The Outline Your Page Already Has
Strip everything off a web page except its headings and you're left with a table of contents nobody wrote deliberately. For a well-structured page, that outline reads like a summary. For most pages, it reads like a list of unrelated fragments.
That gap is diagnostic. A heading structure that doesn't make sense as a standalone outline usually indicates content that doesn't make sense as a standalone argument — because headings are where a writer commits to what each section is about.
Extracting the outline is the fastest way to see it.
Reading Your Own Page as an Outline
The Heading Extractor pulls every heading from a URL with its level:
- Enter the page URL.
- Read the extracted hierarchy.
- Ignore the page design entirely and read the outline as prose.
Four questions to ask of the result:
Does it tell a coherent story on its own? Read the headings in order. If someone who hadn't seen the page could follow the argument, the structure is doing its job.
Are the levels logically nested? An H3 should be a subdivision of the H2 above it. If it's a peer, it should be an H2.
Is anything at the wrong level? Section headings marked as H4 because the designer wanted smaller text is one of the most common structural faults.
Are there gaps in the argument? Sections that assume something not covered by any preceding heading.
The Skimming Pattern
Most people don't read a page top to bottom. Eye-tracking research on web reading has consistently found scanning behaviour — readers move down the page picking up headings, subheadings, and the openings of paragraphs, then commit to reading a section only if it looks relevant.
That means your headings are the first draft of your page that anyone reads. For a substantial proportion of visitors, they're the only draft.
Three practical consequences:
Headings must be self-contained. A reader who lands on your page and jumps to the fourth H2 has no context from what came before. "How It Works" tells them nothing. "How Rate Limiting Works in Nginx" tells them everything.
Vague headings cost you readers. "Getting Started", "The Details", "Some Considerations" — these are placeholders that survived into publication. They give a skimming reader no reason to stop.
The first two or three headings do disproportionate work. They set expectations about what kind of document this is and whether it's worth the time.
Descriptive Beats Clever
The temptation with headings is to make them interesting. Usually the more useful choice is to make them informative.
| Weak | Better |
|---|---|
| Getting Started | Installing the CLI on macOS and Linux |
| The Problem | Why Batch Imports Time Out Above 50,000 Rows |
| Considerations | Cost, Latency and Vendor Lock-In Compared |
| Next Steps | Migrating Existing Data to the New Schema |
| Conclusion | When to Choose Each Approach |
The right-hand versions are longer, which feels like a downside and isn't. A heading is not competing for space with body text — it's competing for attention with the reader's decision to leave.
Front-load the distinguishing word. In a list of headings, the reader's eye catches the first two or three words. "How to Configure Redis Persistence" is scannable; "A Guide to Understanding How You Might Configure Redis Persistence" buries the payload.
Structure That Follows the Argument
The most common structural failure isn't a missing H1 or a skipped level. It's headings that don't correspond to the actual shape of the content.
Symptoms:
Sections of wildly uneven length. One H2 covering 800 words and the next covering 60. Usually the short one isn't a section — it's a paragraph that got promoted.
Headings added after writing. Content written as continuous prose and then chopped into sections by inserting headings at plausible-looking points. The seams show, because the paragraphs weren't written to belong to those sections.
Parallel content at non-parallel levels. Three comparable options where two are H2s and one is an H3, because of how the draft evolved.
Headings that don't match their content. The section under "Performance Considerations" actually discusses pricing. This happens when content is edited but headings aren't.
The fix in every case is to outline before writing, or to re-outline after. Writing the headings first forces you to decide what the sections are — and it exposes gaps and duplication while they're still cheap to fix.
Length and Density
There's no rule about how many headings a page should have, but there are useful heuristics.
A section running past 400–500 words without a subheading is probably doing more than one thing. Either split it or check whether the extra material belongs elsewhere.
A page with an H2 every 100 words is fragmented. Constant headings interrupt reading rhythm and usually indicate that related material has been artificially separated.
Three or four levels deep is almost always too much. H4 and beyond suggests the page is trying to be a book. Split it into multiple pages, each with a cleaner two-level structure.
Depth should reflect genuine subordination. An H3 exists because it's part of the H2 above it. If it stands alone conceptually, it's an H2.
Using Extraction for Content Planning
Beyond auditing your own pages, heading extraction is a fast research method.
Compare outlines across several pages on the same topic. Extract the headings from five pages covering something you're about to write about. Common sections across all five are almost certainly things readers expect — omitting them looks like a gap. Sections that appear on only one are either differentiators or padding.
Find the questions being answered. Pages with question-formatted headings are telling you what queries their authors expected. That's a useful signal about what people actually want to know.
Spot the missing angle. If five competing pages all have the same six sections and none covers something you know matters, that's the reason for your page to exist.
Check depth of treatment. A page whose entire structure is six H2s with no subheadings is treating the topic broadly. One with nested H3s under each H2 is going deeper. Both are valid; knowing which you're up against shapes what you write.
Practical Tips
Write the headings before the body. It's an outline, and outlining is cheaper than rewriting.
Read your headings aloud in sequence. Awkwardness in that reading is a reliable signal of structural problems.
Make each heading work out of context. Assume the reader hasn't read anything above it.
Use levels for hierarchy, never for typography. If a heading needs to be smaller, that's a CSS decision, not a markup one.
Re-extract after editing. Content changes; headings often don't follow. A quick extraction catches the drift.
Avoid heading text that duplicates the H1. Redundant headings waste the reader's attention and make the outline harder to scan.
Watch out for headings that are really labels. "Note", "Important", "Example" describe a formatting treatment, not a section of the argument. They usually belong as styled callouts rather than headings.
FAQ
How long should a heading be? Long enough to be specific, short enough to scan — typically between four and ten words. Specificity matters more than brevity.
Should headings be questions? They work well where the content genuinely answers a question a reader would ask. Used for every section they become repetitive and start to feel formulaic.
How many H2s should a page have? Whatever the content requires. As a rough sanity check, a substantial article usually has somewhere between four and ten, with each covering a distinct part of the argument.
Does heading order matter? Yes, and it's easy to overlook. The outline should progress logically — background before detail, problem before solution, and comparable items grouped together.
Should I use sentence case or title case in headings? Either, consistently. Sentence case is easier to read for longer headings; title case looks more formal. Mixing the two on one page looks careless.
Is it a problem to have two sections with similar headings? Usually a sign that they should be merged or that one needs a more specific title. Near-duplicate headings make the outline harder to navigate.
The Takeaway
A heading structure is the argument of a page made visible. Extracting it strips away everything that might be disguising a weak structure — the design, the images, the prose rhythm — and leaves you with the outline your readers are actually skimming. If that outline doesn't stand on its own, the page probably doesn't either.
Extract the complete heading structure of any page free with the Heading Extractor at sadiqbd.com — no sign-up, instant results.
