Why most listing copy underperforms

When we audit Amazon listings, the copy usually fails in one of two ways. Some reads like a list of search terms strung together. It may rank, but shoppers struggle to work out what the product does, and they leave. The rest is written in brand voice alone. It sounds good, but it leaves out the words shoppers actually type.

Both miss the same thing, the questions shoppers ask before they buy. In our recent audits, live copy answered between a quarter and a third of the shopper questions we tracked for each product. The answers were often somewhere, in a review, on the pack or on the brand's website, but not in the listing.

That gap matters more every month. AI shopping assistants now read the listing to decide whether a product suits a shopper, and a question the copy doesn't answer is a reason not to recommend it.

What the copy has to do now

A listing's words now have four jobs.

• Be found. Carry the terms shoppers search for, on Amazon and off it, in the places Amazon indexes.
• Be chosen. Make the main benefit clear in seconds and deal with the doubts that stop a sale.
• Be recommended. Answer the questions an assistant is asked, in language it can match to a shopper's need. Amazon's own research into shopper intent links products to who they're for and what they're used for, not just to keywords.
• Stay within the claims. Say only what the brand can prove, which matters most in regulated categories like pet health, skincare and medicines.

Doing all four at once is hard to do by feel. So we start with the evidence.

Step 1. Gather the evidence

Before anyone writes a word, we collect four kinds of evidence for every product.

• Reviews. We read every review and group what shoppers love, what puts them off and the words they use to describe the product. Recurring doubts tell us what the copy has to answer.
• Search and question data. Keyword data we license shows the terms shoppers use and how often, on Amazon and on Google. We add the questions they ask on both, and the prompts they put to AI assistants.
• Approved claims. The pack copy, the claims documents the brand has signed off and the claims already live on the listing. Nothing else counts as a claim.
• Competitors. What rival listings promise, which questions they answer and where they leave gaps.

We also take the brand's voice, its science and anything it has told us it can't say. Brand educators and product experts are often the best source of the practical tips shoppers want.

Step 2. Build the information hierarchy

Next we rank every message by how much evidence sits behind it. A benefit that is the top theme in reviews, matches the top search term and is an approved claim goes first. A message with only one of those goes lower down.

The result is a priority order that tells the copywriter, and the designer, what each part of the listing has to say. This is the hierarchy we built for Feliway Optimum.

Search shaped the order as much as the brand did. Shoppers search for "kitten" around 13,600 times a month and "cat calming" around 4,300, so both terms earned a place in the title.

Step 3. Write to the rules

With the hierarchy agreed, the writing follows a few rules we apply on every account.

Only use approved claims

If a claim isn't on the pack, in a signed off claims document or already live on the listing, it doesn't go in. Shoppers often search for something a product can't claim, such as a condition it isn't licensed to treat. Those terms go in the backend search terms, never in the copy. A claim also travels with its conditions. A ranking claim like "number one brand" only runs with the disclaimer that supports it.

Never answer no

FAQs and Q&A modules answer what the product does, not what it doesn't. A "no" gives the shopper a reason to leave. We answer with what the product is for, and point to the right product in the range if there is one.

Lead with the shopper's words

The title and first bullets use the terms shoppers search for and the phrases they use in reviews, not internal product names or category jargon.

Give every term one home

Across a range, we split core and emerging search terms between products, so they don't compete for the same searches.

Choose how far to go

For a live listing that already performs, a light edit may be enough. Most need a keyword boost that adds the most search terms without losing the brand's voice. A weak listing gets a full rewrite. Our tool Freak Copy runs at all three depths, so the choice is about risk, not effort.

Step 4. Check before it goes live

Every new listing goes through the same checks.

• Claims sign off. Each claim is traced to its source and approved by the brand's claims owner.
• Coverage before and after. How many target terms the copy now carries, how much search volume they represent and how many tracked questions it answers.
• A baseline. Rankings and share of voice for the target terms, captured before launch from the right marketplace. Read Amazon from the wrong delivery address and some products simply don't appear.
• Amazon's rules. Character limits and style rules for the category, in every marketplace.

Step 5. Measure one change at a time

Where Amazon lets us, we A/B test new copy against the old and keep the winner. Where it doesn't, we compare before and after against the baseline, allowing for seasonality, promotions and anything else that changed in the same weeks.

We only count products where the new copy actually went live. It sounds obvious, but uplift is often claimed for listings that were never updated.

Feliway and Douxo

We've used this process for Ceva's animal health brands on Amazon since 2024, starting from a master listing in the UK or US and localising it, with fresh keyword research, for every market the brand sells in.

Feliway

Feliway's diffusers help cats feel calm at home. We researched search terms in eight marketplaces, audited the main competitors and built a claims list with Ceva's category team, signed off by regulatory. That fed the hierarchy above, the titles and bullets, and frameworks that give every image and A+ module a single job.

The titles were written to use almost every character Amazon allows, with the words that matter most placed before the point where search results cut a title short. Across 64 products in 25 countries, conversion rose by 24% and ordered revenue by 48%. When we refreshed the UK range in 2026, the new Feliway Optimum image stack beat the old one by 15% in an Amazon experiment and is now live. Read the Feliway case study.

Douxo

Douxo is Ceva's skin and coat care range for cats and dogs, in a category where claims are tightly controlled. Shoppers search by need, so we wrote four title options for every product. The current title, one led by search, one led by the label claims and a short version from the brand. Each came with its search volumes and a recommendation, so Ceva could see exactly what each choice gained and gave up. A term like "dog shampoo" is searched around 295k times a month in the US.

Every claim came from the approved label or the brand's own study references, and Ceva's legal and regulatory teams checked every line. The masters now run in 12 European markets, each localised with its own keyword analysis.

The copy now answers the questions owners ask most, from how often to use a product to whether it suits a puppy or kitten, in the words they search with.

Where to start

Pick your five best selling products and check three things. Does the title lead with the benefit shoppers care about, in their words? Do the bullets answer the questions in your reviews? Is every claim one your brand has approved?

If the answer to any of them is no, our Amazon team can audit the listings, and Freak Copy can rewrite them from the evidence at scale, in every marketplace you sell in.

Sources

1. Building commonsense knowledge graphs to aid product recommendation, Amazon Science
2. A+ Content, Amazon