What a library AI chatbot is, and what it is not
An AI chatbot for libraries is a virtual assistant trained on the catalog, regulations and services of one specific library, not on general internet knowledge. The window looks like any other chat; the difference is where the answer comes from. A generic chatbot such as ChatGPT answers from what it learned in training, so when a student asks whether a book is available, it can reply with something plausible and false. A library chatbot built properly uses retrieval-augmented generation (RAG): before answering, it searches the library's real sources, such as the catalog, service pages, regulations and subject guides, and only then writes the reply on that basis. Our guide to the AI chatbot for libraries explains the mechanics.
| Aspect | Generic chatbot | BiblioLabTech library chatbot |
|---|---|---|
| Source of answers | General training knowledge | Your catalog, regulations and services |
| Book availability | Unknown, or invented with confidence | Checked in the catalog in real time |
| Opening hours and local rules | Unknown unless described in each conversation | Loaded once from official information |
| Identity | Generic | Your library's name, tone and identity |
| Channels | A separate app or site | Your website and WhatsApp |
| Risk of invented answers | High on institution-specific questions | Reduced by answering from verifiable sources |
The questions it actually resolves
- Opening hours, including specific rooms, exam periods and holidays
- Availability of a resource: whether a book is on loan, at which branch, whether a digital copy exists
- Renewals and loans: how to renew, how many times, what happens with a fine
- Personalized recommendations of reading or bibliography by interest or program of study
- Bibliographic guidance: which databases to use for a topic, how to cite in a given style, where to find a digital resource
- Administrative procedures: registering as an external user, accessing subscribed databases from off campus
What it should not attempt is to replace the reference librarian in a complex research consultation or a delicate conversation with a user. The realistic goal, and the one BiblioLabTech designs for, is that the assistant absorbs the repetitive volume so staff can focus on work that requires human judgment. It does not replace the reference team; it frees their hours.
Assistant types
| Assistant | Designed for | Main tasks |
|---|---|---|
| ChatBot Referencista | Libraries | Reference services, reading recommendations, bibliographic guidance, 24/7 support |
| ChatBot de Ventas | Publishers and bookstores | Book search, order management, purchase recommendations |
| ChatBot Estudiantil | Higher education | Academic queries, administrative information, institutional communication |
| Custom chatbots | Any institution | Built around your own processes and sources |
Channels and integrations
The same assistant can live on several channels without duplicating its training: a chat widget on your website that any catalog visitor can open without leaving the page, and WhatsApp, which for many users is the default way to communicate. It can also be integrated into academic portals, virtual classrooms and service desks.
Its real value comes from live data connections. With Koha, it can answer about item availability and loan status in real time; with DSpace, about the theses, articles and datasets in your repository; ABCD and OJS are also supported. Without those integrations, a chatbot can still handle hours, rules and general guidance, but not questions about your live collection. BiblioLabTech implements the chatbot as a layer over what you already have, without asking you to change your library system.
How BiblioLabTech implements it
- Source diagnosis. We review which information exists and how current it is: regulations, hours, tutorials, electronic resource guides, FAQs, catalog and repository. The assistant inherits the gaps of its sources, so this step comes first.
- Conversational design. We define the questions it must resolve, the topics it must never attempt, the tone and level of formality, the institution's own terminology and the rules for escalating to staff.
- Training and integration. The assistant is loaded with the approved content and connected to the catalog and repository for live queries.
- Internal pilot. The reference team tests it with real questions and adjusts answers, tone and handover messages before it opens to the public.
- Launch and monthly upkeep. We track conversation volume, the share of queries resolved without escalation, the most frequent topics and the questions left unanswered. That last list is the monthly work plan for expanding the knowledge base, and it often reveals gaps in the library's public information.
Implementation takes six to ten weeks to public launch, including the pilot. Your institution needs a functional owner in the library who validates content, up-to-date reference material, access to the catalog and repository if live queries are wanted, and a decision on who receives escalated questions. No infrastructure of your own is required.
Data, languages and privacy
The purpose, retention period and access to conversation logs are defined in the contract. When the assistant does not authenticate users, conversations are not linked to identified people; personalized queries, such as loan status, require authentication, and then the institution's data protection policies apply. The assistant replies in the language the user writes in, and for international campuses the set of supported languages and the handling of each one are defined during configuration.
What it costs
| Scenario | Implementation (USD) |
|---|---|
| Basic reference chatbot, one channel (web), no live catalog integration | 800 to 1,800 |
| Web and WhatsApp, integrated with Koha or DSpace | 1,800 to 3,500 |
| Multichannel, several variants and multiple integrations | From 3,500 |
These are reference ranges, not a price list. A recurring monthly cost comes on top: AI model usage, which scales with the number of conversations, plus maintenance, updates and support. WhatsApp adds its own recurring cost for a business account with volume. An assistant whose knowledge base is never updated starts giving outdated answers, the worst possible outcome for a library, so upkeep is part of the service rather than an extra.
Where it runs today
At the Escuela Colombiana de Ingeniería Julio Garavito, the ChatBot Referencista serves the library's reference service as part of a platform that also includes Koha with BiblioScan AI cataloging, a DSpace repository with MetaExtract API and the BiblioMeta metasearch engine, on a dedicated university server administered by BiblioLabTech. Its knowledge base was built from the institution's own material, escalation rules to staff were defined, and it was piloted internally before opening to the public. Read the case study.
Tell us about your library, your channels and your systems and we will scope an assistant built on your real sources.