← Knowledge

I use AI

AI helps you learn and work faster, as long as you know where it gets things wrong. Here's how to get started, check answers, use AI at work and protect yourself from fraud.

Jump to the questions

Questions and answers

Getting started

From the pillar Artificial intelligence

You don't need any prior knowledge or a subscription to get started. You'll learn the most by using AI for something you have to get done anyway.

What can I use AI for in everyday life?

For almost anything you'd otherwise spend a long time writing, researching or thinking through. Have it explain an official letter in plain words, draft a difficult email, plan a trip or prepare you for an important conversation. The more concrete your reason, the more useful the answer.

Which tool should I use?

For everyday use, the well-known chatbots such as ChatGPT, Gemini, Copilot or Claude are enough. They're constantly being developed, and which one is ahead keeps changing. Pick one and stick with it for a while, because you learn more when you know a tool well.

Do I need the paid version?

Not at the start. It can be worth it once you hit the limits of the free version, for example on the number of queries or with large files, or when you use AI every day at work.

How do I ask so that the answer is useful?

Describe your task the way you'd explain it to a new colleague: what you need, what it's for and in what form. If the answer doesn't fit, tell the AI what's missing instead of starting over. Good results usually only come at the second or third attempt.

What's better left out?

Passwords, bank details, ID numbers, health data and other people's data. With many free versions, what you type may be used for training. Whether that's the case and whether you can switch it off depends on the provider and your account settings.

What you can do now

  1. Pick a task you have to do this week anyway and let an AI help you with it.
  2. If the first answer doesn't fit, tell the AI what's missing and ask follow-up questions.
  3. Check in the settings whether your inputs are used for training, and switch this off.

Learning with AI

From the pillar Artificial intelligence

With AI, you can get to grips with a new topic in minutes, whether it's a tax assessment, a tenancy agreement or a technical term at work. Whether the knowledge sticks depends on how you use it.

How do I learn a new topic with AI?

First, have the topic explained in plain words and keep asking until you can put it in your own words. Then ask the AI to quiz you and explain what you got wrong.

Do I really learn anything if the AI gives me the solution?

Often not. In a large study, pupils who had practised with a standard AI chatbot did worse than the comparison group in the exam without AI. Those who practised with an AI that only gave hints didn't have this disadvantage. So ask the AI for hints and questions instead of the finished solution.

Can I practise conversations with AI?

Yes, this is one of its real strengths. Let the AI play your manager, a client or a panel and practise your salary negotiation, your presentation or your answers to difficult questions as often as you like. What to watch out for in its feedback is covered under Presence.

What you can do now

  1. Have the AI explain a topic that's on your mind right now, and keep asking until you can explain it yourself.
  2. Ask the AI to quiz you and give you hints instead of solutions.
  3. Role-play a conversation you have coming up soon.

Checking answers

From the pillar Artificial intelligence

AI answers always sound convincing, even when they're wrong. A few simple steps help you tell what you can rely on.

Why does AI make things up?

A language model calculates which word is likely to come next. It doesn't know whether something is true, so a wrong answer sounds just as confident as a right one. New models make such mistakes too, especially with numbers, dates, people and specialist questions.

Can I rely on the source the AI gives me?

Not without checking. Tests have shown again and again that AI search tools attribute sources wrongly or make up links. So open the source yourself and read whether it really says what the AI claims.

How does AI choose its sources?

It summarises what it can easily find online. So what comes first is often what's most visible, which isn't always the most expert, and sometimes it's advertising. For law, taxes, money and health, it's worth looking at the websites of public authorities, consumer advice centres and experts.

How can I tell whether an answer is up to date?

Ask the AI how up to date its knowledge is and whether it searched the web for the answer. Laws, prices and deadlines change all the time, and a model only knows what was in its training data or its latest search.

What you can do now

  1. For every important answer, open the source given and read the passage yourself.
  2. Look at who's behind the source: a public authority, a specialist publication, a forum or an advertising site.
  3. Ask how up to date the answer is, and take extra care to check numbers, dates and statements about people.

AI at work

From the pillar Artificial intelligence

At work, AI saves the most time with texts, research, summaries and analyses. What matters is which tools and which data are allowed.

Can I just use AI at work?

That's up to your employer. If there are rules, they apply. If there aren't any, ask before you enter company data, and suggest drawing some up together.

What's shadow AI, and is it a problem?

Shadow AI means using tools your employer hasn't provided, such as your private account. Almost one in three people who use AI at work do this (German Economic Institute, 2026). It becomes a problem when customer data, staff data or trade secrets end up in a tool that nobody in the company has checked.

What data doesn't belong in an AI tool?

Other people's names, contact details and health data, confidential company information and anything covered by a duty of confidentiality. Your company may only enter personal data if it has a legal basis for doing so and, if the provider processes the data on its behalf, has signed a data processing agreement with it.

Does my employer have to train me?

Under the EU AI Act, companies that use AI have to take measures to support their staff's AI literacy. Since the end of July 2026, companies no longer have to ensure a specific level of AI literacy for each employee. It's still worth asking for training, because many companies don't offer any yet.

What you can do now

  1. Ask your manager or IT whether there are rules and approved tools.
  2. Don't enter customer or staff data or confidential information into private AI accounts.
  3. Ask for AI training.

Pay and job applications

From the pillar Artificial intelligence

AI learns from what people have written online, and in the process it picks up old gender roles too. When it comes to pay and job applications, that can cost you real money.

Does AI give women different advice than men?

Studies have measured this several times. Researchers at THWS Würzburg-Schweinfurt asked five widely used language models what salary a person should ask for in a pay negotiation. The profiles were identical, only the gender was different. Women got markedly lower recommendations, most of all in medicine and law. In one test, ChatGPT advised a man to ask for 400,000 dollars and a woman for 280,000 dollars (THWS, 2025). A study by UC San Diego found a smaller gap of around 1 % with ChatGPT, but in the same direction (PLOS ONE, 2025).

Does this affect job applications too?

The same pattern shows up in applications. The University of Washington had three AI models rank more than 550 real CVs, swapping only the names. CVs with men's names came out ahead in 52 % of cases, those with women's names in 11 %. This held even for jobs mostly done by women, such as HR. According to the researchers, leaving out the name isn't enough, because AI can also infer gender from word choice or memberships (University of Washington, 2024).

Where does the bias come from?

Above all, the bias lies in the data AI learns from. For ChatGPT-3, researchers estimated that only around 26 % of the training texts were written by women (University of Pittsburgh, 2023). Wikipedia is an important source: in the German edition, 17.9 % of biographies are about women (Wikipedia statistics, 2026). Then there's the question of who builds AI. Worldwide, only about one in five AI professionals is a woman, and in senior AI leadership it's fewer than one in seven (UN Women, 2026).

Does it help to leave my gender out of the prompt?

You can't rely on it. Many AI tools remember details from earlier chats, and the THWS researchers warn that the bias can still find its way into the answer like this. It's safer to check every number the AI gives you against real pay data.

Where can I find reliable pay data?

The most reliable free source is the Entgeltatlas, the pay atlas of Germany's Federal Employment Agency. It shows the median pay for your occupation, broken down by region, age and gender, and is based on what employers report to the social insurance system (Federal Employment Agency). If you work in an establishment with more than 200 employees, the Pay Transparency Act already lets you ask in writing what male colleagues in comparable jobs earn on average. This works if there are at least six of them in the comparison group (Sections 10 to 12 EntgTranspG). The new law, expected in 2027, is meant to extend this to all establishments.

What you can do now

  1. Check every pay figure an AI gives you against the Entgeltatlas, and look at the figure for men there too.
  2. At work, use your right to information under the Pay Transparency Act. It currently applies in establishments with more than 200 employees. The new law, expected in 2027, is meant to extend it to everyone.
  3. Never let an AI decide on job applications on its own, whether you're the applicant or the employer.

How to negotiate your pay is covered under Presence.

Your job is changing

From the pillar Artificial intelligence

For most people, AI changes individual tasks, not the whole profession. If you get involved early, you'll help decide which ones.

Will AI replace my job?

Usually not as a whole. What the labour market shows above all is a shift in tasks: routine work gets done faster, while judgement, relationships and responsibility become more important.

Is it worth learning how to work with AI?

Yes. In many sectors, employers in Germany pay considerably more when a job requires AI skills. Above all, they're looking for people who apply AI in their own field: in 2025, for every job ad in AI development, there were about seven for roles that apply AI (PwC, 2026).

What does this mean for me in practice?

Look at which of your tasks are routine and which need judgement, relationships or responsibility. With routine work, AI takes tasks off your hands, and with the rest it helps you be faster and better prepared. AI also creates new tasks, because someone has to plan it, guide it and check it.

What you can do now

  1. Write down your tasks and mark which are routine and which need judgement or relationships.
  2. Try out how much AI can take off your hands with one routine task.
  3. Ask your employer for AI training.

Deepfakes and fraud

From the pillar Artificial intelligence

Deepfakes mostly target women, and the law is changing a lot right now.

What are deepfakes?

Deepfakes are images, videos or voices that are created or altered with AI and look or sound real. Most of them are sexualised fakes, and almost all of these show women.

Is a sexualised deepfake of me a criminal offence?

Your personality rights and the right to your own image can already apply today. But there's no specific criminal offence yet. On 7 October 2026, the federal cabinet approved a law against digital violence under which even creating sexualised deepfakes would become a criminal offence. It now goes to parliament (Bundestag and Bundesrat) and hasn't come into force yet.

How does fraud with a cloned voice work?

Fraudsters use AI to imitate the voice of a family member and fake an emergency on the phone. A few recordings are enough, for example from social media. You can barely tell from the sound that it's fake.

Who helps me if it happens to me?

HateAid advises people affected by digital violence and may cover legal costs. Save evidence with the date and link before the content is deleted, and report it to the police.

What you can do now

  1. Agree on a code word with your family for emergency calls.
  2. If a call seems suspicious, hang up and call back on the number you have for them.
  3. If it happens, save screenshots with the date and link and report it to the police. For deepfakes or online hate, HateAid can help.

What you can do now

Checklist as PDF ↓

This is general guidance, not tax, legal or funding advice. For your situation, talk to a tax adviser and a lawyer. All information applies to Germany.

Last checked on
November 8, 2026

Getting started, Learning, Checking, At work, Pay and applications, Your job is changing, Fraud

New rules on pensions, taxes or work? You’ll know before they apply.

A two-minute read, every fact checked. Plus AI tips to go, a story from another woman’s life and invitations to free webinars with our experts.

Das hat leider nicht geklappt. Bitte versuch es noch einmal.
Fast geschafft. Wir haben dir eine Mail geschickt. Bitte bestätige darin deine Anmeldung.

Du bekommst unseren Newsletter mit Neuigkeiten, Einladungen und ab und zu einer kurzen Umfrage. Abmelden kannst du dich jederzeit über den Link in jeder Mail. Mehr dazu in der Datenschutzerklärung.

Oder folg uns auf Instagram.