The AEO checklist: website structure, H1s, and FAQs AI can read

The practical shift is that you're writing for extraction. An AI reading your page doesn't scroll the way a person might, so it looks for a heading that matches the question, a paragraph directly under it that answers the question on its own, and machine-readable signals that confirm what the page is about. If your strongest answer is buried in paragraph four or split across three sections, it tends to get passed over in favor of a page that put the answer up top. We dig into the full playbook on our approach to AEO, but the checklist here covers the structural foundation everything else sits on.

VIDEO TRAINING

Get the Growth Playbook.

Learn to plan, budget, and accelerate growth with our exclusive video series. You’ll discover:

• The 5 phases of profitable growth
• 12 core assets all high-growth companies have
• Difference between mediocre marketing and meteoric campaigns

How should you structure a website so AI can read it?

Structure your site so every page owns one clear question and every URL tells the AI what it's about before the page even loads. AI systems map your site through internal links, URL paths, and topic clusters, so a flat pile of unrelated pages reads as noise, while a tight cluster of related pages on one subject signals that you actually know the territory. We go deeper on this in our answers about website structure and navigation for B2B sites.

Run this checklist on your site architecture:

• Give each page a single primary job. A page that tries to answer "what does a website redesign cost," "how long does it take," and "which CMS is best" all at once gives an AI nothing clean to extract. We've found one focused question per page gets cited far more reliably than a sprawling pillar that touches ten topics shallowly.
• Write URLs that describe the content. A slug like /blog/hubspot-website-redesign-cost tells an AI the topic instantly, whereas an opaque slug like /blog/post-4821 forces the system to guess at what the page covers. Keep paths short, lowercase, and keyword-bearing, and avoid changing them once they've earned citations.
• Build topic clusters with internal links. Group related pages and link them to each other so the crawler sees a coherent body of work on one subject. A cost article that links to a process article that links to a CMS comparison signals depth on website strategy, which is how AI assesses topical authority.
• Keep the answer above the fold in the markup. Put the direct answer near the top of the page's HTML rather than gating it behind an accordion or a JavaScript module that loads on click. If a reader has to interact to reveal the text, assume some crawlers never see it.
• Make every page reachable in a few clicks. Orphan pages with no internal links pointing to them rarely get crawled well. A clean navigation and a logical link structure get more of your library indexed and understood.

What makes an H1 and heading hierarchy that AI understands?

A heading that AI understands is phrased as the actual question a person would ask, with a single H1 per page and H2s that each open a self-contained answer. AI systems lean heavily on heading structure to figure out what a page covers and which chunk answers which query. When your H2 matches the phrasing of a real prompt, you've handed the AI a labeled box with the answer already inside.

Work through your headings against this list:

• Use exactly one H1, and make it the page's core question or claim. The H1 is the strongest on-page signal of what the page is about. Phrase it the way someone would ask it out loud, like "How much does a HubSpot website redesign cost," rather than a vague label like "Pricing."
• Phrase H2s as questions people actually type. A heading like "What's included in a website redesign" mirrors how people query an assistant, so it tends to surface in answers far more often than a terse label like "Inclusions" that no one would actually type. We write H2s by imagining the follow-up questions a prospect asks us on a call, then using those almost verbatim.
• Keep the hierarchy clean and sequential. H1, then H2s, then H3s nested under them, with no skipped levels and no headings used purely for visual size. A logical nesting tells the AI how subtopics relate to the main topic.
• Put a standalone answer in the first sentence under each heading. The paragraph directly below an H2 should answer that H2 completely on its own, without needing the paragraph before it for context. That's the snippet an AI lifts, so it has to make sense in isolation.
• Front-load the keyword so the substance lands in the first line. When an AI scans the opening sentence of a section to judge relevance, it picks up a direct answer right away, so a warm-up sentence ahead of it only pushes the useful content lower and weakens the signal. Lead with the point you want quoted.

How should you format FAQs so AI can quote them?

Format FAQs as a literal question in a heading followed immediately by a complete, two-to-four-sentence answer that stands on its own. This is the single most citation-friendly pattern we use, because the structure mirrors exactly how an AI wants to consume a Q&A: clear question, clean answer, no surrounding clutter required to understand it.

Here's what we check on FAQ sections:

• Match the question to real query language. Write the question the way a person phrases it to an assistant, including the natural specifics. A heading like "How long does a HubSpot website build take" is the actual prompt someone types, so it lines up with real queries in a way that a generic label like "Project timelines" never will.
• Answer in the first sentence, then support it. Lead with the direct answer, then add one or two sentences of context or a concrete number. An AI often pulls only the first sentence, so it has to carry the answer by itself.
• Keep each answer self-contained. Don't write "as mentioned above" or "like the previous example," because the AI may extract that answer with zero awareness of what came before. Every answer should read correctly as a standalone block.
• Use real numbers and named specifics. A concrete answer like "a design blueprint runs $6K to $12K over four weeks" gives an AI something quotable, while a vague line like "pricing varies" leaves it nothing to lift. Specificity is how an AI distinguishes a practitioner answer from filler, and it's a strong trust signal.
• Render FAQ text in the HTML, not behind a click. Accordions are fine for humans as long as the answer text lives in the page source rather than loading only when expanded. Confirm the answer is in the rendered markup so crawlers and AI fetchers can read it.

The table below maps the most common formatting choices to how an AI handles them.

What structured data supports an AEO strategy?

The structured data that supports an AEO strategy is JSON-LD schema that labels your content in a language AI and search engines read natively, with FAQPage, Article, and HowTo as the core types. Schema doesn't change what a human sees, but it removes ambiguity for the machine. It confirms that a block of text is a question-and-answer pair, that a page is an article with a known author and date, or that a section is a step-by-step process.

This is where clean content and the support behind an AEO strategy come together, since structured data is the layer that turns readable copy into confirmed, parseable facts. We handle this on HubSpot builds through schema markup for AEO implemented as JSON-LD in the page template, so it stays consistent across the whole site instead of being hand-coded page by page.

Run these schema checks before you call a page done:

• Add FAQPage schema to any Q&A section. Wrap each question and its answer in FAQPage markup so the AI gets an explicit, structured confirmation of the pair rather than inferring it from layout.
• Add Article schema with author and dates. Include author, datePublished, and dateModified so AI systems can assess authorship and recency, both of which feed the trust signals that influence whether you get cited.
• Add HowTo schema to process content. When a page walks through ordered steps, HowTo markup labels each step so an AI can reconstruct the sequence accurately instead of guessing the order.
• Validate before you ship. Run every page through Google's Rich Results Test and Schema.org validator to catch a missing required field, since broken schema can be quietly ignored. We treat a clean validation pass as part of the definition of done.
• Keep schema in sync with the visible content. The structured data must describe what's actually on the page. If the schema claims an FAQ answer the page doesn't show, you risk the markup being disregarded, so update both together whenever content changes.

Schema markup recommendations

For an AEO checklist or guide page like this one, we recommend implementing:

• FAQPage schema for the question-based sections, treating each H2 question and its opening answer as a structured Q&A pair so AI systems can extract them directly.
• Article schema with author, datePublished, and dateModified fields to establish authorship and signal recency, which support citation likelihood in 2025 and 2026.
• HowTo schema if you present the four checklist areas as an ordered implementation sequence, with structure, headings, FAQs, and schema as distinct steps.
• BreadcrumbList schema to reinforce where the page sits inside your topic cluster, helping AI understand its relationship to the rest of your library.