If you've been optimizing your content for traditional search and wondering why AI answer engines still skip your pages, the culprit is often sitting right at the top of your HTML—your page title and H1. Before a large language model reads a single sentence of your body copy, it's already building a mental map of what your page is about, and that map is drawn almost entirely from your headings.
Why LLMs Treat Headings as Semantic Scaffolding
Think of an LLM processing your page the way a researcher skims a research paper. They read the title, glance at the section headers, and form a hypothesis about whether this document answers the question they're investigating. Only then do they invest time in the full text.
This isn't just a rough analogy—it reflects how transformer-based models actually process structured documents. Heading tags (H1, H2, H3) function as explicit semantic signals. They tell the model:
- What the page's primary claim is (H1)
- How that claim is broken into sub-topics (H2s)
- The granular details and supporting points (H3s and below)
When AI answer engines like Perplexity, ChatGPT's browsing mode, or Claude retrieve and cite web pages, they're doing a rapid structural parse before committing to deeper extraction. If your headings don't map cleanly to the query being asked, your page gets deprioritized—no matter how strong your body copy is.
The practical implication: your H1 is a promise, and your H2s are the chapters that keep it. If any of those break the chain of relevance, the model moves on.
The Difference Between a Title Tag, an H1, and Why Both Matter Separately
These two elements get lumped together constantly, and that's a mistake in the AEO world.
Your Page Title (Title Tag)
The title tag lives in the <head> of your HTML and is primarily what search engines show in SERPs—and what LLMs see when they retrieve a URL from their index or a live web fetch. It functions as the page's external label. A well-written title tag answers the question: "If someone were searching for this page, what would they type?"
For AEO purposes, your title tag needs to:
- Contain the specific phrase or question your target query uses
- Be under 60 characters so it isn't truncated in retrieval contexts
- Avoid vague brand-first framing ("Acme Co | Services") in favor of intent-first framing ("Accounting Software for Freelancers | Acme Co")
Your H1
The H1 is the page's internal promise—the first thing a reader sees and the anchor that all subsequent headings hang from. Critically, it doesn't have to be identical to the title tag, and in many cases it shouldn't be.
Your title tag earns the click (or the AI retrieval). Your H1 earns the confidence that the page will actually answer the question. Think of the title tag as the label on the juice bottle and the H1 as the first line of the ingredient list—they're related but doing different jobs.
For AEO, your H1 should:
- Restate the core topic with natural, conversational phrasing that mirrors how people actually ask questions to AI assistants
- Be specific enough that a model can extract it as a standalone answer or context fragment
- Avoid pun-based or clever-but-vague phrasing (save that for your CTAs)
How Small Phrasing Changes Shift Which Queries You Get Cited For
This is where things get genuinely interesting—and where most teams leave visibility on the table.
Consider two versions of an H1 for a page about project management software:
- Version A: "Streamline Your Workflow Today"
- Version B: "Project Management Software for Small Teams Under 10 People"
Version A sounds polished. Version B sounds boring at a marketing pitch but performs dramatically better in AI retrieval contexts because it contains the semantic components of actual queries: a category ("project management software"), a qualifier ("small teams"), and a scope ("under 10 people").
When someone asks Perplexity "what's the best project management software for a small team," Version B's H1 matches three of the five semantic tokens in that query. Version A matches zero. The model isn't being picky—it's being literal.
The "Query Token Overlap" Mental Model
You don't need to understand transformer tokenization to use this concept practically. Just ask yourself: if someone dictated their question to an AI assistant, which words from that question appear verbatim or near-verbatim in my H1?
Aim for at least three meaningful word overlaps (not stop words like "the," "for," "a"). If your heading can't hit that threshold, rewrite it.
H2s Determine Query Breadth
Your H2s expand the set of queries you can be cited for. Each H2 is essentially a second H1 for a subtopic—and AI answer engines often pull individual sections rather than whole pages.
If your page is about email marketing tools, and your H2s are:
- "Easy to Use"
- "Great Features"
- "Pricing That Works"
...you're invisible for any specific query. But if they read:
- "Email Marketing Tools With Visual Drag-and-Drop Editors"
- "Automation Features: Sequences, Triggers, and Behavior Tracking"
- "Email Marketing Pricing: Free Tiers vs. Paid Plans Compared"
...now each H2 independently catches queries. You've turned one page into six or seven retrievable answer fragments.
Common Heading Mistakes That Kill AI Visibility
1. Starting the H1 with Your Brand Name
"Acme Co: Digital Marketing Solutions" is nearly useless for AEO. The brand name adds no semantic value for a user who doesn't know you yet—and AI retrieval is almost always triggered by people who don't know you yet. Lead with the topic, trail with the brand if necessary.
2. Using the Same Text for the Title Tag and H1
Repetition isn't inherently bad, but it wastes a slot. Your title tag can be optimized for retrieval (keyword-dense, concise), while your H1 can be optimized for confidence (natural, specific, slightly longer). Using the same text for both means you're doing neither job as well as you could.
3. Clever Headlines That Obscure the Topic
Marketing-speak H1s like "The Future Is Now" or "Where Ambition Meets Results" tell an LLM absolutely nothing about your page. They'll be ignored in favor of a competitor whose H1 reads "Business Loans for Startups: Rates, Requirements, and Applications."
4. Skipping Heading Hierarchy
If your page jumps from an H1 to H3s without H2s, or uses H2s as styling choices rather than structural ones, the semantic scaffolding collapses. LLMs infer topic relationships from heading nesting. A broken hierarchy means broken context inference.
5. Generic H2s That Could Belong to Any Page
"Overview," "Benefits," "How It Works," "Contact Us"—these aren't wrong, but they're semantic dead weight for AEO. Every heading is an opportunity to capture a specific query. Don't waste them on labels.
A Quick Audit Process You Can Run Today
You don't need a sophisticated tool to start improving your heading structure for AI visibility. Here's a five-step process you can run on any page in under 15 minutes:
Extract all headings. Paste your page into a free heading extractor tool or check the source HTML. List every H1, H2, and H3 in order.
Read the headings alone. Ignore the body copy. Do the headings alone tell a complete, coherent story? Could someone understand your page's argument from headings only?
Map headings to queries. For each H2, write down the query a real user might type to an AI assistant that this section could answer. If you can't write one, the H2 needs to be rewritten.
Check query token overlap. For your most important target queries, count how many meaningful words appear in your H1. Aim for three or more.
Verify hierarchy. Confirm you have exactly one H1, that all major sections use H2, and that H3s are used only for sub-points within an H2 section.
This isn't the full picture—body copy, structured data, internal linking, and external citations all matter too—but fixing heading structure is the highest-leverage first pass you can make.
How AEO Juice Approaches This for You
Manually auditing every page on your site for heading structure, query alignment, and semantic hierarchy is genuinely tedious. It's also the kind of work that falls off calendars the moment anything shinier demands attention.
That's why AEO Juice builds heading analysis directly into our automated visibility workflow. Our free 26-check AEO report flags heading issues as part of a broader diagnosis—including whether your H1s are query-aligned, whether your title tags are optimized for retrieval, and whether your heading hierarchy is structurally sound for LLM parsing.
For teams on our Pro and Prime tiers, we go further: our AI-generated fix recommendations surface specific rewrite suggestions for underperforming headings, matched against the queries your competitors are already getting cited for. The weekly LLM-visibility tracking shows you whether heading changes are actually moving the needle in AI answer surfaces—not just traditional SERPs.
If you've been doing solid content work but still feel invisible when AI assistants field questions in your category, start with the free report. It takes about two minutes to generate and almost always surfaces at least two or three heading issues that are straightforwardly fixable.
FAQ: Page Titles, H1s, and AI Answer Engines
Does the H1 tag affect whether AI answer engines cite my page?
Yes, significantly. AI answer engines use heading tags as primary semantic signals to understand what a page covers and whether it's relevant to a specific query. A vague or keyword-poor H1 reduces the likelihood that your page gets retrieved and cited, even if your body copy is excellent.
Should my title tag and H1 say exactly the same thing?
Not necessarily. Your title tag is optimized for retrieval (concise, keyword-forward), while your H1 is optimized for on-page comprehension and confidence. They should reinforce each other, but each can do its job more effectively when they're distinct but aligned.
How many H2s should a page have for good AEO?
There's no magic number, but each H2 should correspond to a distinct subtopic that could independently answer a user query. Most well-structured pages have four to eight H2s. Fewer than three suggests the page is covering its topic too broadly; more than ten often signals that the content needs to be split into multiple pages.
Can I improve AI visibility just by fixing headings, without rewriting body copy?
Heading fixes alone can meaningfully improve AI visibility, especially if your current headings are vague or poorly structured. That said, headings and body copy work together—headings set the context, body copy provides the supporting evidence. The biggest gains come from aligning both.
What's the fastest way to check whether my headings are hurting my AI visibility?
Run the free AEO Juice 26-check report. It checks heading structure as part of a full AEO audit and takes about two minutes to generate. That's genuinely the fastest starting point.
Heading structure isn't glamorous. It doesn't have the novelty of AI-generated content or the drama of a core algorithm update. But it's the foundation that every other optimization sits on—and in the age of AI answer engines, it's the first thing the machines read. Get it right, and everything else works harder.