Common mistakes people make with AI chatbots (and what to do instead)

A lot of complaints about AI chatbots — "it makes things up", "it always says the same thing", "it doesn't get what I mean" — have less to do with the tool than with how it's used. This article covers the most common usage mistakes and what to do instead. It isn't about the model's own failures, such as hallucinations — those are covered in our guide to fact-checking AI answers — but about user habits that make them worse.

#1. Treating it like a search engine

Typing "student laptop" works in Google, but it leaves a chatbot guessing almost everything. You get a generic answer.

Instead: ask a full question that includes your situation and the decision you need to make. "I'm studying architecture and need a laptop for AutoCAD that I can carry to class. What minimum specs should I look for, and which ones don't matter?" The full technique is in our guide to writing better prompts.

#2. Relying on it for current facts

Models are trained on data up to a cut-off date. Some can search the web, but not all of them do so every time or on every plan. Asking about prices, opening hours, recent software versions or news can return outdated information that sounds current.

Instead: for anything that changes, ask where the information comes from, use the tool's search mode if it has one, and confirm on the official website.

#3. Letting one conversation run forever

In very long chats, the model can lose track of early instructions, mix up versions of a document, or carry forward a mistake it made earlier. Every tool has a context limit, and quality drops as you approach it.

Instead: when you change topic or the thread has become muddled, start a new chat. Paste a short summary of what matters ("here are the decisions we've already made…") and continue from there.

#4. Asking for several things at once

"Proofread this, translate it into Spanish, summarise it and suggest a title" usually produces four half-finished jobs.

Instead: one task per message — or a numbered list of steps, asking it to work through them in order and show you each result.

#5. Assuming it knows context you haven't given

The assistant doesn't know your "report" is for a client in Japan, that your company bans certain phrases, or that your tutor requires APA style — unless you say so.

Instead: spend two or three sentences on the situation before asking for the output. It saves several rounds of corrections.

#6. Sharing data you shouldn't

Pasting a contract with real names, medical details, passwords or confidential company information is one of the mistakes with the most serious consequences. Depending on the service and your settings, conversations may be stored and, in some cases, used to improve models. OpenAI and Google both explain their data controls — and how to switch certain uses off — in their help centres.

Instead: anonymise before you paste (change names, amounts and identifying details), check your tool's privacy settings, and read your employer's AI policy. There's more detail in our guide to AI privacy.

#7. Pushing until it agrees with you

Contradict a chatbot firmly enough and it will often concede — even when you're the one who's wrong. This people-pleasing tendency is well known, which means "winning the argument" proves nothing.

Instead: rather than pushing, ask for both sides: "What are the arguments for and against this claim? Which source would settle it?"

#8. Copying and pasting without reading

Text containing phrases like "as a language model…", invented references or unchecked figures is a giveaway of careless use, and in school or at work it can have real consequences.

Instead: read everything you plan to use, rewrite it in your own voice, and check every specific fact.

#9. Ignoring the rules where you study or work

Many universities, schools and employers have their own rules on AI: what's allowed, what must be disclosed and which tools are approved.

Instead: check the policy before using AI on assessed work or with company information. If disclosure is required, disclose.

#10. Thinking there's only one tool

Using a general chatbot to transcribe audio, edit photos or analyse a spreadsheet with thousands of rows can work poorly when there are dedicated tools that do the job better.

Instead: pick the tool for the task. See our guide on which AI tool to use for each task.

#Quick reference

#The bottom line

Most of these mistakes disappear with two habits: give enough context, and never treat the answer as final. Add protecting your data and choosing the right tool for each job, and you'll see a bigger improvement than you would from switching assistants.