AI in e-commerce: how an online shop can make money from artificial intelligence and what to watch out for

What is AI in e-commerce?

Artificial intelligence in e-commerce covers all the uses of AI that help a shop sell: generating and translating product descriptions, a chat that answers customers, recommendations, a site search that understands questions asked in the customer's own words, review analysis and, increasingly, the presence of the offer in AI assistants such as ChatGPT. The shop is only one of the areas in which AI is changing how a company works; the others are described in guide to AI in business.

This last change is the newest. Customers ask a language model (LLM, i.e. a program that understands and produces text; hence the term ecommerce LLM): “which trail running shoes under PLN 400?”, and get ready-made suggestions. If your shop is not among the sources the assistant uses, you are not part of that conversation. We describe the mechanism in more detail in the article on ranking in ChatGPT.

How many e-commerce companies use AI? There is no separate data for online retail alone in Poland. According to Eurostat, in 2025 8.4% of all companies in Poland with at least 10 employees used AI, against an EU average of 20%, while on the customer side it is the other way round: 53% of internet users surveyed by Gemius say they use chatbots or AI tools. Customers are moving faster than shops.

Where does AI really make money in an online shop?

Assess AI solutions for e-commerce against two questions. How close are they to revenue? How much risk do they carry? We have collected the most common applications in a table.

Shopping in ChatGPT: Allegro, AI agents and a shop ready for assistants

On 12 May 2026, Allegro launched an app in ChatGPT. The customer describes what they are looking for, the assistant browses the offers and shows product cards (price, photo, ratings), and a “Buy on Allegro” button takes them to the order on the platform. The search terms “ChatGPT Allegro” and “Allegro AI assistant” appeared in search engines precisely after this launch.

The takeaway for sellers is practical. In this model, the choice depends on how well the offer is described: attributes, photos, reviews, availability. The assistant will not read a “unique, top-quality product”, but it will read the size, material, weight and delivery time. The next step is shops with AI agents – programs that not only advise but place the order on the customer's behalf; we explain how such agents work in the article on agentic AI and AI agents.

How do you optimise a shop for AI? Five things make the biggest difference. Complete attributes in the product data. Structured data on product pages, i.e. a description in the page code that machines can read without guessing. An up-to-date product feed for price comparison sites and platforms. Genuine customer reviews. And finally, mentions of the shop in independent media and guides from which models draw their knowledge.

Who is AI in e-commerce for, and when does it pay off?

Shops with a large catalogue (hundreds or thousands of products) gain the most, because writing and correcting descriptions by hand simply does not scale. The second group is shops selling in several markets, where AI-assisted translation checked by a human is much cheaper than translating from scratch. We describe how to do this without slip-ups in the article on AI translation in business.

A separate case is B2B e-commerce with AI, i.e. ordering platforms for businesses. There, AI helps customers find the right product in an extensive technical catalogue, prepare an order based on previous ones and answer questions about specifications. The time saved for sales reps can be greater here than in retail, and if there is a sales team behind the B2B shop, connect it to the system we describe in the article on CRM with AI.

A small shop with a few dozen products and a small number of enquiries will gain the least. There, start with the tools built into the e-commerce platform rather than with separate implementations.

How to prepare an online shop for AI implementation step by step

The question “how to use AI in e-commerce” usually ends with buying a tool, but it is better to start from the other end – the data the tool will work with, because that is what decides whether the implementation has any chance of success. Five steps for the first quarter:

1. STEP 01Review your product data

   How many products have complete attributes, consistent names, good photos and unique descriptions? Gaps here undermine everything downstream: the chat, the site search, recommendations, visibility in AI assistants.

2. STEP 02Choose one area to start with

   Most often product descriptions in one category, or a chat answering questions about delivery and returns. One area, a clear goal, a measurable result.

3. STEP 03Measure your starting point

   Conversion in the chosen category, the number of customer service queries, the basket abandonment rate, the time needed to prepare a description. Without this, you will have nothing to compare with after implementation.

4. STEP 04Set control rules

   Who checks descriptions before publication? What promises must the chat not make (discounts, delivery dates, exceptions to the terms and conditions)? When is the conversation handed over to a human?

5. STEP 05Compare with a control group

   Roll out changes in some categories and compare them with the rest or with the same weeks of the previous year. Seasonality and promotions can have a bigger effect than AI itself.

How much does AI in e-commerce cost and what drives the budget?

Some tools are already included in the price of the e-commerce platform, some are billed per use, but the biggest cost is usually the work on product data, which does not appear in any price list because it is done by your team or by someone you outsource it to. Prices below are as of September 2026.

Can I use AI in my shop and what does the law say?

Yes, you can use AI in e-commerce, but with a few obligations. Legal position as of 24 September 2026:

• Customer chat. From 2 August 2026, Article 50 of the AI Act requires that customers know they are talking to a program, unless this is obvious.
• AI-generated content. The AI Act provides for labelling certain synthetic content, especially realistic images and recordings. We explain what this means for marketing in our piece on labelling AI content from 2 August 2026.
• Description matching the product. The seller is responsible for the description and photo, regardless of who prepared them. A generated attribute the product does not have is a straight road to a complaint.
• Chat replies. The shop is responsible for what its chatbot has promised. We describe how to set the boundaries of such a conversation in the article on AI in customer service and chatbots.
• Personal data. Chat conversations and data for recommendations are customer data, so you need a data processing agreement with the vendor and information in your privacy policy.

When should you implement AI in your shop, and when is it better to wait?

The best moment is two or three months before the peak season, which for most shops means before the pre-Christmas period. There is then time for testing and fixes, followed by calm measurement, so the effect arrives just when traffic is at its highest. Rolling out a chat in the week of a big promotion is asking for trouble.

It is better to wait while the e-commerce platform is being migrated, when product data is in disarray or when nobody has time to review content. In that case, tidy up the catalogue first. Also look at the whole customer journey rather than just one element; for help with this, see our earlier article on how to design the purchasing process.

The future of AI in e-commerce and what will change for shops

How will AI affect the future of e-commerce? Two directions are certain because they are already visible. First, some customers start shopping not in a search engine or on a marketplace but in a conversation with an AI assistant – the Polish example being the Allegro app in ChatGPT. Second, AI agents are emerging that compare offers and place orders themselves. In both cases, the winner is the shop whose data is complete and trustworthy.

The rest is, for now, forecasts. A sensible strategy is not about rebuilding the shop around one vision of the future, but about doing things that help regardless of how fast things change: tidy data, good descriptions, genuine reviews, independent mentions. Can AI build an e-commerce shop? It will help with the template, descriptions and graphics, but decisions on the product range and prices are made by people, as are decisions on how to serve customers.

Want to check how your shop performs in AI answers and what to do to make it appear more often? Visit the page AI search optimisation. We write about how ChatGPT is also becoming an advertising channel in the article on ads in ChatGPT.

Mistakes that cost a shop customers and margin

Read next