Heading structure: how H1/H2/H3 levels work and where keywords go
1. What headings tell Google
Many SEO guides call headings “the second most important place for keywords,” and that was true in 2012. In 2026, Google judges topics from full-text semantics and user behavior, so the direct ranking weight of headings has dropped very low.
Here is the counterintuitive part: the direct ranking impact of stuffing keywords into headings is often overestimated. The steady payoff comes from two things: sitelink-style rich results in the SERP, and citations in AI Overviews and large models, which use headings to split sections and decide which passage to cite.
So when you write a heading, first be clear about what the section covers and whether readers would click that section in the SERP directory, rather than which keyword to stuff in.
| Common Belief | Reality | Rationale |
|---|---|---|
| H1 must contain the main keyword | Recommended, but not a hard ranking requirement | Google has repeatedly downplayed H1 weight |
| A page can have only one H1 | HTML5 allows multiple, but SEO best practice is one | Multiple H1s dilute the topic signal |
| More H2s means broader keyword coverage | More than 8 H2s dilutes instead | Topic focus drops |
| Headings determine rankings | Small direct impact, large indirect impact | Indirect effect through AI citations and CTR |
How headings and title divide the work: the title sits in the head and is written for the SERP, while the H1 sits in the body and is written for visitors and AI summaries. The two can be worded differently, but the core term should appear in both. See Keyword placement in title, H1, and H2.
2. One H1 only: one main title per page
The HTML5 spec allows multiple H1s per page (one per section), but SEO practice still favors templates that output only one H1 sitewide. The reason is that Google’s 2026 AI summary extractor treats the first H1 as the page topic, and multiple H1s make it hesitate.
| Checklist | Pass criteria | Common mistake |
|---|---|---|
| Number of H1s | Exactly 1 | Templates output 2 H1s |
| H1 position | At the very top of the body content | Logo or breadcrumb marked as H1 |
| H1 and title | Same core term, wording can differ | H1 has nothing to do with title |
| H1 length | 50 to 70 characters | Long sentences over 100 characters |
| CSS styling | Don’t use display:none | Hide the H1 with visual font size |
Case 1: WordPress themes often output the site title (logo text) as an H1, and the post title is also an H1. To check, count the H1 tags in the body with your browser’s developer tools, not the visual font size. If there are too many, change the logo to a div or p in the theme template.
Case 2: On a product listing page (category page), the H1 should be “Stainless Steel Flange,” not the category name “Products.” The latter is the template default, so all 200 category pages look identical and Google flags them as templated thin pages.
3. H2 and H3 hierarchy: no skipped levels, no stacking
H tag hierarchy works like folders: H1 is the root directory, H2 is a first-level subfolder, H3 is a second-level subfolder. Jumping from H1 to H3 is like creating a grandchild folder directly inside a folder — machines can’t read it.
| Tier | Suggested count for a 2,000-word article | Role | How it’s written |
|---|---|---|---|
| H1 | 1 | Main page title | Question or topic, 50 to 70 characters |
| H2 | 4 to 7 | Main sections | Start with a verb or phrase as a question |
| H3 | 0 to 3 under each H2 | Subsections | Specific steps or comparisons |
| H4 and below | Don’t use | — | Too many levels bring no benefit to B2B articles |
Counterintuitive follow-up: more H2s isn’t better. Cram 12 H2s into a 2,000-word article with only 100 words under each, and the topic gets chopped into fragments that Google struggles to identify as main sections. 4 to 7 H2s is the rule of thumb: fewer than 4 means the sections aren’t broken up, more than 7 means the paragraphs are too fragmented.
A common mistake is skipping levels: going straight from H1 to H3, or from H2 to H4. To check, run the page through a browser outline tool (such as HTML5 Outliner) and see whether the hierarchy tree is continuous.
4. How to place keywords in H tags
The principle for placing keywords in H tags is “main keyword in H1, secondary keywords in H2, long-tail in H3.” Don’t repeat the same word in every H.
| Location | What to put | Example |
|---|---|---|
| H1 | Main keyword once | Stainless Steel Flange PN16 Pressure Ratings and Specifications |
| H2 | Secondary keywords and question words | 1. How to Choose Between PN16 and PN10 Pressure Ratings |
| H3 | Long-tail variations | Differences Between 304 and 316L Materials at PN16 |
| What not to put | Synonym stacking | Flange / flange disc / flange plate / Flange |
Case 1: An article about first response to B2B inquiries has the H1 “Inquiry First-Response SLA and Conversion Rate,” with H2s “1. How to Put a 5-Minute SLA Into Practice,” “2. Three Things in Your First-Response Script,” and “3. Fallback When There’s No Reply in 30 Minutes.” Each H2 covers a secondary keyword and doesn’t repeat “inquiry first response” from the H1.
Case 2: For an article about content audits, don’t write all your H2s as “Content Audit Methods,” “Content Audit Steps,” and “Content Audit Tools” — that’s keyword stuffing. Use specific section names like “Three-Dimensional Evaluation Matrix,” “Quarterly Execution Rhythm,” and “Common Misjudgments” instead, and AI summaries will be more willing to cite them.
5. Common mistakes and how they affect AI summaries
These four H tag mistakes are the most common, and every one of them hurts your citation rate in AI Overviews.
| Mistake | Impact | Fix |
|---|---|---|
| Multiple H1s | Your topic signal gets diluted | Keep only the article title as an H1 in your template |
| Skipping levels (H1 straight to H3) | The hierarchy tree breaks | Add an H2 or promote the H3 to H2 |
| Using H tags for styling | Tagging something H2 just because the font is big | Control font size with a CSS class |
| H tags used as pure decoration | “Part one” or “next step” tells readers nothing | Every section H2 must contain a specific question or action |
Pages with a clear hierarchy are 2 to 3 times more likely to be cited in AI Overviews, which makes this the most underrated investment of the GEO era. To check where you stand, use the fixed question set in GEO experiment design and compare the citation rates in AI answers between pages with a clear hierarchy and pages with a messy one.
When you refresh old articles, H tag structure is a low-cost, high-return fix. Using the decision table in Refreshing Old Articles you can see that for pages still ranking in the top 20 for their main keyword but with low CTR, restructuring the H2s usually works faster than adding more body copy.
6. How to verify that your H tag changes actually earned AI citations
Don’t judge the results of an H tag restructure by gut feel. Build a fixed question set: pick 10 to 15 questions closely related to your business, search each one once a month on Google AI Overviews, ChatGPT and Perplexity, and note whether your site was cited and which section was cited. Run a baseline before the changes, then run it again three months after. If citations go from 1 page to 4, the hierarchy work paid off; if it’s still 1 page after three months, the problem is probably not your H tags but the content itself lacking original data.
One thing to watch when you log citations: AI usually cites the paragraph under the H2 heading, not the whole page. So an H2 like “How to choose between PN16 and PN10 pressure ratings” gets cited more often than “Pressure ratings overview” — the first is a question that can be answered directly, the second is just a topic. Rewriting H2s from topic-style to question-style is a concrete, zero-cost way to raise your AI citation rate, and you don’t have to touch a single word of the body text.
One more detail: numbers and years in H2s also get pulled out by AI. “2026 SEO checklist for DTC sites” gets cited by Perplexity more often than “SEO checklist for DTC sites,” because when the model answers questions like “what should I do this year,” it favors the version with a year. Evergreen methodology articles don’t need a year, but adding one to the H2s of tool and standards articles is a small move with a solid payoff — change one heading and you get one more shot at a citation.
FAQ
How many H1 tags can a page have?
How many H2 tags should you use?
Does stuffing keywords into H tags improve rankings?
Does the H1 have to match the title exactly?
Can H tags skip levels, like going straight from H1 to H3?
Can I just use H2 for every visual subheading?
Final thoughts
Optimizing H tags costs very little: you change the heading hierarchy without touching the body content. But it’s the most underrated investment of the GEO era: AI summary extractors split sections by H tags, and pages with a clear hierarchy are 2 to 3 times more likely to be cited.
Three steps to get it done: this week, run the HTML5 Outliner across your top 50 highest-traffic pages and list the ones with multiple H1s, skipped levels, or too many H2s; next week, restructure them to the standard of “1 H1 + 4 to 7 H2s + 0 to 3 H3s under each H2”; in week three, compare citation rates before and after on AI Overviews to verify the Monthly monitoring of a fixed question set data.
One line worth remembering: heading tags aren’t for stuffing keywords. They’re the table of contents you hand to Google and AI. Keep the hierarchy clear, and summaries will quote you.
