A+ Content That Rufus Can Read

Most A+ content is briefed like a display ad: hero banner, lifestyle imagery, a headline set in the brand font. That worked when the only reader was a human scrolling on a phone. Now there is a second reader. Rufus, Amazon's AI shopping assistant, pulls from your listing content to answer shopper questions — and it treats A+ not as decoration but as evidence. The listings that win Rufus answers are the ones whose A+ states facts in text a machine can extract. Most A+ we audit does the opposite: it locks the best facts inside artwork where no machine will ever find them.

Does A+ content help with Rufus? Yes — as evidence

The direct answer: yes. Rufus assembles answers from what it can read on and around your listing — the title, bullets, structured attributes, customer reviews, and the text content of your A+ modules. A+ sits in that stack as corroboration. The title and bullets assert a claim in a few words; A+ is where the claim gets substantiated with detail; reviews confirm it from lived experience. When those layers agree, Rufus can answer confidently in your favor. When your A+ contains nothing a machine can parse, you have voluntarily removed a layer of evidence from your own case.

Concrete example. A shopper asks Rufus whether your skillet is oven-safe. Your bullet says "oven-safe." If your A+ states "oven-safe to 260 C / 500 F, including the lid" in a standard text module, Rufus now has a specific, consistent, substantiated fact: a temperature, a scope, and two sources on the listing that agree. If that same fact exists only as stylized type on a lifestyle banner, Rufus finds a one-word bullet claim with nothing behind it. Same design intent, very different machine outcome — and the difference decides whether the answer a shopper hears is "yes, up to 260 C" or "the seller says it's oven-safe."

It helps to picture where this surfaces. Rufus shows up as a chat panel and as suggested questions on the detail page — and in both places the shopper sees extracted facts, never your layout. Nobody admires a banner inside a Rufus answer. The design work still matters for humans who scroll, but the machine pathway strips your A+ down to whatever survives extraction: text, values, table cells. That residue is what competes for the answer.

Machine-readable corroboration

Restating and expanding a product claim in indexable text so an AI assistant can verify it against the title and bullets. A+ content is the natural home for this on Amazon: it has the space the bullets lack and the seller-controlled detail that reviews cannot provide.

Not all modules are equal: where extractable facts live

The A+ module library splits cleanly into modules that carry text as text and modules that carry text as pixels. Machines can only use the first kind. Before touching your layout, know which side of the line each module sits on:

• Standard comparison chart — a real table: attribute rows against up to six ASINs. The most structured content you can publish anywhere in A+.
• Standard text and product description text — full paragraphs of indexable copy. Unglamorous, and doing far more machine-readable work than anything prettier.
• Standard tech specs — label-value pairs. The closest A+ gets to genuinely structured attribute data.
• Image-and-text modules (four image & text, three image & text, single image & sidebar) — the headlines and body captions are real text; the images are not. These earn their keep only if you use the text fields for facts.
• Image-led modules (image header with text overlay, full-width banners, company logo) — visually dominant, near-zero extractable content. The overlay text fields are tiny and the artwork itself is a black box.

The uncomfortable implication: an A+ layout scored on how it looks and an A+ layout scored on what a machine can extract are close to inverses. The banner-heavy design that sails through creative review is usually the one carrying the least evidence.

The comparison chart is the single highest-value module

If you change one thing after reading this, build or fix your comparison chart. It is the only A+ module that publishes a genuine table — attributes as rows, sibling products as columns, discrete values in cells. That is the same shape as the structured data agents reason over best, and it directly serves one of the most common question shapes on Amazon: "what's the difference between this one and that one?" A well-built chart lets Rufus answer that question with your data, in your framing.

It also keeps the comparison inside your brand. When a shopper — or Rufus on their behalf — compares your 12-inch skillet against your 10-inch, the chart frames the tradeoff in your terms and both outcomes are your sale. Skip the chart and the comparison still happens; it just happens against a competitor, framed by whoever did publish the data. A shopper deciding between two of your sizes is one ambiguous answer away from opening someone else's listing — a chart that states the tradeoff plainly resolves the doubt inside your product family instead of widening the search.

• Rows are attributes shoppers actually ask about — capacity, material, weight, compatibility, what's in the box. Not marketing rows like "style" or "inspiration."
• Values are concrete and consistent — numbers with units, yes/no, named standards. A checkmark beats a paragraph; a real value beats a checkmark.
• Every sibling ASIN carries the same chart, so the data agrees no matter which listing Rufus reads it from.
• The row set mirrors your [backend attributes](/insights/amazon/amazon-backend-attributes-guide), so structured data and visible content corroborate each other instead of drifting apart.

Can Rufus read text inside images? Treat the answer as no

The direct answer: assume no. Whatever image-text extraction Amazon does or does not run over A+ assets, none of it is promised, none of it is observable, and none of your baked-in text behaves the way real listing text demonstrably does. The operating rule that keeps you safe is simple: a fact that exists only as pixels does not exist for Rufus. That covers spec callouts on banners, infographic-style feature grids, badges, and the beautiful type your designer set over a lifestyle shot.

A+ images do carry one machine-readable field: the image keyword — alt text — you set per image in the A+ manager. Fill every one of them; it helps indexing and accessibility at zero cost. But it holds a few plain words, not a spec sheet. It is a caption, not a substitute for a text module.

Module-by-module priority order

When budget or bandwidth forces choices, work down this list. It is ordered by how much machine-readable evidence each module can carry per hour of effort:

Reworking existing A+ without starting over

You rarely need a from-scratch rebuild. Most A+ already contains the right facts — they are just trapped in the wrong containers. The rework is a migration, and it runs in a fixed order:

1. 1

   Inventory the trapped facts

   Open your current A+ and list every factual claim that exists only inside an image: specs on banners, feature grids, compatibility callouts, certification badges. That list is the scope of the work.

2. 2

   Move claims into text modules

   Add or repurpose standard text and image-and-text modules so each trapped fact is stated in indexable copy — with units, scope, and phrasing that matches the bullets word for word.

3. 3

   Build the comparison chart

   Attribute rows shoppers ask about, concrete values, every sibling ASIN on the same chart. If a chart already exists, replace the marketing rows with factual ones.

4. 4

   Fill every image keyword field

   Plain-language alt text for each image: what it shows, in a few words. A small win at zero cost — do it in the same pass so it never becomes its own project.

5. 5

   Reconcile with the rest of the listing

   A+ that contradicts the title or bullets is worse than silence — conflicting facts give an assistant a reason to hedge or decline. One fact, one phrasing, everywhere. Our Rufus listing optimization guide covers the title-and-bullet side of the same discipline.

Common A+ mistakes in the Rufus era

• Designing first, writing second. When banners come before copy, facts end up as decoration. Fix the brief, not just the output.
• Charts full of checkmarks and no values. A grid of ticks against vague rows tells a machine almost nothing. Real values in real units.
• Different facts in different places. A+ that says 500 F where the bullet says 450 F does not add evidence — it subtracts trust.
• Treating Brand Story as a spec sheet. The carousel is image-led with thin text fields. Use it for who you are; keep product facts in A+ text and the chart.
• Stopping at Amazon. The off-site assistants — ChatGPT, Claude, Gemini, Perplexity — reward exactly the same facts-in-text discipline on your own storefront.

Where A+ sits in the wider evidence stack

A+ is one layer of a stack that has to agree with itself. Backend attributes structure the facts. The title and bullets assert them. A+ substantiates them. Reviews confirm them. The same structure-over-prose discipline drives catalog enrichment on your own storefront, where the identical facts feed the assistants shaping what shoppers already want before they ever type "amazon.com." Merchants who treat A+ as an isolated design asset end up re-litigating the same facts channel by channel. Merchants who treat it as one output of a single fact source fix it once and publish it everywhere.

The honest summary: your A+ was almost certainly approved by people judging aesthetics, and it is now being read by a machine judging evidence. Both readers matter. Only one of them was in the brief. Update the brief.