Product descriptions for an online store: method, template, batches

A good product description in an online store has three parts: a short description that says in 2–4 sentences in what situation the customer needs the product, a long description that answers pre-purchase questions, and a specs table with units. For SEO, one main phrase in natural places and your customers' vocabulary are enough. With dozens of products from one brand, you can prepare descriptions as a batch from the manufacturer's catalog, as long as every spec comes from exactly that model. The stakes are concrete: a wrong dimension in a boiler description means a return of goods worth thousands, and writing a hundred descriptions from scratch is a week of work.

Below I show how I do it in the termo-expert.com.pl store (boilers, heat pumps, ventilation with heat recovery), where I prepared 777 descriptions this way. At the end you'll find a description template, an AI prompt, a downloadable specs spreadsheet and a checklist.

What makes a good product description?

A product description has to answer three buyer questions: is this for me, how is it different from similar models, and what exactly do I get. Each answer is served by a different part of the product page.

The short description is 2–4 sentences. It says who the product is for and what its main feature is. An example for a 16 kW pellet boiler (a general example, not a specific model): “A pellet boiler for a home of about 1,300–1,700 sq ft (120–160 m²). Fuel feeds automatically from the hopper, so you top up pellets once every few days.” It takes 10 seconds to read, and the customer knows right away whether to keep looking.

The long description expands on what customers ask before buying: how to choose the output, how much space the boiler room needs, how it's controlled, what fuel it uses, what the warranty is. In a real description from the Defro batch I prepared on April 29, 2026, the long description has 9 sections and 931 words: from design and controls, through fuel and warranty, to buyers' questions. For a boiler, that length is justified. For a USB cable, it isn't.

How long should a description be: a decision rule

The length of a description follows from the number of questions customers ask before buying, not from a word count. Count the questions you hear about this product and give each one a short paragraph.

The numbers in the table are my working rule, not a Google standard. Google doesn't reward length, only whether the page answers the question better than other results.

How to write a product description that sells: a four-step method

A description sells when customers recognize their own situation in it, not when they read a list of benefits. The most popular guides tell you to write “in benefits”. That's a good start, but a benefit asks customers to imagine the future, and imagination is easy to brush off. It works better to show the product at the moment the customer is already in.

The copywriting methods I use have no dedicated course on product descriptions. The method below combines two of them: placing the product in the customer's situation (place, time, goal) and building one sentence with a tangible promise. I call it a method because it's a procedure, not a guaranteed result.

1. List the buying situationsWhen does a customer look for this product? An example for a boiler: the old boiler failed inspection, the heating season starts in a month, a new house is at the boiler-room stage. Each situation is a candidate for the first sentence and for a subheading.
2. Pick one situation for the short descriptionThe short description has one place and one goal. Don't mix a family home, a workshop and a subsidy in it. Move the other situations to the long description.
3. Write a sentence with a tangible promiseWhat the customer gets and what it changes for them, in numbers from the catalog. Instead of “convenient to use”, write “you top up pellets once every few days”. If you can't show it with a number or an action, the promise is too vague.
4. Answer questions in the order of the decisionFirst “is this for me” (sizing), then “how is it different” (comparing models), and last “what do I get” (specs, what's included, warranty). Each question is a subheading with the answer in its first sentence.

The same order helps when you hire a copywriter. Professional product descriptions don't come from writing talent alone: the copywriter needs the buying situations, the customers' questions and the specs spreadsheet from you. Without them, you'll get a nice text about “top quality”.

A before → after example for one product

A made-up example with no real brand: a heat recovery ventilator for a 1,500 sq ft (140 m²) home.

All the numbers in the “after” version come from the same catalog as in the “before” version. What changed is the layout and the starting point: the customer's situation first, then the data.

How should you show product specs in a table?

Put specs in a real table: one feature per row, the value in a separate cell, the unit in the feature name. A buyer comparing two models reads this layout, and so do search engines that pull “feature – value” pairs into results and AI answers. Specs written into a paragraph (“the boiler has an output of 16 kW and an efficiency of over 90%”) are harder to compare and easier to get wrong.

The template I use for boilers (values in parentheses are filled in from the manufacturer's catalog):

Three rules that save you complaints:

• One value per cell. Split “16 kW / 4.8–16 kW” into two rows.
• The unit in the feature name, not in some cells and not in others.
• The same numbers everywhere. The output in the table, in the short description, in WooCommerce attributes and in the data for Google Merchant Center has to be identical. A mismatch in one place undermines trust in the whole page, for both the buyer and the search engine. Google describes consistent product data in its guide to sharing product data.

Collect specs in a spreadsheet first, and only then in the description. The spreadsheet below has a column for the page and the catalog column each value comes from, so reviewing a batch means comparing two numbers instead of searching a PDF.

How do you write an SEO product description without keyword stuffing?

An SEO product description answers search questions in full sentences with specific vocabulary. It isn't a description with the phrase repeated ten times. Put the main phrase, usually the product type with a feature (“16 kW pellet boiler”), in the product name, the first sentence of the long description and one subheading. That's enough.

Vocabulary does the rest. Google weighs words in the context of the paragraph: a paragraph with “hopper”, “auger feeder”, “burner”, “class 5” and “room thermostat” is clearly about a pellet boiler. A paragraph with the phrase repeated three times and filler (“high quality, modern design”) says nothing. Google explicitly lists keyword stuffing in its spam policies.

What helps in practice:

• Subheadings as buyers' questions: “What boiler output do I need for a 1,600 sq ft home?”, with the answer in the first sentence below it.
• Paragraphs that make sense outside the page. “This model also works well in this case” means nothing without context. “A 16 kW boiler heats a home of about 1,300–1,700 sq ft with good insulation” does. AI assistants that answer product questions quote sentences like that too.
• A separate meta title and meta description. That's the text in the search results, not part of the description. I explain how to write it in the post Meta title and meta description for your store.
• A Google Merchant Center description from the same data. The description attribute in the product feed has its own requirements, different from the description on the page; they're explained in Merchant Center Help.

Why shouldn't you copy the manufacturer's description?

Dozens of stores use the manufacturer's description at once, so for the same text Google usually shows one version, most often the manufacturer's or the biggest seller's. Your product page adds nothing new and loses before ranking even starts.

The second problem is less obvious: the manufacturer's marketing department writes the description for all markets at once. It doesn't contain what your customers ask about. On the phone at Termo Expert, I most often hear questions about matching the output to the floor area, space in the boiler room and subsidies. The manufacturer's catalog doesn't cover that, but a good description should.

From the manufacturer's description, take the data: specs, what's included, standards, warranty. Write the text yourself, for your customer. Google explains this in its guide to helpful, people-first content: what counts is whether the page offers something beyond what other results already have.

Which product description template can you copy?

The template below has every part of the description in the order of the customer's decision, with instructions in square brackets. Copy it, delete the instructions and fill it in with data from the specs spreadsheet.

SHORT DESCRIPTION (2–4 sentences)
[Sentence 1: who it's for and in what situation, with a number from the catalog, e.g. floor area, capacity, size.]
[Sentence 2: one main feature shown as something the customer does, e.g. “you top up pellets once every few days”.]
[Sentence 3, optional: how this model differs from the closest one you sell.]

LONG DESCRIPTION

[First paragraph: the main phrase (product type with a feature) and the answer to “is this for me”. 2–3 sentences.]

Who [product type] is for
[Buying situations: when the customer needs it. 2–4 sentences, no vague talk of quality.]

How to choose [output / size / capacity]
[The sizing rule in one sentence, then an example with numbers. If sizing depends on conditions, say which ones.]

How it differs from [neighboring model / lower version]
[One or two differences in numbers. Skip this section if you don't sell a similar model.]

Design and what's included
[What's in the box, what isn't and has to be bought separately. Catalog data only.]

Technical specs
[Table: Spec (unit) | Value. One value per cell. Skip a missing value; don't guess.]

Installation, maintenance and warranty
[Who installs it, what needs servicing, the warranty period from the catalog. Don't promise dates or services you don't provide.]

Frequently asked questions
[3–5 questions you really hear on the phone and by email. Answer in the first sentence, max. 3 sentences.]
- [Question 1?]
[Answer.]
- [Question 2?]
[Answer.]
- [Question 3?]
[Answer.]

When does AI help with product descriptions, and when does it make up specs?

AI helps when it gets the data and only writes the text. A language model such as Claude or ChatGPT is good at arranging sections, writing clearly and keeping the tone across a hundred descriptions. It's bad with numbers it wasn't given: asked about the output or dimensions of a specific model without a catalog, it will give a value that sounds credible but may be made up.

How to prevent it:

• Give it data; don't make it guess. Specs from the manufacturer's catalog or your spreadsheet, for every description.
• Exact match, not approximate. When the catalog doesn't have exactly this model, the description shouldn't be written. An empty row in the batch is better than a boiler with its sibling's specs.
• Gaps flagged, not filled in. Tell the model to list missing data separately instead of filling it in.
• No promises you don't keep. Write in the instructions what it must not promise: delivery dates, installation, subsidies, if you don't decide on them.
• Review before saving. Nothing goes into the store automatically. You check the table and the promises, and skim the rest.

The prompt below applies these rules in a regular AI chat. You paste the data between the markers, and the model has two ways out: a description or a message that the model is missing. It doesn't guarantee the model will never make a mistake, which is why the last part asks it to list the source of every number, and that shortens your review.

Write a product description for an online store based only on the data below.

PRODUCT: [SKU and exact model name, e.g. as in the catalog]
CUSTOMER AND SITUATION: [who buys and when, e.g. owner of a 1,300–1,700 sq ft home replacing a boiler before the season]
CUSTOMER QUESTIONS: [3–5 questions you hear on the phone]
DO NOT PROMISE: [e.g. delivery dates, installation, subsidies]

CATALOG DATA (the only source of numbers):
<<<
[paste the catalog excerpt or rows from the specs spreadsheet: attribute; value; unit; page]
>>>

RULES:
1. First check whether the data is about exactly the model in the PRODUCT field. If the data covers several models (e.g. columns with different outputs), use only this model's column. If this model isn't there, don't write a description. Reply only: “MODEL NOT IN DATA: [name]” and stop.
2. Every number, dimension, standard and warranty period must come from the DATA section. Don't convert units, don't round and don't fill in values from your own knowledge.
3. If a template section lacks data, skip it or the table row. Don't write “no data” in the description.
4. Don't use the words: best, innovative, top quality, perfect. Instead of adjectives, show an action of the customer or a number.
5. Use the main phrase [product type with a feature] in the first sentence of the long description and in one subheading. Nowhere else on purpose.

FORMAT:
- Short description: 2–4 sentences, the first about the customer's situation.
- Long description: subheadings as customer questions, the answer in the first sentence of each section, paragraphs of 2–4 sentences.
- Specs table: Spec (unit) | Value, one value per cell.
- FAQ: only questions from the CUSTOMER QUESTIONS field that the data lets you answer.

AT THE END, separately, outside the description:
- CHECK: a list of every number in the description with the page or data row it comes from.
- GAPS: data a buyer would need that isn't in the DATA section.

WooMate Studio works the same way. It takes specs only from the exactly matching model in the catalog, and when it can't find it, it shows a message instead of a description. In the Defro batch I got that message for the Klimosz UNIPELET S boiler: the Defro catalog didn't include that model, so no description was written. Every finished description goes to the “Changes” view, where you see it before it's saved in WooCommerce. More about the module on the Studio page, and the step-by-step for one product in the article Single product description.

How do you prepare a batch of descriptions from a PDF catalog for many models?

You prepare a batch of descriptions in five steps: one catalog as the source of specs, a product list with exact model names, a specs spreadsheet, a test on one product, and then the whole batch with a review before saving.

1. Pick one catalogThe manufacturer's current catalog as a text PDF, or a technical data sheet, with its publication date. Run a scan without a text layer through text recognition (OCR) first. One manufacturer, one catalog.
2. Prepare the product listSKU and model name exactly as in the catalog. “Defro Komfort Eko Mini” and “Komfort Eko” are two different boilers. Remove duplicates.
3. Extract specs into a spreadsheetAttribute, value, unit and catalog page for each model. Flag gaps as NOT IN CATALOG. You'll find the pattern in the CSV file above.
4. Test on one productCheck the structure, length, tone and every number in the table. You correct the instructions now, not after a hundred descriptions.
5. Run the batch and review the resultsIn each description, compare output, dimensions and capacity with the right catalog column. Send only the descriptions you've checked to the store.

The most common mistake from practice: specs from another model in the same series. A boiler catalog usually has one table for the whole series, with columns for 10, 16, 20 and 25 kW. The description for the 16 kW version then gets the dimensions or hopper capacity of the 25 kW version, because it's the next column. You can't see it at a glance, because the numbers look credible. The customer who measures their boiler room sees it. That's why, for a batch from a catalog, I check three numbers in every description: output, dimensions and capacity, against the right column.

Edge cases worth settling before the batch:

• The model is in the catalog under a different trade name. Write both names and the page in the spreadsheet. Don't count on the AI connecting the names by itself.
• The catalog has two editions with different data. Take the newer one and write the publication date in the source column.
• A variable product with output variations. One description for the product, with a table that has a column for each variation, or specs in the variations. More on variation changes in the post Bulk price change in WooCommerce.
• The products aren't in the store yet. Import them from a file first, then write descriptions. I cover choosing an import method in the post WP All Import and its alternatives.

In the Termo Expert store I prepared 777 descriptions this way, in 21 batches. Most were for the Atmos (273), Defro (234) and KHT (130) brands. One of those batches, from April 29, 2026, was 52 Defro products: pellet and coal boilers, heat recovery ventilators, heat pumps, gas boilers. The steps in the app are covered in the article Batch descriptions from a catalog.

How much does it cost to prepare product descriptions?

When you write by hand, the cost is time: yours or a copywriter's. With AI you pay the model provider for tokens, and the cost depends on the model and the length of the description. In WooMate you pay the AI provider with your own key (Anthropic, OpenAI or Google), you see the estimated batch cost before you start it, and you set the monthly spending limit yourself.

The biggest item in a batch isn't tokens; it's human time. The calculation below is an example with assumed times; run it with your own.

Two things really lower the AI bill for batches:

• A background batch costs half the price of regular generation. AI providers handle tasks that don't need to be ready right away more cheaply. A batch takes up to 24 hours, but you can close the tab and get back to work.
• A smaller model for repetitive descriptions. I prepared the Defro batch on Claude Haiku 4.5, the cheapest Claude model available in WooMate. For descriptions with a table from a catalog, it's enough, because the model doesn't have to invent anything, only put the data into words.

One batch in WooMate has at most 100 products. With a bigger catalog, I split the list by manufacturer or by category, which makes the review easier anyway.

What should you check in a description before publishing?

Before publishing, you check numbers, promises and data consistency, not style. The list below remembers what you check, so you can use it for every batch.