How to Write Amazon Bullet Points That Rufus Actually Recommends

Your Bullet Points Have a New Reader

There's a bot reading your bullet points right now. Not a search crawler. Not an indexing spider. An AI shopping assistant called Rufus that answers customer questions using your listing content — and decides whether to recommend your product or your competitor's.

Rufus is now processing over 20% of Amazon's roughly 2 billion daily searches. When a shopper asks "what's a good water bottle for hiking," Rufus doesn't just match keywords. It reads your bullets, your description, your reviews, your images, and it decides whether your product actually answers that question. If your bullets say "STAINLESS STEEL WATER BOTTLE 32OZ VACUUM INSULATED BPA FREE SPORTS GYM HIKING CAMPING" — that's not an answer. That's a word salad. Rufus skips it.

This changes what "optimized bullet points" means. The old playbook — cram keywords, follow the character limit, slap on a CAPS opener — still matters for baseline compliance. But if that's all you're doing, you're optimizing for an algorithm that's being replaced.

The Algorithm Shift You Can't Ignore

Amazon's search stack now has three layers that read your bullets differently:

A9 (legacy search) still does keyword matching. It checks whether "insulated water bottle" appears in your listing when someone searches that phrase. This is the system most bullet point guides are written for. It still matters, but it's no longer the whole game.

COSMO (knowledge graph) is the layer underneath. It's a massive AI framework with over 6.3 million nodes and 29 million relationship edges that maps products to shopping intents. COSMO doesn't just see that your bullet says "insulated" — it understands that insulated bottles relate to hiking, gym use, keeping drinks cold, and outdoor activities. It builds a web of intent around your product based on how completely your content describes it.

Rufus (customer-facing AI) is what the shopper actually interacts with. It's a text-based LLM that generates answers by reading your product detail page. The more complete and clearly structured your content is, the more likely Rufus recommends your product in response to a customer question.

Here's the critical insight: COSMO powers Rufus. COSMO is the knowledge infrastructure; Rufus is the interface. When you write better bullets, you're feeding both systems at once.

The 1,000-Byte Indexing Limit Nobody Talks About

Most sellers know about the per-bullet character limit (500 characters for most seller categories, 255 for vendors, with category-specific caps as low as 100-150 for some Apparel subcategories). We covered the specifics in our character limit breakdown.

But here's a limit that barely gets discussed: Amazon only indexes the first 1,000 bytes total across all five bullet points for search ranking.

Your bullets can display up to 2,500 characters total (5 x 500). Customers see all of that text. But for search purposes, only the first ~1,000 bytes count toward keyword indexing. That's roughly 200 characters per bullet if you distribute evenly.

This means if you're writing 400-character bullets and putting your important keywords in the back half, those keywords might not be indexed at all. The full text displays to the customer, but the search algorithm stops reading after 1,000 bytes.

Practical takeaway: Front-load every bullet with your most important keywords in the first 200 characters. Save the back half for persuasive copy that helps the human buyer, not the search engine.

Write Bullets as Answers, Not Feature Lists

This is the single biggest shift in how to write bullets in 2026. Rufus answers questions. Your bullets need to be answers.

Think about how customers actually use Rufus. They don't type "stainless steel vacuum insulated water bottle 32oz." They ask things like:

• "What's a good water bottle that keeps drinks cold all day?"
• "Is this water bottle dishwasher safe?"
• "Will this fit in my car cup holder?"
• "What's a good gift for someone who hikes?"

Your bullet points need to contain the answers to these questions — phrased in natural language that an LLM can extract and present to the customer.

The second version answers "how long does it keep drinks cold?" and "does it sweat?" and "does it taste metallic?" — three real questions Rufus gets asked. The first version just lists a specification without explaining what it means for the buyer.

Category-by-Category: What Rufus Cares About

Different product categories generate different kinds of customer questions. Your bullets need to anticipate the questions your specific buyers are asking.

Supplements & Health Products

Shoppers in this category ask about dosage, ingredients they're avoiding, and whether a product fits their specific health situation. Rufus needs clear, factual answers.

The second bullet anticipates "how much should I take," "when should I take it," and "why olive oil?" Those are the questions that show up in Rufus queries for supplements.

Note: supplements have strict limits on health claims. You can't say "cures" or "treats" anything. Stick to structure/function claims ("supports bone health") and be specific about dosage and ingredients. Amazon's enforcement on supplement bullets is particularly aggressive — prohibited claims get your listing suppressed fast.

Electronics & Tech

Tech buyers want compatibility details, specifications that matter, and answers to "will this work with my setup?"

The second version answers "will these work with my iPhone," "what's the range," and "do they disconnect randomly" — a top complaint in earbud reviews that Rufus knows about because it reads reviews too.

Home & Kitchen

Home products get comparison questions. "Is this big enough?" "Will it match my kitchen?" "Is it easy to clean?" Your bullets need dimensions, materials, and practical usage information.

The second version answers "how big is it," "is it heavy," and "does it have juice grooves" — without wasting a single word on "professional grade" (a phrase that means nothing and Amazon increasingly flags as an unsubstantiated claim).

Clothing & Apparel

Clothing categories often enforce shorter bullet point limits — some subcategories cap at 100-150 characters. That makes every character count double. Apparel shoppers ask about fit, fabric feel, care instructions, and occasions.

Both fit within typical Apparel limits. But only the second one answers "what's the fit like," "what's it made of," "how do I wash it," and "will it shrink" — all within 128 characters.

The Structure That Feeds Both Algorithms

Here's a bullet point formula that satisfies keyword indexing (A9), intent mapping (COSMO), and question-answering (Rufus) simultaneously:

CAPS BENEFIT HEADLINE - [Specific claim with number or measurement] + [Why it matters to the buyer] + [Use case or scenario]

Each component does a job:

• CAPS HEADLINE (3-6 words): scannable for human shoppers, contains primary keyword for A9
• Specific claim: gives COSMO factual data to map your product to intents
• Why it matters: natural language that Rufus can extract as an answer
• Use case: triggers COSMO's intent graph ("good for hiking" links to outdoor, travel, fitness intents)

Here's that formula applied five times for a single product — a portable phone charger:

1. CHARGES AN iPHONE 15 TWICE ON ONE CHARGE - 10,000mAh capacity delivers 2 full charges for iPhone 15 or 1.5 charges for Samsung S24. Throw it in your bag and stop hunting for outlets at the airport
2. SMALLER THAN A DECK OF CARDS - 4.1 x 2.7 x 0.6 inches and 6.5oz. Fits in a front jeans pocket. Most 10,000mAh chargers are nearly double this size
3. CHARGES 2 DEVICES AT ONCE - One USB-C port (20W fast charge) and one USB-A port (12W). Your phone and earbuds at the same time, no dongles or adapters needed
4. SHOWS EXACT BATTERY REMAINING - LED display shows percentage (not vague dots). You'll know if you have enough juice before you leave the house
5. BUILT TO SURVIVE YOUR BAG - Passed 1,000 drop tests from 4 feet. Scratch-resistant casing. Includes a USB-C to USB-C cable and a travel pouch. 18-month warranty with US-based support

Each bullet answers a question Rufus fields regularly: "How many charges does it hold?", "How big is it?", "Can it charge two things?", "How do I know when it's dead?", "Is it durable?"

Each bullet also hits keywords naturally: "10,000mAh," "iPhone 15," "Samsung S24," "USB-C," "fast charge," "portable charger" — without stuffing a single one.

Stop Writing Bullets Your Competitor's VA Wrote

I want to say something blunt here. Most Amazon bullet points read like they were written by someone who has never used the product, doesn't speak the customer's language, and was paid $3 per listing to hit a word count. Because that's exactly what happened.

You can tell immediately. The bullets are stuffed with keywords that don't form coherent sentences. The benefits are vague ("premium quality," "professional grade," "ultimate performance"). The use cases are generic laundry lists ("great for gym sports hiking camping traveling outdoor activities daily use"). There's no specificity because the writer doesn't know the product and doesn't know the customer.

Rufus is trained on millions of reviews. It knows what customers actually care about for your product category — because customers wrote about it, in their own words, thousands of times. If your bullets don't reflect the language and concerns in your reviews, Rufus will favor the competitor whose bullets do.

One approach that works: read your competitors' 3-star reviews. The 3-star reviews are gold because they come from people who mostly liked the product but had specific complaints. Those complaints tell you exactly what buyers in your category care about. Address those concerns directly in your bullets. "No metallic taste" isn't a random claim — it's a response to the #1 complaint in water bottle reviews.

Testing What Works: Manage Your Experiments

If you're brand-registered, Amazon's Manage Your Experiments tool lets you A/B test your bullet points directly. You create a Version B, Amazon splits traffic, and after enough data accumulates, you see which version drove more conversions.

A few things most guides don't mention about bullet point experiments:

• Run tests for at least 8-10 weeks. Shorter tests rarely reach statistical significance unless you're doing serious volume.
• Test one change at a time. If you rewrite all five bullets, you won't know which change drove the result.
• Amazon now offers multi-attribute experiments that test title + bullets + images simultaneously. Useful for a full listing overhaul, but the results are harder to attribute to a single element.
• Machine learning suggestions are available — Amazon will suggest bullet point variations derived from experiments on similar products. These are hit-or-miss, but worth reviewing as a starting point.

You need a Professional selling account and Brand Registry to access this. If you have it and you're not running experiments, you're leaving data on the table.

Where a Tool Actually Helps

I'll be honest about what's hard here: writing five bullets that simultaneously hit keyword targets, follow the CAPS convention, stay within your category's character limit, front-load keywords in the first 200 characters for indexing, answer the questions Rufus cares about, and avoid Amazon's growing list of prohibited phrases — that's a lot of constraints to hold in your head at once.

SellScope handles the constraint management part. It knows your category's character limits, flags prohibited claims, structures the CAPS headline + benefit + use case format, and generates bullets with the important keywords front-loaded. You still need to know your product and your customer — no tool replaces that. But the formatting and compliance grunt work is exactly the kind of thing AI handles better than a person manually counting characters in a text box.

The Bottom Line

Your bullet points now have two audiences: human shoppers scanning on their phones, and an AI assistant that reads your content to answer questions on your behalf. The sellers who treat bullets as keyword containers are going to lose visibility to sellers who treat bullets as clear, specific answers to the questions their customers actually ask.

Pick your top-selling product. Open Rufus in the Amazon app. Ask it questions about your own product category — "what's the best [your product type] for [common use case]?" See if Rufus mentions your product. If it doesn't, read the bullet points of the products it does recommend. You'll see the difference immediately.