Answer engine optimization.

A narrower question than "is this good content"

Answer engine optimization gets talked about like a content-quality exercise, and that's the wrong frame. AEO asks something much narrower and far more mechanical: if an engine wanted to quote two sentences answering a buyer's question, do two such sentences actually exist on your page, attached to a heading genuinely phrased as a question, in a format the engine can parse?

A site can be well-written and rank well and still fail this completely. The information is often all there — it's just scattered across paragraphs under headings like "Our Approach" or "What We Do," which read as confident marketing copy but answer nothing an engine can lift. There's no question mark, no self-contained sentence, nothing to quote. The content exists; it just isn't shaped to be pulled out and handed to someone asking a question.

What AuditSpark found across 40 real sites

AuditSpark's own testing measured this directly across a sample of 40 real small-business homepages — not a synthetic benchmark, actual sites doing actual business. Treat the figures below as specific to that segment and accurate to roughly ±7 percentage points, not a universal claim about every website.

67.5% of the sites — 27 of 40 — had zero question-shaped headings anywhere on the homepage. Not weak ones, none. And 92.5% had no answer schema at all: no FAQPage, no QAPage, no HowTo, no Speakable markup of any kind.

That second number is easy to misread as "these sites have no structured data," which isn't quite right and matters for how you check your own. 65% of the sample — 26 of 40 — did carry some JSON-LD, but none of it was answer-oriented: Organization, WebSite, BreadcrumbList, LocalBusiness — the defaults most website builders emit automatically, whether anyone asked for them or not. A shallow check that just asks "does this site have structured data?" would pass all 26 of those sites. That pass would be technically accurate and still wrong for the thing that actually matters, because none of that markup tells an engine "this block of text is an answer to a question."

The trap in measuring "questions"

There's a specific mistake worth naming because it's easy to make and it inflates scores exactly where sites are weakest. An early pass at this measurement counted any heading that started with an interrogative word — "What," "How," "Who" — as a question. On real small-business copy, 44% of those matches turned out to be false positives. "Who We Are," "What We Do," "How We Help" are declarative marketing headers that happen to open with a question word. They are not questions. A real question needs a question mark, and it needs to actually be answered by what follows, not just gestured at.

Two related gaps are worth checking for on your own site. In a strict reading, only 5% of the sample had genuine FAQ-style content that simply wasn't marked up (12.5% under a looser reading) — real question-and-answer content sitting there unmarked. And 15% — 6 of 40 — answered questions only on a separate, dedicated FAQ page rather than on the page actually being read. If your homepage is the page a buyer lands on and an engine scans, FAQ content living three clicks away on its own page may as well not exist for that page's purposes.

What a liftable answer actually requires

Four properties have to be true at once. Missing any one of them means the content isn't liftable, even if the other three are perfect.

#	Property	What it means
1	Real question heading	States the actual question and ends in a question mark. "What's the difference between X and Y?" qualifies. "Our Approach" doesn't, no matter how question-shaped it looks.
2	Self-contained answer	Survives being lifted out of context — roughly 15–80 words, the direct answer in the first sentence, no pronouns that only make sense with the surrounding paragraph still attached.
3	Parseable format	Plain paragraphs, lists, and tables extract cleanly. Content that exists only as a styled visual — an infographic, a div-based layout with no real markup — doesn't.
4	Identifying markup	FAQPage, QAPage, HowTo, or Speakable schema, populated with real content that matches the visible text exactly — not an empty wrapper.

The first two properties are about the words themselves; the second two are about whether a machine can find and trust them. A page can nail the writing and still fail on format or markup, and vice versa — which is why all four have to hold together, not just the ones that are easiest to fix.

Why this is graded, not scored

You might expect this to come back as a number out of 100, the way a lot of readiness checks work. It doesn't, and the reason is the same distinction we cover in the visibility ecosystem piece — some things are better classified than scored. With 67.5% of sites at zero question headings and 92.5% lacking any answer schema, a weighted numeric score would read near-zero for roughly two-thirds of any sample. That fails the one job a score has: discriminating between sites, and showing whether a site is actually improving over time.

Three honest categories do the job better. Pass — question headings, self-contained answers, and identifying markup are all present. Partial — real question-shaped content exists, but it's unmarked, or too long and diffuse to extract cleanly. Gap — no answer-shaped content found at all. It's a coarser read than a score, but it's an honest one, and it moves in the direction you'd expect as you fix things.

What this looks like in practice

The good news buried in those numbers: this is almost entirely a restructuring exercise, not a content-production project. You're editing headings and tightening paragraphs on pages that already exist, not writing anything new. It's also one of the lowest-cost, most under-claimed levers available — and one that fails completely silently, since nothing in ordinary analytics shows a competitor getting quoted in your place. The upside is that unlike LLMO measurement, which takes real waiting to see engine behavior shift, AEO gives fast, deterministic feedback: you can check whether the markup validates and the answer actually renders in your served HTML today.

• Rewrite your highest-value headings as real questions. Take your five or ten most commercially important sections and turn generic headers into the actual question a buyer would ask — question mark included — followed immediately by a self-contained answer.
• Bring FAQ content home to the page it belongs on. If real Q&A content only lives on a separate FAQ page, an engine reading your service or product page never sees it. Answer the question where the buyer is actually looking.
• Check what your accordion actually serves. A collapsed panel is fine — the problem is only when the answer text is lazy-loaded by JavaScript on click, because a crawler that never clicks never sees it.
• Add matching markup last, not first. Once the visible answer exists and reads cleanly, add FAQPage or the matching schema type — populated with the same text that's actually on the page, not an empty wrapper standing in for it.

Frequently asked

Do collapsed accordion panels hurt AEO?

Not by themselves. What matters is whether the answer text exists in the HTML your server actually sends. If it's already in the served markup and JavaScript just handles the visual collapse, an engine can still read it. The problem is only when the answer is lazy-loaded on click — then a crawler that doesn't execute that click never sees the content at all.

We already have structured data. Isn't that enough?

Not necessarily. AuditSpark's testing found 65% of sites carried some JSON-LD, but almost none of it was answer-oriented — Organization, WebSite, and LocalBusiness are the defaults most site builders add automatically, and they tell an engine nothing about which text answers which question. Check specifically for FAQPage, QAPage, HowTo, or Speakable schema that matches your visible content.

Does fixing this mean writing a lot of new content?

Usually not. Most of the value here comes from restructuring what's already on the page — turning a generic heading into a real question, tightening a paragraph into a self-contained answer — rather than producing new material. That's also why it's a fast, low-cost place to start.