Verdict: the easiest place to start, and the free plan is enough for all of Level 1. Switch off training before you paste anything, and use your organisation's account, not a personal one, for work material. Move to Plus (€23 a month) only when you hit the free limits several times a week.
Other good fits
Anyone who wants to learn on a free plan before deciding whether a paid assistant is worth it.
+ Pros
ChatGPT is the assistant most of your colleagues already know. One window covers drafting, reading PDFs, web search, images and data files, so tips and help are easy to find. The free plan is enough to learn everything in Level 1.
Pick it when…
You're starting out and want the most widely used assistant, so colleagues can help you.
You want drafting, file reading, web search and images in one place without installing anything.
You want a free plan that already includes search, limited file uploads and limited deep research.
− Cons
Read cons
Answers can be wrong in a confident tone, so check anything that matters. On personal plans (Free, Go, Plus, Pro) your chats are used for training unless you switch it off, and a thumbs-up or thumbs-down can still send that conversation for training. For anything sensitive, use a Temporary chat, or better, don't paste it at all (see T12-L01). Some features arrive later or not at all in the EU, for example public publishing of Sites and Record & Replay, and memory of past chats is off by default in the EEA.
Pick something else when…
Your organisation already provides an assistant. If you work in Microsoft 365, Microsoft 365 Copilot may already be approved under your contract; in Google Workspace, Gemini. Ask IT first; that beats any feature difference.
You work with long documents most of the day. Try Claude side by side on one of your own documents and keep whichever needs fewer corrections.
Data must stay on your own machines. Use Open WebUI with a local model instead.
Avoid if
Your material may not leave your own machines; run a local model behind Open WebUI instead.
You would paste sensitive data into a personal account where training is on by default.
Workflow
State the goal and permitted source material → attach only appropriate files or data → ask for a draft, explanation or comparison → inspect claims, citations and calculations → revise or use the result as a starting point. Human review is still required before relying on output.
GPT-5.6 Luna, web search, limited uploads and deep research
Go
€8 ($8)
More room than Free, same Luna model, may show ads
Plus
€23 ($20)
GPT-5.6 Sol with a Thinking slider, higher limits; monthly billing only
Pro
from €103 ($100; the $200 tier is closed to new sign-ups since 10 September 2026)
GPT-6 Pro in chat, heavy daily use
Business
€21 per user yearly, €26 monthly (Standard seat)
Shared workspace for 2–200 people, no training on your data by default
Stay on Free until you hit the limits several times a week, then move to Plus. Go keeps the smaller model; Pro is rarely needed at Level 1 or 2. There is nothing to install or maintain. The running effort is checking answers and keeping the training switch off on every personal account. API use for your own software is billed separately.
Restrictions
Data controls distinguish training preferences, chat history and temporary chats. Turning off model improvement does not delete saved chats; temporary chats have their own retention and memory behaviour. Managed workspace policies can also limit individual controls.
First project
Documentation-based suggestion — not run here.
Read first project
Turn a messy page of notes into a one-page summary you would send to a colleague.
Switch off training first: Settings → Data controls → Improve the model for everyone → off.
Paste a page of non-sensitive notes, for example from a public seminar or a supplier call with names removed.
Ask: "Summarise this in five bullet points for a colleague who missed the meeting. Mark anything you are unsure about."
Ask a follow-up: "Which of these points came directly from my notes, and which did you add?"
Check every bullet against your notes and fix what's wrong.
Done when the summary fits on one page, every point traces back to your notes, and you corrected at least one thing the assistant got wrong or added. T01-L02 covers this checking habit.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on OpenAI's pricing page (viewed from Germany), help centre articles on Plus, Pro tiers, Business billing, data controls and GPT-5.6, and the ChatGPT release notes, all read on 2026-09-23. Nothing was run in ChatGPT for this review. The EUR price of the Pro $200 tier and whether older EEA restrictions on connectors still apply could not be confirmed.
ChatGPT pricing · Sep 23, 2026 — Plan prices in EUR as shown in Germany; training opt-out on personal plans; feature rows for Free.
What is ChatGPT Plus? · Sep 23, 2026 — Plus is $20 a month; no annual billing.
About ChatGPT Pro tiers · Sep 23, 2026 — Pro $100 and $200 tiers; new $200 sign-ups paused since 2026-09-10.
Data controls FAQ · Sep 23, 2026 — "Improve the model for everyone" switch, Temporary chat, feedback still used for training.
GPT-5.6 and GPT-6 Pro in ChatGPT · Sep 23, 2026 — Default models per plan (Luna on Free/Go, Sol on Plus and up, GPT-6 Pro on Pro/Business).
ChatGPT release notes · Sep 23, 2026 — EU/EEA feature gaps and model release dates.
Further review notes & alternatives
Alternatives and trade-offs
Claude wins when you read and write long documents; it also has a free plan and Pro costs $20 a month.
Gemini wins when your organisation runs on Google Workspace, or when you want a cheap paid tier (Google AI Plus is €4.99 in Germany).
Open WebUI wins when you need a chat window your team hosts itself, in front of a local or EU-hosted model.
Analyze uploaded files with Python, draft text, explain concepts, and search with citations.
Choose it for a broad everyday assistant with built-in data analysis.
Its practical distinction is file-based statistics and plotting through Advanced Data Analysis.
Owner-requested editorial use of public official interface screenshots, 15 September 2026. Copyright retained by the original publisher; attribution preserved. No open licence or hands-on course execution is asserted. Source asset: https://images.ctfassets.net/j22is2dtoxu1/intercom-img-dfd7518cb9b5ab4588cb814d/6ca9c53aec99d74eda875f16468276a2/Screenshot_2026-02-24_at_11.48.11%C3%A2__AM.png. PNG conversion preserves the complete frame.
