ChatGPT has long since found its place in law firms, legal departments and sole practices. Lawyers use it to write emails, summarise contracts, research legal terms or produce first drafts of clauses. The tool is fast, easy to access and, for a general-purpose language model, remarkably capable.
At the same time, using it in everyday legal work isn’t automatically plain sailing. ChatGPT wasn’t built for legal work, and that shows in practice. If you understand where the tool helps, where its limits are and when alternatives make sense, you have a clear advantage.
What lawyers actually use ChatGPT for
In practice, the following use cases in particular have become established:
- First drafts of contracts, memos, NDAs and client letters
- Summaries of long documents, files or legislation
- A first overview of legal concepts by jurisdiction
- Creating structured checklists and issues lists
- Making complex legal language understandable for clients
- Comparing clauses across different document versions
The efficiency gains are real. Tasks that would otherwise take 45 minutes can often be prepared in ten, not because the AI replaces legal judgement, but because it does the groundwork.
The risks lawyers need to know about
ChatGPT is a general-purpose language model. It wasn’t trained on curated legal databases, doesn’t know recent judgments and has no access to a firm’s precedents or client data. And it hallucinates.
This means the model can produce information that sounds plausible but isn’t true. In many contexts that’s merely annoying; in the legal field, though, it can have serious consequences.
German courts specifically identified this problem in 2025. In the order of the Higher Regional Court (OLG) of Celle of 29 April 2025 (ref. 5 U 1/25), counsel for the defendant cited four supposed decisions from various higher regional courts. None of them could be found in juris or beck-online.
Shortly afterwards, the Local Court (Amtsgericht) of Cologne (ref. 312 F 130/25) reached a similar conclusion: every source named in the brief had been made up. The court made clear that this kind of conduct undermines trust in lawyers’ submissions.
On top of this, there’s a structural confidentiality problem: entering client information into a public AI tool risks breaching professional secrecy and the GDPR. Standard ChatGPT configurations aren’t designed for this.
How to use ChatGPT sensibly in everyday legal work
Despite these risks, ChatGPT can be a helpful tool, provided it’s used deliberately and in a structured way.
Well-written prompts are essential
The quality of the output depends heavily on the input. Vague questions lead to vague answers. A good legal prompt contains four elements:
- Role: What perspective should the AI take?
- Context: What information is relevant (facts, jurisdiction, goal)?
- Task: What exactly should happen (analyse, draft, compare)?
- Format: What should the result look like (list, table, memo)?
Example: “You are an experienced employment lawyer in Germany. Review the following termination clause and identify any provisions that could breach German employment law. Present your findings as an issues list with a short rationale and a severity rating.”
This produces a structured, usable result. A prompt like “Have a look at this clause” doesn’t.
Which tasks ChatGPT is suited to
ChatGPT is particularly helpful for first drafts and structuring. These include:
- Drafting standard clauses
- Explaining legal terms
- Creating checklists
- Summarising non-confidential documents
The tool is less suitable for anything that has to be legally watertight, for example:
- Exact case citations
- Working with confidential client data
- Compliance reviews
- Briefs without subsequent review
Always check the output yourself
Every AI output must be professionally reviewed before it’s used. That’s not a recommendation; as a rule, it’s required by professional regulations.
Case citations should always be checked against primary sources. Statutory references must be verified independently. It makes sense to treat AI output like the work of a good trainee: a helpful basis, but nothing you adopt without checking.
Where ChatGPT reaches its limits
As soon as larger-scale or particularly demanding legal work is involved, the limits become clear. Examples include:
- Due diligence across large volumes of documents
- Structured extraction of specific data points
- NDA review with redlining
- Systematic risk flagging
Tasks like these require more than a text-based answer.
What specialised legal AI does differently
This is exactly where specialised tools such as BEAMON come in. They aren’t designed for general text, but for specific legal workflows.
One key difference is how they deal with hallucinations: answers are based on verified legal sources, citations are checked against databases, and traceable source references are provided.
There are clear differences on data protection too: BEAMON is certified to ISO 27001:2022 and SOC 2 Type II, meets the requirements of the German Federal Lawyers’ Act (BRAO) and § 203 of the German Criminal Code (StGB), and processes data exclusively within the EU/EEA. Client data is neither used for training nor passed on to third parties.
The key features:
- Assist: research, analysis and drafting with source references
- Extract: structured data analysis across multiple documents
- Word Plug-in: contract review and redlining directly in Microsoft Word
The results are traceable, structured and tailored to legal use.
ChatGPT or specialised legal AI
The key question isn’t whether to use AI, but how.
For simple, non-confidential tasks and first drafts, ChatGPT is often enough. But as soon as client data, legally robust results or auditable processes are required, specialised tools are the better choice.
The 2025 court decisions show clearly: the risk of using the wrong tool is real, and can lead to sanctions, loss of trust and reputational damage.
What legal professionals should do now
- Define clear internal guidelines (permitted use cases, tools, review processes)
- Train teams in how to write prompts
- Use specialised tools if complex legal tasks come up regularly. If you’re curious, you can try BEAMON for free at: ai.bryter.io/sign-up.
