The product description is the most neglected piece of text on most online stores. It gets pasted from the supplier’s spreadsheet during the catalog import, nobody reads it again, and it ends up identical to the description on fifty other stores selling the same item. Then the owner wonders why the page never ranks and why shoppers keep asking support the same three questions.
A description has two jobs that most stores treat as one. It has to give search engines enough unique, specific text to understand what the page is and which searches it deserves, and it has to give a shopper enough confidence to buy without picking up the phone. Most of the reasons a product page fails to rank come back to the description, which is why it sits near the top of the list in our breakdown of why most ecommerce product pages never rank.
The cost of getting it wrong is measurable. Salsify’s 2026 consumer research, a survey of nearly 3,000 shoppers in the US, UK, and Canada, found that 38% have abandoned a purchase because product information was inconsistent across websites, and 45% returned an online purchase in the past year because the product content was wrong. Those are lost sales and refunded sales caused by text the store controls completely.
This guide covers how to write descriptions that do both jobs: what to research first, how to structure the page, where keywords belong, how to handle variants and manufacturer copy, how to use AI without creating a spam problem, and how to keep the same product described consistently everywhere it is sold.
What a product description actually does
A description is read by more systems than the shopper on the page. It feeds the snippet Google shows under your listing, the product data that marketplaces and shopping feeds pull from, the answers AI assistants give when someone asks about the product, and the on-page relevance signals that decide which searches the page can rank for. Google’s documentation on how search snippets are created is explicit that snippets come mainly from the page content itself, with the meta description used only when it describes the page more accurately. A thin or generic description therefore produces a thin or generic snippet, whatever you put in the meta tag.
That makes the description the piece of content with the widest reach on a product page, and the one where effort compounds: every channel that reuses it gets better when it gets better.
Start with the buyer’s questions, not the spec sheet
The supplier’s specification sheet answers the questions a warehouse asks: dimensions, weight, material code, carton quantity. Shoppers ask different questions. Will it fit? Is it the right size for my use? How does it compare to the cheaper one? What happens if it breaks? Nielsen Norman Group’s ecommerce product page guidelines make anticipating and answering product questions one of the core principles of a good page, and the description is the natural place to do it.
Before writing a word, collect the questions from the places where shoppers already ask them:
- Support tickets and chat logs, filtered by product or category, for the questions your team answers repeatedly
- Product reviews, your own and competitors’, for the details buyers mention after purchase (sizing, durability, what surprised them)
- Returns reasons, which point directly at what the description failed to communicate
- “People also ask” and autocomplete for the product type, which show the questions people type before they find a store
- Marketplace Q&A sections on Amazon and similar sites, where buyers ask sellers directly
- Your sales or showroom team, if you have one, for the objections they hear in person
The output is a short list of five to ten questions per product type. Those questions become the skeleton of the description, and they are the reason a rewritten description converts better: it answers what the shopper was going to ask instead of what the factory chose to print.
Pick the words before you write the sentences
Product pages target narrow, specific searches, usually the product type plus one or two attributes (“waterproof hiking boots wide fit”), the brand and model, or a use case (“boots for wet trail running”). The category page above it targets the broader term. Getting this split right keeps product pages from competing with their own categories, and the mapping process we described in our guide to ecommerce keyword research applies directly: one primary term per page, a handful of variations, and the category keeping the head term for itself.
Once the terms are chosen, they go in a small number of specific places:
- The product name and H1, using the term shoppers search rather than the internal range name
- The first sentence of the description, stating plainly what the product is and who it is for
- One subheading, where the page structure allows it
- The image alt text, describing the product as a person would
- The page title, if your platform lets it differ from the product name
Everywhere else, write naturally. Repeating the primary term in every paragraph adds nothing that the first mention did not already establish, and it makes the copy harder to read.
A structure that works for most products
There is no single template, but most strong product descriptions share a shape that serves both skimmers and careful readers. A useful default:
- Opening summary (two or three sentences): what it is, who it is for, the one reason to choose it
- Benefit sections (one short paragraph each): each tied to a feature and one of the buyer questions from your research
- Specification table: dimensions, materials, weight, compatibility, everything a comparison shopper needs to check
- Use and care: how to set it up, use it, clean it, store it
- What is in the box: especially for anything sold as a kit or set
- Short FAQ: the questions that did not fit naturally above
The opening summary carries the most weight because it is the part most shoppers read and the part most likely to become the search snippet. Write it last, once the rest is clear, and make every word in it specific. “Premium quality” tells nobody anything; “full-grain leather, stitched by hand, resoled for free for five years” tells them what they are paying for.
Manufacturer copy, duplicates, and variants
Copying the manufacturer’s description is the fastest way to a page that cannot rank, for a simple reason: Google has already seen that text on every other store that imported the same feed, and it has no reason to rank your copy of it above the others. Duplicate descriptions do not trigger a penalty in the usual sense; they make the page indistinguishable, and indistinguishable pages lose to the stores with more authority.
Rewriting every description in a large catalog is rarely realistic, so prioritize. Start with the products that already sell and the ones in categories where you want to rank, then work outward. Keep the manufacturer’s specification data, which is factual and expected to match, and rewrite the prose around it.
Variants need their own rule. Products that differ only by color or size should share one description on one page, with the variant selector doing the work. Variants that differ in ways a buyer cares about (a different material, a different capacity, a different intended use) deserve their own description, and sometimes their own page, because they answer different searches.
Using AI without creating a spam problem
AI has made it possible to draft thousands of product descriptions in an afternoon, and that capability is exactly where the risk sits. Google’s guidance on AI-generated content is clear that the method of production is not the issue: using automation, including AI, to generate content primarily to manipulate rankings violates its spam policies, while using it to produce helpful content does not. The question is what the output is for and whether it adds anything.
Feed the model what a writer would use
In practice, AI works well for descriptions when it is fed the inputs a human writer would use: the specification data, the buyer questions you collected, review excerpts, and your brand’s tone guidelines. It works badly when it is fed only the product name and asked to “write an SEO description”, because it then produces the same generic paragraph a thousand other stores generated from the same prompt. Our guide to generative AI for product descriptions, emails and ads covers the prompting and workflow side in more depth. For SEO purposes, the rule is that every AI draft gets a human edit that checks facts against the spec sheet and adds at least one detail the model could not have known.
Templates and the line Google draws at scale
Large catalogs tempt stores into description templates: “The [Brand] [Model] is a [color] [type] perfect for [use].” A template filled with attributes produces technically unique text that says nothing a shopper could not read from the specification table, and at scale it starts to look like what Google’s spam policies call scaled content abuse: many pages generated primarily to manipulate rankings rather than to help users.
Templates are not banned, and for some product types (replacement parts, consumables, items with dozens of near-identical sizes) structured, attribute-driven text is exactly what a buyer needs. The test is usefulness: does the description help someone choose between this item and the one next to it? If the template only rearranges the attributes, put the attributes in the specification table, where they belong, and keep the prose for what actually distinguishes the product.
Let customers write part of the description
Reviews and customer questions are product content that you did not have to write, and they cover the angles a store’s own copy usually misses: how the item holds up after six months, how it fits a particular body type, what it looks like in a real room rather than a studio. They also add fresh, relevant text to the page every time a new review arrives. How reviews shape purchase decisions in general is covered in our analysis of how online reviews influence purchases; for description writing, the practical moves are these:
- Render reviews in the page HTML, not only inside a widget loaded after the page, so their text counts as page content
- Answer customer questions publicly on the product page and fold the best answers into the description over time
- Quote recurring praise in the description itself when it is specific (“fits a 16-inch laptop with room for a charger”)
- Fix the description when reviews reveal a gap, for example when several buyers mention the sizing runs small
The Salsify survey cited at the top of this article found that 57% of shoppers rank ratings, reviews, and user-generated content among the most important elements on a product page. That is close to the 60% who name pricing and availability, and it reflects how much shoppers trust other buyers over the seller’s own claims.
Premium products need more words, not fewer
There is a persistent belief that shorter is always better for conversion. For low-price, low-consideration items, it often is: nobody needs four paragraphs about a phone charger. For expensive or considered purchases, the opposite holds. A buyer about to spend several hundred on a single item wants the material origin, the construction details, the warranty terms, the comparison with the model below it, and evidence that the claims are true. Thin descriptions on premium products read as either careless or evasive.
The positioning side of selling at the top of a category, including how pricing and presentation work together, is the subject of our guide to high-ticket ecommerce. For descriptions, the rule of thumb is to match length to the decision: the more a shopper has to think before buying, the more questions the description should answer.
One product, consistent everywhere
The same item often appears on your store, on Amazon, in a Google Merchant Center feed, and on a reseller’s site. When those descriptions disagree on dimensions, materials, or what is included, shoppers notice, and the Salsify finding about inconsistent information driving abandonment is the result. Keep one master description and specification set per product, then adapt it per channel rather than writing each version from scratch.
Adaptation matters because each channel has its own rules. Amazon limits title length, prohibits promotional phrases, and reads bullets and backend terms differently from a store’s product page, as we covered in our guide to Amazon SEO. Feeds have character limits. Your own store has none of those constraints. The facts stay identical everywhere; the format changes to fit.
Measuring whether rewrites worked
Rewritten descriptions should be judged on numbers, not on whether they read better. Before rewriting a batch, record each page’s organic clicks and impressions from Search Console, its conversion rate, and its return rate, then compare the same metrics over the same length of time after the change. Look at the right signals:
- Impressions for new queries, which show the page is now relevant to searches it previously missed
- Click-through rate, which reflects whether the new opening summary makes a better snippet
- Conversion rate, which shows whether the description answers the questions that used to stop buyers
- Return rate and “not as described” returns, which show whether expectations now match reality
- Support tickets per order for the product, which should fall when the description answers questions up front
Rewrite in batches of 20 to 50 products in one category, measure for six to eight weeks, then decide whether to roll the approach out to the next category. That rhythm makes it possible to separate the effect of better descriptions from seasonality and from everything else changing on the store at the same time.
Where to start
Pick your 20 best-selling products and your 20 products with the highest return rates. For each, collect the buyer questions, rewrite the opening summary, add a proper specification table, and fold in what reviews say. That small batch will show you more about what your customers need to read than any template, and it will give you a working pattern for the rest of the catalog.