ChatGPT for your website
6 min read · updated August 27, 2026
Searching for ChatGPT for your website usually means one of two things: putting a chat box on your site, or getting a model to answer questions about your business. Those need different things, and the most popular option — a Custom GPT — does not do the first one at all.
The Custom GPT misunderstanding
A Custom GPT lives inside ChatGPT. You can upload files to it, give it instructions, and share a link. What you cannot do is put it on your own website — visitors would need a ChatGPT account and would be leaving your site to use it.
That distinction catches people out because building one feels like building a website assistant. It is closer to building a saved prompt with attachments. Useful for your own team, invisible to your customers.
What the model knows about your site by default
Nothing reliable. A general model has absorbed a large amount of public text during training, which may include pages from your site if it was crawled before the training cutoff. That is not the same as knowing your current prices, stock or policy.
The failure mode is specific and worth naming: asked about a policy it has not seen, a model will usually produce a plausible one rather than decline. A confidently wrong refund window is worse than no answer, because the customer acts on it.
The three paths that actually work
Each of these puts real answering on your own site. They differ mostly in how much engineering you own.
- Build it yourself on the API: crawl your pages, chunk them, embed them, store the vectors, retrieve per question, prompt with the retrieved passages, and build the widget. Full control, and you own crawling, re-crawling, retrieval quality, streaming, rate limiting and abuse handling forever.
- Use a retrieval product plus your own front end: a hosted vector store handles indexing and search; you still build the widget, the conversation handling and the re-crawl schedule.
- Use a hosted answering layer: point it at your URL and embed a script tag. You give up control of the retrieval internals in exchange for not maintaining any of it.
What to check before you commit to any of them
Ask what happens when the answer is not on the site. A system that always answers is a system that will invent one. The correct behaviour is to say it does not know and hand over.
Ask whether answers cite the page they came from. Citations are the only practical way for you, or the customer, to check an answer without knowing the content by heart.
Ask how often the content is re-read. A site that changes weekly and an index that refreshes monthly will disagree, and the customer sees the disagreement before you do.
Questions
- Can I embed a Custom GPT on my website?
- Not as a chat widget for visitors. A Custom GPT is shared as a link into ChatGPT, where the person needs their own account. Putting answering on your own site means the API or a hosted product built for it.
- Will ChatGPT know my prices if my site is public?
- Not dependably. Training data has a cutoff and no guarantee your pages were included. Anything that changes — pricing, stock, hours, policy — needs to be retrieved at answer time rather than recalled.
- Is the API approach cheaper than a hosted tool?
- The model calls usually are. The total rarely is, once crawling, embedding storage, retrieval tuning, the widget and ongoing maintenance are counted as work someone has to do.
Related
tool
AI chatbot generator
Generates the configuration an AI chatbot needs — persona, tone, opening line, starter questions and the topics it can cover — derived from a real website rather than a template.
tool
Website chatbot generator
Reads your website and generates a complete chatbot configuration — name, greeting, suggested questions, and sample answers with the pages they came from — so you can see exactly how a chatbot would sound before installing anything.
tool
Chatbot ROI calculator
Estimates the support cost an AI chatbot would remove, based on your ticket volume, handling time and loaded hourly cost — plus payback period against a monthly subscription.
term
What is RAG (retrieval-augmented generation)?
RAG is what separates a chatbot that knows your business from one that guesses. Documents are split into passages and indexed by meaning; each question retrieves the closest passages, and the model is instructed to answer only from them.
term
What is a RAG chatbot?
The distinction is where the answer comes from. A general chatbot answers from whatever the model absorbed in training. A RAG chatbot looks something up first — in your pages, your documents, your help centre — and answers from what it found, which is why it can cite a source.
term
What is an AI hallucination?
Language models predict plausible text, not true text. Asked a question with no supporting material, a model does not fall silent; it produces the most likely-sounding answer, which is often a wrong price, a policy that does not exist, or a citation to a page that was never written. The behaviour is not a bug to be patched out of the model. It is what the model does when nothing constrains it.
Want to skip to the working version?
RubyRep does everything in this guide from a single URL — reads your site, builds the knowledge, and gives you one line of code to install.
Train My RubyRep