Header Tags & Content Structure

Headers are the bones of your page. They tell search engines what each section is about, they tell screen-reader users where to jump, and they tell skim-readers (i.e., everyone in 2026) whether to keep scrolling. Most pages get header structure wrong because they treat H tags as styling shortcuts instead of semantic structure.

TL;DR

One H1 per page, matching the title’s intent. Multiple H1s aren’t a fatal SEO sin, but they confuse the document outline algorithm and dilute your strongest semantic signal.
Headers must form a logical hierarchy: H1 to H2 to H3, no skipping levels. Skipping levels breaks the accessibility tree, which Google’s quality models now read alongside HTML structure.
Headers are a moderate ranking factor and a strong scannability factor. Google has confirmed they help with topic signals; user behavior data confirms scannable pages with descriptive headers earn lower bounce and higher engagement.

The mental model

Think of header tags like the chapters and sections of a textbook. The H1 is the book title — there is only one, and it tells you what the whole thing is about. H2s are chapters, H3s are sections inside each chapter, H4s are subsections, and so on. A book with 12 random chapter-1s and no section dividers wouldn’t pass an editor; neither should your page.

The “outline” produced by these tags is what assistive technology (screen readers like JAWS, NVDA, VoiceOver) uses to let a user jump around the page. It’s also what Google’s passage indexing system uses to extract sub-sections of long pages and rank them independently for specific queries. A clean outline means more of your page is rankable; a chaotic outline means only the H1’s worth of context is recognized, and the rest of your work is lost in the parser.

A useful rule: if you printed the page’s headers as a table of contents, would it actually represent what’s on the page? If the answer is “kind of” or “no,” your structure needs work.

Deep dive: the 2026 reality

Headers and ranking. John Mueller has stated repeatedly (2020, 2022, 2024) that headers are “helpful” but “not a strong signal.” Practical observation in 2026: H1s and H2s carry roughly the same weight as the body text they live in, weighted slightly higher because they summarize. They are not a magic boost — the top of an H1 is not 10x stronger than a paragraph — but using them well is a meaningful, free improvement.

Passage indexing, launched February 2021, expanded significantly through 2024. It allows Google to rank a section of a page independently for a specific query, even when the page as a whole is about something else. Strong, descriptive H2s and H3s are what passage indexing uses to identify the boundaries of those sections. A 4,000-word guide with 12 well-named H2s is far more passage-rankable than the same 4,000 words under a single H1.

Featured snippets and AI Overviews citations both prefer pages with clear header structure. AI Overviews, AI Mode, Perplexity, and ChatGPT Search all extract headers as candidate question-anchor pairs. A header phrased as a question (“How long should email subject lines be?”) followed by a 50–80 word direct answer is the canonical recipe for both featured snippets and AI Overview citations.

Accessibility as a quality signal. Google’s 2024 Quality Rater Guidelines update added accessibility as a Page Quality factor for the first time. Pages with broken heading hierarchy (e.g., H1 → H4 → H2) fail automated accessibility checks and are flagged by raters. Lighthouse and axe DevTools both surface heading-order issues; raters now reference these in their reports.

Multiple H1s. HTML5 technically permits multiple H1s within sectioning elements (article, section, nav). In practice, Google’s John Mueller said in 2020 it’s “fine” to have more than one. But — most CMSes don’t generate proper sectioning, which means multiple H1s on a typical page just mean two top-level signals competing. The safe rule remains: one H1 per page.

Length and density. A 2025 Backlinko analysis of 11M pages: top-ranking pages averaged 1 H1, 8 H2s, and 12 H3s. Pages with fewer than 4 H2s on long-form content (>1,500 words) ranked materially worse than pages with 8–14, suggesting that under-structured long content is harder for both humans and algorithms to extract value from.

Headers as modular content. The modern approach uses headers not just for hierarchy but for answer-shaped content design: every H2 and H3 is the question, the first paragraph beneath it is the direct answer (50–80 words), and the rest is elaboration. This pattern is what wins featured snippets, AI Overview citations, and the “People Also Ask” expansions.

Visualizing it

flowchart TD
  H1["H1: Best CRM Software for SaaS in 2026"]
  H1 --> H2A["H2: How we tested"]
  H1 --> H2B["H2: The 4 that won"]
  H1 --> H2C["H2: 6 that quietly raised prices"]
  H1 --> H2D["H2: Free-tier comparison"]
  H1 --> H2E["H2: How to choose"]
  H2A --> H3A1["H3: Test methodology"]
  H2A --> H3A2["H3: Pricing data sources"]
  H2B --> H3B1["H3: HubSpot"]
  H2B --> H3B2["H3: Attio"]
  H2B --> H3B3["H3: Pipedrive"]
  H2B --> H3B4["H3: Close"]
  H2D --> H3D1["H3: Comparison table"]

Bad vs. expert

The bad approach

Multiple H1s, skipped levels, decorative use, vague text.

<h1>Welcome</h1>
<h1>About Our Company</h1>

<h3>Why Choose Us</h3>
<p>We are the best...</p>

<h2>FAQ</h2>
<h6>Question 1</h6>
<p>Answer...</p>

<h4>Visit Our Other Pages</h4>

Two H1s confuse the page topic. The H3 has no H2 above it. The H6 under “FAQ” should be an H3. “Welcome” and “Visit Our Other Pages” are content-free headers. Screen readers will produce a broken outline; passage indexing will extract nothing useful; AI Overviews won’t find a citable question/answer pair.

The expert approach

One H1, descriptive H2s as questions or section topics, H3s nested under H2s, plain HTML — no decorative classes pretending to be headers.

<h1>Best CRM Software for SaaS in 2026 (Tested)</h1>

<p class="lede">We re-tested 14 CRMs in April 2026 with fresh accounts and live
pricing. Here are the four that won and the six that raised prices quietly.</p>

<h2>How we tested</h2>
<p>For each tool, we created a fresh account, imported a sample dataset of
1,200 contacts, and ran six common SaaS workflows...</p>

<h3>Test methodology</h3>
<p>Each CRM was tested against the same workflows...</p>

<h3>Pricing data sources</h3>
<p>Pricing was pulled from official pricing pages on April 28, 2026.
We confirmed each price by starting checkout but not completing it...</p>

<h2>The four CRMs that won</h2>
<p>HubSpot, Attio, Pipedrive, and Close stood out for distinct reasons...</p>

<h3>HubSpot: best free tier</h3>
<p>HubSpot's free CRM remains the most generous of any tool tested,
with unlimited users and 1M contacts...</p>

<h3>Attio: best for product-led teams</h3>
<p>Attio's relational data model is the only one in our test that
let us model accounts, opportunities, and product usage in one schema...</p>

<h2>How long should an email subject line be?</h2>
<p>Most analyses agree on 41 characters or fewer. Gmail mobile clips
subject lines at 33–43 characters depending on device width, and
Litmus's 2025 study of 10,084 emails found subjects under 41 characters
had a 22% higher open rate than longer ones.</p>

The H2-as-question format is what wins featured snippets. The first paragraph under each H2 is a direct, citation-ready answer.

/* Style with CSS — never use heading tags for visual size */
h1 { font-size: clamp(2rem, 4vw, 3rem); }
h2 { font-size: clamp(1.5rem, 3vw, 2rem); margin-top: 3rem; }
h3 { font-size: clamp(1.125rem, 2vw, 1.25rem); }
.lede { font-size: 1.125rem; line-height: 1.6; }

Visual hierarchy comes from CSS, semantic hierarchy comes from the tags. Don’t conflate them.

Element	Bad	Expert
H1 count	Multiple or zero	Exactly one, matches page topic
Hierarchy	Skipped levels (H1 to H4)	H1 to H2 to H3 in order
H2 phrasing	”Section 1"	"How we tested” / “How long should…”
Visual styling	Tag-based	CSS-based
Density (long-form)	1–3 H2s	6–14 H2s for >1,500 words

Do this today

Run Lighthouse in Chrome DevTools (F12 > Lighthouse > Accessibility) on your top 10 pages. Flag any “Heading elements are not in a sequentially-descending order” warnings.
In Screaming Frog, click H1 and H2 tabs. Filter to Missing, Multiple, and Duplicate. Each row is a fix candidate.
For pages with multiple H1s, decide which one represents the page’s primary topic and demote the others to H2 or H3.
Open each top-traffic page in Chrome DevTools > Elements, expand the body, and scan the heading flow. If you can’t read the headings as a coherent table of contents, the structure is wrong.
Rewrite generic H2s (“Overview,” “Section,” “More info”) as specific questions or topics that match queries you want to rank for. Use Ahrefs Keywords Explorer > Matching terms > Questions to find candidate phrasings.
Under each H2 phrased as a question, ensure the first paragraph is a direct 50–80 word answer. This is the featured-snippet recipe and the AI Overview citation pattern.
Add HeadingsMap (Chrome extension) and inspect the document outline of your top pages. The visual tree should match the section structure of the article.
For pages > 1,500 words with fewer than 4 H2s, add structural breaks. Aim for an H2 every ~250–400 words on long-form content.
In your CSS, audit any <div class="header">, <p class="title">, or styled spans that visually look like headers but aren’t. Convert them to the proper <h2>/<h3> element if they represent section divisions.
Run axe DevTools (free Chrome extension) on key templates. Fix any “Heading levels should only increase by one” or “Page must contain a level-one heading” issues — these are accessibility AND SEO fixes in one.