INP replaced FID in March 2024 and is the metric most sites are now failing on, so test it specifically.

Quick wins, in rough order of impact:

• Compress and lazy-load images.

• Set explicit dimensions on every image and embed.

• Defer or remove third-party scripts that block rendering.

• Serve from a CDN.

Run PageSpeed Insights on your page for a baseline, then re-run it after each fix.

✓ Adds structured data where it makes sense

Schema markup is JSON-LD code that tells search engines what kind of content your page is. It powers most of the rich results you see in Google, and there is mounting evidence that AI engines also use it as a parsing aid.

The schema types worth adding for most blog content are Article, FAQPage, and HowTo. Product pages should add Product and Review. Local pages should add LocalBusiness.

If you use WordPress, plugins like Yoast or RankMath generate the JSON-LD for you. Otherwise, Merkle’s free generator gives you the code to paste into your page’s <head>. Validate everything in Google’s Rich Results Test before publishing.

✓ Internal links to and from the page

Internal links pass authority and tell search engines what a page is about. They also keep readers on your site, which is its own ranking signal.

Two passes for every new page:

Pass 1: link from the new page to existing relevant pages. Pick 5 to 10 of your strongest related pages and link to them where they fit naturally in the body. Use descriptive anchor text, never “click here.”

Pass 2: link from existing pages to the new page. Find old pages that mention your target topic and add a link to the new page. The simplest way to find these is a site:yourdomain.com "your topic" search in Google.

For larger sites, a crawler will surface internal link opportunities at scale. The Content Optimizer view inside Analyze AI also flags pages that are losing organic traffic, which often turn out to be the best candidates for new internal links pointing into them.

For more, see our 10 Internal Linking Tips for SEO and Single Page SEO guide. You can also run the free Broken Link Checker to make sure none of your existing internal or outbound links are dead before publishing.

✓ Cites credible external sources

Linking out to authoritative sources does not directly boost rankings, but it does three things that matter.

It builds reader trust. It supports your claims with evidence. And it gives AI engines additional signals about your topic and your reliability as a source. LLMs are trained to weigh content that cites primary research more heavily than content that does not.

The rule is simple. When you make a non-obvious claim, link to the source. Cite the original study, not the blog that summarized it. And do not be afraid to link to sources that include your competitors. Useful is more important than insular.

✓ Optimizes images for search and speed

Three boxes to tick on every image:

• Descriptive filename. core-web-vitals-2026.png beats IMG_4827.png.

• Descriptive alt text. Write what you would tell someone over the phone, not a keyword stuffed string.

• Compressed file size. WebP is the default format in 2026. Tools like Squoosh or TinyPNG will compress without visible quality loss.

If you publish a lot of images, set up automatic compression at the CMS layer. Manual compression is a tax you will eventually skip.

One more move worth making in 2026. Add original diagrams, charts, or annotated screenshots to your page. Original visuals are a strong information-gain signal, and Google has confirmed that images can be cited inside AI Overviews.

✓ Wins featured snippets and AI Overview citations

Featured snippets and AI Overviews are now the two most valuable spots on a SERP. Both pull short, structured answers from the top-ranking pages, and the rules for winning them have converged.

The pattern that works for both:

1. Identify a question your audience asks.

2. Add a clear H2 with the question phrased as a heading.

3. Answer it in the first 40 to 60 words below the heading, in a single tight paragraph or a short list.

4. Expand on the answer in the body that follows.

Look at the example below. The opening paragraph defines the term in one sentence, then the rest of the section adds context. That structure is snippet-friendly and citation-friendly at the same time.

To find featured snippet opportunities you already half-rank for, run your domain through the free Keyword Rank Checker and look for keywords where you sit in positions 4 to 10. Those are the cheapest snippets to steal.

For AI Overview citations, the Citation Analytics view in Analyze AI shows you which of your URLs are getting cited, in which AI engine, and for which prompts. If a competitor is being cited for a prompt you should be winning, that is your next page to optimize.

For more, see our deep dives on how to get mentioned in AI search and how LLMs cite sources.

✓ Tracks how the page performs in AI search

This is the step most on-page SEO checklists skip, and it is the one that tells you whether anything you did actually worked.

Default analytics will show you Google traffic. They will not show you which AI engines are sending you visitors, which prompts you are mentioned in, or which of your pages are being cited.

Three things worth tracking for every page that matters:

1. AI-referred traffic per landing page. The AI Traffic Analytics view in Analyze AI breaks down visitors by source engine (ChatGPT, Perplexity, Claude, Copilot, Gemini), shows engagement per page, and lists the prompts that drove the visit.

2. Citation count per page. Even pages that drive zero AI-referred traffic can be heavily cited, which still influences brand visibility. The Sources view tells you which of your URLs are being pulled into AI answers and which competitor URLs are getting cited instead.

3. Prompt-level visibility. Track 5 to 20 prompts that matter to your business inside the Prompts view, and watch how your visibility moves week over week as you optimize pages.

For a weekly summary that ties everything together, the Weekly Email Digest pulls the biggest movements across visibility, citations, and AI traffic into one short email so you do not need to log in to spot what changed.

If you do not have a tool yet, start with the basics. Filter your Google Analytics referrer report for chatgpt.com, perplexity.ai, gemini.google.com, and claude.ai, and check it weekly. It will not give you citation data, but it will tell you whether AI traffic is showing up at all.

The 17-point on-page SEO checklist, in one place

Here is the full list one more time, so you can copy it into your own template.

Work through the list once for every new page, and once a quarter for every page you want to keep performing.

Keep learning

For more on the topics covered in this guide:

• 4 Pillars of an Effective SEO Strategy for AI Search

• What Is Answer Engine Optimization?

• GEO vs SEO: Key Differences and Similarities

• SEO Content Strategy: 10-Step Breakdown

• How to Use Keywords in SEO: 14 Practical Tips