For online shops
A chatbot for online shops
Most questions in a shop come before the purchase: how much is delivery, when will it arrive, does it fit my model, can I send it back. Anyone who cannot find a quick answer often does not buy from you, but from the next result.
A chatbot answers these questions around the clock from the knowledge you provide. From the Medium plan upwards it also knows your product catalogue: it recommends from your real products, shows them with picture, price and link, and asks a follow-up question instead of guessing. This page shows what that means day to day, and says just as clearly what it cannot do without a link to your shop system.
The questions every shop hears again and again
The answers are almost always somewhere on the site already, in the delivery information, in the terms and conditions, in the product description. It is just that nobody looks there while a chat window is closer than the footer.
- How much is delivery, and when is it free?
- When will my order arrive?
- Can I send the item back, and who pays for that?
- Does it fit my device, my vehicle, my size?
- Which of these twenty similar products should I choose?
- How can I pay?
- Do you also deliver to Ireland or mainland Europe?
We enter all of these answers in the customer portal once, on every plan. The bot answers from them, and only from them. If it finds nothing for a question, it says so plainly instead of guessing a delivery time.
It knows your range, not just your text
From the Medium plan upwards you enter one address in the customer portal: your Google Shopping feed, the WooCommerce Store API or the product list of your Shopify shop. No password, no login details, no extension to install. We recognise the format ourselves, fetch the products straight away and then again every night.
When a visitor asks about a product, our server searches your catalogue beforehand and passes the matches along. The bot recommends from those, and the visitor sees up to three cards with picture, price, tax and delivery note, availability and a link to the product page.
- It knows more than name and price. For each product it reads attributes from your catalogue, such as size, colour, material or power output, and dimensions from the name and description: “95 x 50 x 90 cm”, “Ø 57 cm”, “1.65 kg”, “8 kW”. Ask for a “fire bowl up to 50 cm” and it finds the 50 cm one, not just any.
- It knows where the visitor is. Someone asking “does this fit my model?” on a product page means that product. The bot is given the page along with the question and does not have to ask which one was meant.
- It compares two products when someone is torn between them, and names the difference that matters for the decision.
- It shows your cancellation policy word for word, in a box of its own. It never phrases that text itself: a wrong deadline would bind you.
It asks before it guesses
“I'm looking for a barbecue.” A bot that shows three random barbecues out of two hundred has not given advice, it has rolled the dice. That was exactly the point where we made improvements.
For every search, our server works out how many products match about as well as the best one, and what distinguishes them. If there are seven or more and the question contained nothing to narrow it down, the bot asks one short question about the feature that splits the selection most: fuel, diameter, material. Only then does it recommend.
One question, not three. If the answer is “I don't mind”, it stops asking and recommends. If someone mentions a size, a price or a model of their own accord, it does not ask at all. Just as a shop assistant asks once and then picks something off the shelf.
What the bot cannot do in a shop
The bot is not connected to your shop system. It sees no orders, no stock levels and no customer accounts. So when asked “Where is my order?”, it cannot give a tracking number; instead it takes down the request with contact details and files it in your portal.
We say this here because it is the point where this page differs from some providers' advertising. A bot that answers from the rules you have set is built for the questions before the purchase. Looking into an individual order remains a job for your customer service, and the bot sorts the requests for them in advance.
On the Small plan the bot answers in English and German. From Medium upwards it answers in each visitor's own language, and you set the chat window itself, meaning its buttons and notices, from 20 languages: 5 of them on Medium, 10 on Big. For a shop with customers abroad, that is the difference between an answered question and an abandoned purchase.
See it on a real shop example
The quickest way to see it is our example website Werkhof (in German), a made-up tool shop with 120 products. Ask the bot there for a cordless drill for drywall or a ladder for a working height of three metres. It gives reasons for its recommendation, rules out the wrong tool and shows the products as cards underneath, with picture, price and link. No sign-up needed.
The demo of our customer portal (in German) shows the same thing from the other side: what the bot makes of a conversation, and how you find it in the portal. Your portal would look exactly like that, just with your own rules.
There you can also see what comes out of a conversation: the question answered, the enquiry taken down with contact details, and the list of questions it had no answer to. In a shop that list is worth real money, because it shows which detail is missing from your product pages.
Accessible, and why that is no side issue in a shop
In Germany, the Accessibility Strengthening Act (BFSG) has applied since June 2025. Anyone there who runs an online shop for consumers must keep their website accessible. Micro-enterprises are exempt, meaning businesses with fewer than ten employees and an annual turnover or balance sheet total of no more than two million euros. Everything that runs on the site is covered, including a chat window from a provider like us. Breaches can bring fines of up to €100,000. This is German law; we do not assess here what applies to a shop in the UK.
The act does not apply to us, as we are too small. It does apply to many of our customers, and then our window is not a matter of taste but the condition for being allowed to use it at all.
- The chat window is measured against WCAG 2.1 level AA, with calculated contrast ratios rather than estimated ones.
- Tested on 15 September 2026 with the NVDA screen reader on Windows: finding the button, opening it with the keyboard, asking a question, having the answer read out, hearing who is speaking, closing it and getting the focus back.
- This turned up three faults, all three of which an automated testing tool had missed. All three were fixed the same day and checked again by ear.
- Zoom to 200 per cent, text spacing and operation without a mouse were measured on a device, not inferred from the code.
Adding it to your shop system
The bot goes into your existing site as one line of code; your shop system is left untouched. For Shopify and Shopware, videos on the installation page show every click; a WooCommerce shop follows the WordPress route shown there. If it does not work first time, we do it together with you online.
Delivery terms, returns rules and your main categories: we import those from your shop when we set the bot up, on every plan, and you only need to correct them. Later you can re-import changes there yourself, new delivery rules before the Christmas rush, for example.
After that it runs free of charge for 14 days. With your own content, on your own website. No payment details, no automatic renewal. If you decide not to carry on, simply say nothing.
Frequently asked questions from shop owners
- Can the bot tell customers the status of an order?
- No. It is not connected to your shop system and sees no order data. It takes down the request with contact details and passes it on to you, so that nobody is left sitting in front of a silent window.
- Where does the bot get its answers from?
- Solely from the knowledge you provide in the customer portal, such as delivery costs, deadlines and product details. If it isn’t there, the bot says so instead of guessing.
- Does the bot answer in other languages too?
- On the Small plan the bot answers in English and German. From Medium upwards it answers in each visitor's own language, and you set the chat window itself, meaning its buttons and notices, from 20 languages: 5 of them on Medium, 10 on Big.
- What happens with a complaint?
- The bot passes the conversation on to you, tells the visitor it has done so and files the case in the customer portal. That applies on every plan.
Related
See the bot with your own product range.
Tell us what you sell and which questions your customer service answers all the time. We will get back to you within 24 hours, in person.
Enquire without obligationLast checked: 3 October 2026