Answer-engine readiness
for Shopify product pages.

“Answer engine optimisation” has no single canonical origin — industry accounts of where the term came from disagree — so Rank Sniper states its own definition of AEO and scores against it:

Whether the catalogue can be lifted as a direct answer to a shopper's question.Rank Sniper pillar definition — AEO

The definition is deliberately narrow. It is about the page, not about any engine: whether the catalogue contains passages that answer a shopper’s question directly, specifically and in a shape that can be quoted. Rank Sniper does not claim to know how any engine selects the passages it quotes or the sources it cites, and nothing here is a ranking recipe.

Google’s own position points the same way. On its AI features page:

“There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”Google Search Central — AI features and your website

So readiness for an answer engine is not a special format. It is a set of ordinary page qualities — the opening sentence, the numbers, the lists, the headings — that make a product page easy to quote accurately, whether by an engine, a search snippet or a person skimming on a phone. On Shopify almost all of them live in one field: the product description.

Most product descriptions open with mood. Mood sells, but it answers nothing. Compare:

Inspired by slow mornings on the coast, our favourite throw brings a sense of calm to any room.

The Harbour throw is a 130 × 170 cm blanket woven from 100% merino wool, weighs 1.2 kg and is
machine-washable at 30 °C. Inspired by slow mornings on the coast, it brings a sense of calm to any room.

The second version keeps the voice and puts four facts ahead of it: what the product is, its size, its material and its care. Each is a claim a machine can repeat without paraphrasing and a shopper can check.

How the scan reads the opening

The answer-first structure check takes the description’s first sentence and asks two things: is it between 40 and 300 characters, and does it make a statement — is, are, has, have, features, includes, measures, weighs or comes? It is a blunt heuristic, not a judge of quality. It cannot tell a good sentence from a bad one; it can tell a statement from a slogan.

Numbers with units

A number with a unit is the most quotable thing on a product page. The GEO pillar’s quotable claim density check (16 of 82) looks for exactly that — gsm, oz, cm, mm, kg, ml, W, mAh, per cent and similar — anywhere in the description. Rank Sniper will not invent the figure: if the width is not in your copy, only you can add it.

Enough text to quote

The description depth check (20 of 86) measures the description with its HTML stripped. The rubric gives its reason in one line: Under ~180 characters there is nothing for an agent to extract facts from.

A specification buried in a paragraph can only be quoted as the paragraph. The same facts as a list can be quoted one at a time. A list made in the product description editor is stored in the description’s HTML as ul and li elements — structure a machine can see, not just a visual indent.

Headings do the same for questions. A description that asks and answers what shoppers actually ask — size, care, compatibility, what is in the box — gives an engine a question to match and the answer directly beneath it:

<p>The Harbour throw is a 130 × 170 cm blanket woven from 100% merino wool.</p>

<h3>What size is the Harbour throw?</h3>
<p>130 × 170 cm, sized for a two-seat sofa or the foot of a queen bed.</p>

<h3>How do I wash it?</h3>
<p>Machine-wash at 30 °C on a wool cycle, then dry flat.</p>

<ul>
  <li>Material: 100% merino wool, 400 gsm</li>
  <li>Weight: 1.2 kg</li>
  <li>Made in Portugal</li>
</ul>

What the scan counts

Two checks read this structure. List and table density (16 of 86) reads the description and passes it when it contains a list, a measurement with a unit, or the words dimension, material or composition. Despite its name, it does not look for a table element — a table passes when its cells carry units. Question-shaped headings (14 of 86) reads every heading on the sampled product pages, theme and description alike, and counts one that ends in a question mark or opens with how, what, why, when, where, which, can, does, do, is or are.

Keep the heading levels honest

The SEO pillar’s H1 structure check (10 of 100) expects exactly one h1 per product page. If your theme renders the product title as the page’s h1, keep description headings at h2 or h3, so the questions add structure without competing with the title.

The dated record

• 2023-09-14. Google’s documentation-updates log records FAQ rich results being limited to “well-known, authoritative government and health websites”, and the removal of the How-to documentation because that rich result “is no longer shown in search results”.
• 2026-05-08. The log records that the FAQ feature “will no longer appear in Google Search starting May 7, 2026”.
• 2026-06-15. The FAQPage documentation is removed because “The FAQ rich result feature is no longer shown in Google Search results.”

Any offer to win FAQ rich results is now selling something that no longer exists, for every site. Our journal lists the other retired types.

What the markup is still for

FAQPage remains part of the schema.org vocabulary, and a page with real questions and answers can still state them as structured data. What it buys now is structure — a machine-readable pairing of each question with its answer — not a Search feature. Rank Sniper’s rule for FAQ markup is the one it applies to its own pages: mark up only questions and answers that are visible on the page, word for word, in the merchant’s own voice. The question below matches the heading in the description above exactly.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What size is the Harbour throw?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "130 × 170 cm, sized for a two-seat sofa or the foot of a queen bed."
      }
    }
  ]
}
</script>

“AI is built into Search and integral to how Search functions, which is why robots.txt directives for Googlebot is the control for site owners to manage access to how their sites are crawled for Search.”Google Search Central — AI features and your website

“To limit the information shown from your pages in Search, use nosnippet, data-nosnippet, max-snippet, or noindex controls.”Google Search Central — AI features and your website

Two consequences for a store. First, the snippet controls are page-level and element-level tools that apply to Search as a whole, not a separate switch for AI answers: data-nosnippet on an element keeps that element out of snippets, max-snippet caps snippet length for the page, nosnippet removes snippets for the page, and noindex removes the page. Second, Google-Extended is not one of them. It is a control token for whether content may be used to train future Gemini models and for grounding in some of Google’s other systems, and Google says “Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search.”

<!-- Keep a time-limited promotion out of snippets -->
<div data-nosnippet>Free shipping this weekend only.</div>

<!-- Cap the snippet length for the whole page -->
<meta name="robots" content="max-snippet:160">

<!-- Bing, per Microsoft's 2023 post -->
<meta name="bingbot" content="nocache">

On Shopify the meta tag belongs wherever your theme renders the page’s head, conditioned on the product template if it should not apply site-wide; data-nosnippet goes on the element itself, in the section that renders it. Use both sparingly: every character withheld from a snippet is a character that cannot be quoted.

Bing

Microsoft documented its controls on 2023-09-22 for what it then called Bing Chat. With NOCACHE, content may be included in answers but only URLs, titles and snippets are shown, and only those may be used in training Microsoft’s generative AI foundation models. With NOARCHIVE, content is not included in Bing Chat answers and is not used for that training. The post predates the product’s later name and we have not read a newer official restatement, so treat it as the last documented position rather than a current guarantee.

What the scan reads here

Rank Sniper reads your robots.txt rules for Googlebot and Bingbot and counts a full block against the SEO pillar only — 6 points per blocked search crawler, capped at 30, never against the composite. It also reads Google-Extended as one of the 10 AI agents the scan checks by name: a disallow there counts toward the capped crawler penalty like any other agent on the list, because the score treats it as an AI-use policy — while Google says it has no effect on Search. The crawler index lists every agent.

Two rules decide how the numbers behave. A check whose input could not be read — no product page fetched, no products returned — is excluded from the pillar rather than scored zero, so the pillar is computed over the checks that actually ran. And a catalogue check is a share: if half your products open with a statement, that check scores half.