Chatbot, Callback or Live Chat: What Brings Inquiries on Your Website

Which inquiry route brings the most inquiries on the website?

Since language models became widespread, many providers have offered website chatbots that are supposed to answer questions around the clock and capture prospects. At the same time there are proven routes: the callback form, in which the visitor gives a number and a time window, and live chat, in which an employee answers in real time. Owners face the question of which of these pays off.

This comparison answers it: Which inquiry route brings the most good inquiries on a small B2B business’s website: chatbot, callback form or live chat? We first clarify the terms and a model from organizational research that explains which channel fits which question. Then we look at what field experiments show about chatbots and which obligations apply. At the end there is a comparison and a decision aid.

The best inquiry route is not the most modern one but the one on which visitors can most easily get their question across.

Chatbot, callback, live chat: what is what?

• Rule-based chatbot: a program that guides visitors through predefined questions and answers, for example via selection buttons. It can only do what was set up in advance.
• AI chatbot: a program based on a language model that understands freely worded questions and generates answers, usually drawing on content from your own website. It is more flexible but can also give wrong answers.
• Callback form: a short form in which the visitor enters name, phone number, concern and a preferred time window. An employee calls back.
• Live chat: a chat window in which an employee answers in real time. If nobody is available, it usually turns into a message form.

How the classic contact form, appointment booking and callback work together is described in the article on the inquiry route on the website. Here the question is whether and when a chat usefully complements these routes.

Which channel fits which question? Media richness theory

Daft and Lengel (1986) developed a model that is still used to choose communication channels: media richness theory. Channels differ in how much they can convey at once: immediate feedback, multiple cues such as voice and tone, natural language and personal focus. A face-to-face conversation is very rich, a form not very.

The core statement: a channel works best when its richness fits the task. For unambiguous questions with a clear answer, such as opening hours or delivery times, a lean channel suffices. For ambiguous questions, where it first has to be clarified what the customer actually needs, a rich channel is necessary. A channel that is too lean leads to misunderstandings; one that is too rich wastes time.

For the website this means: a chatbot is a rather lean channel, even if it phrases things in a friendly way. Live chat is richer because a person can ask follow-up questions. A callback is the richest, because voice and conversation are added. In business-to-business markets, many first inquiries are ambiguous: the prospect knows they have a problem but not exactly which solution fits. The model has limits. It was developed for communication within organizations, not for websites, and it does not take into account that some people prefer a lean channel because it saves them time.

What do field experiments show about chatbots?

The best-known study comes from Luo, Tong, Fang and Qu (2019). In a field experiment, more than 6,200 customers of a financial services company were randomly assigned to receive a highly structured sales call either from an AI voice chatbot or from an employee. As long as customers did not know they were talking to a machine, the chatbot sold as well as experienced employees and four times as well as inexperienced ones. When it was disclosed before the conversation that a chatbot was calling, the purchase rate fell by more than 79.7 percent. Customers considered the disclosed chatbot less knowledgeable and less empathetic and ended the calls sooner.

Gnewuch, Morana, Hinz, Kellner and Maedche (2024) examined a related question on so-called hybrid service agents, in which a chatbot and employees work behind the same interface. A field experiment with 8,966 customers of a multinational telecommunications company and an online experiment showed: when it was disclosed that humans were involved, customers wrote more like they would to a human, that is, at greater length and more politely instead of in short keywords. The chatbot could answer such messages itself less often and had to pass them to employees more frequently. Their workload rose. The authors explain this with customers’ wish to make a good impression in front of a person.

For small businesses there are so far mainly case studies. Marcineková, Sujová and Ďurica (2025) describe the introduction of an AI chatbot at a Slovak micro-enterprise with an online shop. The chatbot reduced staff workload, shortened response times and improved customer satisfaction. This is a single business with consumer customers and rather simple service inquiries.

Two readings of the findings

The first reading sees the studies as a strong argument for chatbots: they can be as good as people and relieve the team. The second sees a problem: the effect depends on customers not recognizing the machine, and exactly that will no longer be permitted. I consider the second reading the more important one for business-to-business companies. Trust is the core of selling there, and an effect that rests on ignorance is not a sound foundation. That does not mean chatbots are useless. It means they should be used for tasks where disclosing them does no harm: unambiguous questions, pre-qualification, routing.

Which obligations apply to chatbots?

The EU AI Act provides for transparency obligations that, according to the regulation’s timetable, apply from 2 August 2026. Anyone offering an AI system that interacts directly with people must ensure that they can tell they are talking to an AI, unless that is obvious anyway. How the obligations affect sales and marketing is described in the article on the EU AI Act.

Data protection comes on top. A chat collects personal data, often including information visitors type in spontaneously. The provider usually processes it on behalf of the business, which requires a data processing agreement and information in the privacy policy. If a chat tool is loaded when the page opens and sets cookies that are not strictly necessary, consent may be required. The basics are described in the article on data protection in sales. This section is not legal advice.

Chatbot, callback form and live chat compared

The following overview compares the three routes by the same criteria. It describes typical properties. Prices and features of individual providers vary widely and should be checked before a decision.

The callback has an advantage that is often overlooked: it moves the ambiguous question into the richest channel and gives the business time to prepare. Its success stands or falls with speed. A callback the next day is already too late for many prospects.

The inquiry switch: which route suits your business?

Model and findings yield a simple decision aid. I call it the inquiry switch. It does not ask which technology is modern but which questions your visitors have.

1. Collect questions

   Review the inquiries of the last three months and classify each one: unambiguous or ambiguous. Also ask sales which questions keep coming up on the phone.

2. Determine the share

   If most questions are unambiguous, a chatbot or a good FAQ section pays off. If most are ambiguous, the callback comes first.

3. Check the staffing

   Live chat only if someone can reliably answer within a few minutes during business hours. Otherwise the chat becomes a slow form.

4. Define the handover

   Every route ends with a person. Decide who takes over an inquiry, within which deadline and how it gets into the CRM.

5. Measure and compare

   Count inquiries with real need per route, not opened chats. After three months, compare with the period before.

Which inquiries then take priority and how to assess them is described in the article on lead qualification. Whether the website’s inquiry rate improves overall you measure as described in the article on measuring the website inquiry rate.

How to get started

• Divide the inquiries of the last three months into unambiguous and ambiguous questions.
• Check how quickly callbacks happen today and set a deadline, for example within four hours during business hours.
• Create a callback form with at most four fields or shorten the existing one.
• Put the most frequent unambiguous questions with answers on the website as an FAQ section.
• If a chatbot is already running: read ten conversations and check whether the answers are correct and the disclosure is clear.
• For every inquiry route, decide who takes over and how the inquiry gets into the CRM.

Where else AI saves time in sales and where its limits lie is compared in the article on AI in sales. The basics of a website that leads to inquiries are summarized in the checklist for a website that brings inquiries.

Conclusion: what brings inquiries on the website?

The guiding question was which inquiry route brings the most good inquiries on a small B2B business’s website. Media richness theory sets the direction: the channel has to fit the question. In business-to-business markets many first inquiries are ambiguous and belong in a conversation. For them, a fast callback form is usually the most effective route. Live chat pays off only if it is reliably staffed.

According to the field experiments, chatbots can be surprisingly good, but their effect depends heavily on whether customers recognize the machine, and disclosure is becoming mandatory. Their place is therefore with unambiguous questions, pre-qualification and handing over to a person. More inquiries come not from the newest tool but from the route on which visitors get their question to the right place without a detour.