Frequently asked questions

Usually under 15 minutes. You add the site address, 4chaty crawls the public pages and builds the initial knowledge base, then you paste a single line of script into the site. No developer and no plugin required.

Yes. The script goes into the theme custom-scripts section, just like any analytics tag. It also works on a site built from scratch — the only requirement is that you can add one line of script to the page.

No. The bot starts from your existing site content, which it indexes automatically. You only fill in what is missing: unpublished rates, internal procedures and the questions you get most often that appear nowhere on the site.

You ask it the real questions your team fields every day and correct whatever it gets wrong by adding an explicit FAQ. Only after that short cycle do you publish the widget. You can also choose which pages it appears on.

Yes. You can bring the question/answer pairs across from Intercom, Tidio, Zendesk or Crisp and load them as FAQs, then top them up with your own site pages.

Yes. Colour, position, avatar, greeting and tone of voice are configured in the panel to match your brand. You can also set how long before the widget opens by itself.

Yes. Romanian is the primary language of the product, not a translation layered over an English bot. The instructions sent to the model, the interface strings and the answers are all built in Romanian, with diacritics. If the visitor writes in English, the bot replies in English.

Before generating an answer, the bot searches your knowledge base for the relevant fragments and is instructed to answer strictly from them. If it cannot find a sufficiently relevant fragment, it says plainly that it does not know and offers an agent.

Yes. Every answer can show its source — the page or document the fragment came from. For the visitor it is a link to the full detail; for you it is the fastest way to spot an old page left in the index.

It says it does not have the information and offers to escalate to a human agent. The conversation reaches the team inbox with the full history attached, so the customer does not repeat what they already wrote.

Yes, through refusal rules. For sensitive subjects — medical, legal, financial, complaints, refunds — the message skips the bot entirely and goes straight to a human.

Almost always the cause is the content, not the model: an old page carrying a price that no longer applies. Delete or exclude the page from indexing, or add an explicit FAQ — the answer corrects itself with no retraining.

Not yet — the WhatsApp and Messenger integration is in the works. The website widget is available today, and once the channels launch they will be different entrances to the same assistant: one knowledge base, the same escalation rules and a shared inbox, with rates updated in a single place.

Through three triggers: no answer grounded in the knowledge base, an explicit request from the visitor, and the sensitive topics you define yourself.

The bot keeps answering from the knowledge base, but on escalation it states transparently when a reply will come and collects contact details. The conversation stays open and is picked up in the morning with full context preserved.

Yes, in full: the questions, the bot answers, the cited sources and the details already collected. They can add internal notes visible only to the team and assign the discussion to a specialist colleague.

Because per-seat pricing punishes exactly what you want to do: put more colleagues on conversations. With 4chaty usage is metered in AI credits based on conversations processed, so you can add the whole support team without the subscription rising per seat.

From the usage report, which shows credits spent per day and per channel with an end-of-month forecast. If the forecast climbs unexpectedly, the usual causes are visible in the same place.

In the European Union, in line with the GDPR. Conversations and the knowledge base never leave the European space. The retention period is configurable, following the storage-limitation principle in art. 5(1)(e) GDPR.

Yes, and 4chaty does so from the first message. It is the requirement of art. 50 of Regulation (EU) 2024/1689 (the AI Act). When a colleague takes over, the change is announced explicitly.

Explicitly, in the conversation, before the data is saved. The moment, the text displayed and the visitor reply are kept as evidence under art. 6(1)(a) GDPR. Consent from a pre-ticked box is not valid.

It is not advisable. Content there can surface in answers to other visitors, which would be an unauthorised disclosure. Personal data is handled separately, by escalating to an agent entitled to consult it.