AI Chatbots: What They Actually Do, and When an FAQ Page Wins

Few tools get louder marketing in 2026 than the AI chatbot: available around the clock, answers everything, qualifies leads, saves headcount. The truth is more modest, and in many cases more inconvenient: a bot that reliably answers five questions is worth more than one that improvises on the sixth, and plenty of businesses would get further, faster, with a tidy FAQ page than with a chat window.

What an AI chatbot does differently from the old click menu

An old chat system worked with fixed decision trees: every possible question had to be set up in advance, or the reply was just "I did not understand that." A modern AI chatbot understands free-form questions and replies in natural language. Most usable solutions work through retrieval: the bot answers not from general model knowledge, but from a fixed, checked knowledge base the business supplies itself. That distinction is exactly what makes it credible for business use in the first place.

What it reliably delivers - and what it only appears to

Realistically, and without exaggeration, a well-built chatbot delivers this: answering recurring standard questions around the clock, pulling concrete answers from a business's own content instead of generic phrases, routing visitors to the right form or contact, and handing off cleanly to a person when needed.

What it only appears to deliver, once the knowledge base runs thin or the question is unusual: sounding confident without actually knowing anything. A language model writes fluently even when it lacks the grounding for an answer, and that fluency is what many people mistake for competence.

Where the limits sit

• Made-up answers: a freely running language model can state something false with full confidence, especially on prices, deadlines or legal matters
• Complex or emotional cases: complaints, edge cases and negotiations belong with a person, a bot that tries to force these through tends to frustrate rather than help
• Only as good as its knowledge base: if content is outdated or contradictory, the bot reflects exactly that - it does not replace well-maintained information
• Not a set-and-forget tool: a chatbot needs setup, testing and ongoing tuning, one left alone after launch quickly starts giving poor answers

The point where a better FAQ page is the smarter choice

A chatbot earns its cost where many people ask many different, but recurring, questions and a natural-language answer adds real value. It does not earn its cost everywhere, and admitting that honestly saves businesses unnecessary spend.

An FAQ page carries no ongoing upkeep for a language model, no check against hallucination, and no GDPR review for a new system. Where it serves the same purpose, it is simply the cheaper, more robust solution.

Data protection in brief

Once a chatbot talks to real visitors, it processes personal data. Four points need settling before launch: a data processing agreement with the provider, clarity on the model used and its server location, a visible notice in the chat window, and data minimisation in what gets stored. The details on provider types and selection criteria are covered in AI Chatbot: Build or Buy.

Concrete steps for the next 90 days

• Days 1-30: collect the five to ten most common questions and check whether an FAQ page would already cover them completely
• Days 31-60: if a chatbot is genuinely justified, structure the knowledge base and pick a provider with a data processing agreement and a suitable server location
• Days 61-90: launch the bot or the FAQ page, review real questions and refine the knowledge base or the page accordingly

Conclusion

Keep a tally for one week of the questions that actually come in. If it is ten of the same ones, they belong on an FAQ page, and that is finished in an afternoon. If it is sixty different ones that somebody answers by searching the same three documents every time, a bot carries its cost - but only if you can name, at the same time, the person who will maintain that knowledge base permanently. How the economics work out in detail is covered in AI Chatbots for Websites: Cost-Benefit Check, and the full process from idea to a running bot in AI Chatbot: Build or Buy. A broader look at automatable processes is in AI in Business: Twelve Processes.