How to connect ChatGPT to your website, e-shop or system
What connecting to ChatGPT means, the three routes to it, what you need for each and roughly what it costs — explained without programmer-speak.
Updated: August 2026
What connecting ChatGPT means
Connecting ChatGPT means linking your website, e-shop or company system to the artificial intelligence from OpenAI, so that your system sends it questions and receives answers on its own — with nobody at the keyboard. It is done through the OpenAI API, and three routes lead there: no code at all, ready-made plugins, or a custom integration.
Think of an API as a socket in the wall. OpenAI has built a ready-made “brain”, and the API is the socket through which you plug it into your own program — just as you plug a washing machine into the mains without building a power station.
In practice it looks like this: a customer types a question on your website, your system sends it to ChatGPT, ChatGPT returns the answer, and the customer sees it on your site — under your logo and by your rules.
ChatGPT vs. the OpenAI API — the difference in one sentence
ChatGPT is the page where you talk to the artificial intelligence in person. The OpenAI API is the same thing for programs — through it, your system talks to the AI, automatically and around the clock.
In practical terms: when AI is meant to answer in a window on your website, sort your e-mails or write product descriptions in your e-shop, you need the API. The chatgpt.com page is not enough for that — it only works while somebody is sitting at it.
3 ways to connect ChatGPT
There are three routes, and they differ in price, in effort and in what they can do. The prices are typical market ranges in Slovakia, not our price list.
| Route | What it means | Typical price | What it suits |
|---|---|---|---|
| No-code (Make, Zapier) | Make and Zapier are services where you assemble the connection by clicking — without writing code. | on the order of hundreds of € for setup | Simple automations: AI summarises an order, sorts messages, drafts a reply. |
| Ready-made plugins | A pre-built component that clicks into an existing website or e-shop. | depends on the plugin, usually the cheapest start | Ordinary websites and e-shops on widespread platforms a plugin exists for. |
| Custom integration | A developer connects ChatGPT directly to your system and your data, by your rules. | from the low thousands of € upwards | When AI should answer customers, draw on your data or take several steps at once. |
What you will need
Whichever of the three routes you choose, you always need these three things:
- An API key — a long password your system uses to identify itself to OpenAI. You create it in your OpenAI account, and usage is billed to it.
- Rules for what the AI may say — written guardrails: what it answers, what it stays silent on and when it hands over to a person. Without them, the AI will happily answer things it has no business answering.
- Someone to wire it up — for the click-together route a capable colleague is enough; for a custom integration, a developer or us.
What the integration and the running cost
You pay for two different things, and it pays not to mix them up: the one-off integration and the monthly running costs.
Running costs: OpenAI bills for so-called tokens — pieces of words. Put simply, you pay for the amount of text the AI reads and writes, much as you pay for electricity by consumption. For a small company this typically comes to a few euros to a few tens of euros per month.
The integration: the click-together route via Make or Zapier comes to hundreds of euros; a custom integration into your own system starts in the low thousands. As before — these are typical market figures in Slovakia; the exact price is set by scope.
“Am I not sending them my company data?” An honest answer
The most common worry — and the right question to ask. The good news: your integration decides what gets sent. It is set up so the AI receives only what it truly needs to answer — the text of the question, for example, but not the customer's personal identification number or payment details.
On top of that, a data processing agreement can be concluded with OpenAI, and data sent through the API is not used to train their models by default. In a proper integration this is addressed at the start of the design, not as an afterthought.
When to hand it to professionals
An internal automation that summarises your e-mails or drafts replies — assemble that yourself by all means. When it slips up, you are the only one who sees it.
Call in professionals the moment the AI is to speak with customers or reach into your data. There, a mistake hurts: a badly configured AI can promise a discount that does not exist, or reveal something it should not have.
At Grew Studio we have 60+ projects across 24 industries behind us, and we build AI into our own products too: our voice assistant Olivia answers the phones for more than 50 medical practices today and handles 8 calls at once. Exactly the kind of integration where the rules and the security cannot be an afterthought.
Frequently asked questions
Do I need a programmer to connect ChatGPT?
Not always. Simple automations — say, an AI that summarises an order and e-mails it to you — can be assembled with tools like Make or Zapier without any programming. You need a programmer, or us, once the AI is to answer customers, draw on your data or take several steps at once — then the rules and the security must be set up properly.
How much does running ChatGPT cost for a small company?
As a guide, a few euros to a few tens of euros per month. OpenAI bills for the tokens consumed — the pieces of words the AI reads and writes — so you pay by actual usage, much like electricity. You can watch the consumption in your OpenAI account as you go and set a monthly cap on it, so the bill can never catch you off guard.
Will the AI answer from our company materials?
Yes, that is exactly what a custom integration is for. The AI is connected to your price lists, manuals or internal documentation and answers from them, not from the whole internet. Put simply: alongside the customer's question it receives your materials attached, and it sticks to them. The quality of the answers then stands or falls on how well that connection is built.
Will the AI make things up?
Without rules, occasionally yes — which is why the rules are half of the whole job. The risk is tamed by having the AI answer only from your materials, say it does not know when a question falls outside them, and hand sensitive situations to a person. No honest supplier will promise you a 100% guarantee; what can be achieved is a state where the AI makes fewer mistakes than a tired person at the end of a shift.
What happens to the data we send through the API?
Data sent through the OpenAI API is not used to train their models by default, and a data processing agreement can be concluded with OpenAI. The main thing, though, is that your integration decides the content — it is set up so the AI receives only the necessary minimum, with no personal identification numbers or payment details. This is the first thing addressed in the design.
How long does connecting ChatGPT take?
A simple automation via Make or Zapier is usually done in a few days. A custom integration — with rules, testing and a connection to your data — typically takes several weeks depending on scope. You receive an exact schedule together with the fixed quote within 24 hours of the intro call, so you know in advance when each part will be ready.
Want ChatGPT inside your system?
Book a free call with us. Tell us what the AI should do, and within 24 hours you receive a fixed quote with a precise scope and an estimate of the monthly running costs.
You might also like
Artificial intelligence for business
What AI can genuinely handle in a Slovak company, what it costs and how to start without risk. Explained without jargon.
Learn MoreAI chatbot for your website
How a chatbot that understands questions works, what it costs and when an off-the-shelf tool for a few euros a month is enough.
Learn More