RechtGPT Alternative: Research Is Only Half the Work

RechtGPT is an Austrian research product: legal questions answered with citations across five jurisdictions, drafts, billed in queries per month. For research that is good and inexpensive. CASUS covers the same Austrian primary sources and adds the contract work on top. The question is not who can research, but what happens to the document afterwards.

This text places both, using RechtGPT's own published information as of August 2026. It is written by CASUS, one of the providers in the same market. Every statement about RechtGPT therefore comes from its own presentation and can be checked there.

What RechtGPT is

RechtGPT is built by Prime AI GmbH in Baden, founded in February 2025 by Stefan Fenz and Michael Berg-Müller. By its own account more than 650 legal teams rely on it. Its audience is broader than most competitors' and expressly includes tax advisors, in-house legal teams and public administration alongside law firms.

The feature set goes well beyond research:

• Legal research with citations, RAG-based, with filters and the relevant passages shown in the original source

• Appeal agent for structured drafts, supporting deadlines, structure and lines of argument

• Contract drafts based on a description of the facts

• Document processing with OCR, including scanned and legible handwritten originals

• Custom sources from the Business tier upwards, stored encrypted and shared within the subscription

• Integration with Microsoft Word, Copilot, Teams and Outlook

Added to that are a pseudonymisation function that replaces personal data before upload, a visualisation of the reasoning path, and selection of historical statute versions from 1 January 2015 onwards. Models used include OpenAI models hosted in the EU Azure cloud, according to the provider.

This is a serious product, and dismissing it as a search engine with a language model attached does not do it justice.

Where the line runs

The difference between the two products shows most clearly in the billing unit. RechtGPT bills in queries per month: 50 on Starter, 250 on Business. That is a sensible unit for research, because a legal question is a discrete request.

A contract review is not a query. It is a pass over an entire document in which every clause is held against a set of interests, and it produces not an answer but a list of findings with attribution and weight. Reviewing a thirty-page framework agreement is not asking a question but having a document worked through.

Three differences follow, and all three deserve settling before a decision.

First: review from a party's perspective. A contract is not good or bad in the abstract but favourable or risky for one of the two sides. A risk review identifies the contracting parties and analyses from the perspective of your own side, with attribution, relevance and severity for each finding, and with wording options that can be adopted directly. In practice this takes one to three minutes and returns around twenty findings per contract.

Second: benchmarking against a standard. The second question on every incoming contract is whether it matches what the firm or company has set as its standard. A benchmark checks against an internal playbook or against best practice, names missing topics and incomplete provisions such as liability without a cap or confidentiality without a deletion duty, and reports the match as a percentage.

Third: work across document sets. In due diligence or a contract inventory the subject is not one document but two hundred. What is needed then is a table in which each column rests on a self-defined extraction field, so that liability, termination, IP and service levels can be compared across the whole set.

RechtGPT states no equivalent for these three jobs. A deliberate product decision, with a consequence: after the research, the work continues by hand.

Jurisdictions and sources

On jurisdictions, RechtGPT is the broader offering. By its own account it covers Austria, Germany, Switzerland, Liechtenstein and Union law, with the list expressly naming the CJEU, the General Court, the EFTA Court and EUR-Lex. For advisors working across borders that is a tangible argument.

CASUS covers Switzerland, Germany, Austria and Liechtenstein. For Austria that means statutes plus the case law of the OGH, VwGH and VfGH, together with Findok, the financial market authority and the competition authority. Union law is currently not covered, and that is the one gap that remains against RechtGPT. Anyone regularly working with CJEU case law should weigh that in the decision.

On Austrian primary sources both providers draw on the same federal legal information system. Queried through the official interface, RIS held around 895,000 judicial decisions in August 2026. CASUS covers that body in full; RechtGPT names the same RIS sources but does not quantify the extent.

RIS is freely accessible, so the real question is not who reaches it but what happens to the results afterwards. That is precisely where a research result and a contract review part ways.

Price and usage model

RechtGPT publishes its prices, which is the exception in this market and speaks for the provider. As of August 2026, per user and month excluding VAT:

Starter costs EUR 29 and includes 50 queries; Business costs EUR 49 and includes 250 queries along with agent mode, the appeal agent and custom sources. A free trial with seven questions and an Enterprise tier from five licences complete the range; all are cancellable monthly.

For research alone RechtGPT is considerably cheaper than CASUS, and anyone staying inside the allowance carries no cost penalty even under heavy use. CASUS prices are likewise public on the pricing page, and carry no query allowance in return.

Comparing two subscription prices answers only half the question. A research tool is measured against the search time it saves; a platform that also reviews, against the review itself: a risk review takes one to three minutes and returns around twenty findings with a severity rating and wording options. What those minutes replace, every firm knows better than any brochure, and can hold against the difference between EUR 49 and EUR 155 a month.

The calculation shifts only when two things coincide. First, Business stops at 250 queries a month; beyond that RechtGPT points to its sales team, and the price is no longer public. Second, a research tool does not cover contract work. Anyone needing both runs two tools side by side and should compare the total, not the individual price.

Security and professional rules

Both providers argue on data protection compliance but differ in how they implement it. RechtGPT states that it works in line with professional rules and the GDPR, stores no queries or answers server-side (storage being optionally available), and offers pseudonymisation as an upstream step.

At CASUS, client content sits on Google Cloud in the Zurich region, encrypted with AES-256 at rest and TLS 1.2 or higher in transit. Inference runs through Google Vertex AI in Belgium and Microsoft Azure OpenAI in Switzerland and Sweden, not in the United States. Zero data retention is agreed with the model providers, and for Azure an opt-out from abuse monitoring is in place. One practical consequence follows: pseudonymisation before upload is not required. It exists as a function for onward disclosure to third parties, not as a precondition of use.

For Austrian firms, section 40 para. 3 RL-BA 2015 applies on top. It permits engaging external providers for electronic data processing, but requires among other things that the provider be contractually obliged to report a search of the premises without delay, and that clients be informed about the categories of provider engaged.

That reporting duty is the fastest way to tell providers apart: it is either in the contract or it is not. At CASUS it sits in the Austria annex to the terms: the company undertakes to inform the customer without delay in the event of official seizure, a search of the premises or any other official access, as far as legally permitted, with express reference to section 40 para. 3 no. 3 RL-BA 2015. Anyone comparing should ask each provider to point to that commitment in the contract text rather than accept a general statement of compliance. What CASUS commits to on hosting, encryption and professional rules is set out on the security page.

The differences at a glance

n/a means: not stated by the provider, as of August 2026.

Which tool when

Both tools research the same Austrian primary sources. The decision therefore falls not at the research stage but after it: does it end with the answer to a legal question, or does work on the document follow?

RechtGPT suits anyone resolving legal questions across several jurisdictions, working regularly with Union or Liechtenstein law, drafting appeals and official decisions, operating on a tight budget, or coming from tax advisory or public administration.

CASUS suits anyone who, alongside research, receives third-party contracts daily, benchmarks against internal standards, negotiates from a party's position, analyses document sets, or does most of the work in Word anyway. In most firms that is the larger part of the day. That last point is concrete: at CASUS around 80 per cent of usage happens in the Word add-in, not the web application.

Two tools side by side make sense for a practice centred on Union or Liechtenstein law. For most firms they mean two subscription prices, two roll-outs and two places where client data sits. The Austrian provider comparison covers the whole market.

Anyone wanting to test whether document work makes the difference can trial CASUS free for fourteen days with no minimum term. That is only worthwhile with real contracts from live work: a trial with a sample document shows that the software runs, not whether it changes anything day to day. Access can be set up in a few minutes.

FAQ

What does RechtGPT cost?

As of August 2026 RechtGPT lists four tiers per user and month excluding VAT: a free test with 7 trial questions, Starter with 50 queries at EUR 29, Business with 250 queries at EUR 49, and Enterprise with 100 queries from five licences at EUR 29. All tiers are cancellable monthly.

Does RechtGPT have a Word integration?

Yes. By its own account RechtGPT integrates with Microsoft Word as well as Copilot, Teams and Outlook, and documents can be processed directly in Word. The occasional claim that RechtGPT runs solely as a Copilot agent is inaccurate.

Can RechtGPT review contracts?

RechtGPT analyses documents, answers questions about them and produces contract drafts. It does not state a structured risk review from a party's perspective with severity per finding, nor a benchmark against a stored playbook with a percentage match. Both are standard in CASUS, together with analysis across whole document sets.

Which jurisdictions does RechtGPT cover?

Austria, Germany, Switzerland, Liechtenstein and Union law, including the CJEU, the General Court, the EFTA Court and EUR-Lex. The data is updated daily by its own account, and statute versions can be selected by reference date from 1 January 2015 onwards.

What is the difference between research and contract review?

Research answers a legal question from statute and case law and ends in a sourced answer. A contract review takes an existing document and names what is problematic about it for a particular party, with weighting and wording suggestions. The second job presupposes the first but is not the same thing.

Is CASUS an alternative to RechtGPT?

Only in part, and that is the honest answer. For pure legal research across five jurisdictions RechtGPT is cheaper and broader on Union law. For work on the document, meaning review, benchmarking against standards and analysis of entire sets, CASUS covers jobs for which RechtGPT states no equivalent.

Must documents be anonymised before upload?

Not with CASUS: processing takes place in Switzerland and the EU, and zero data retention is agreed with the model providers. RechtGPT offers pseudonymisation that replaces personal data before upload. Anyone weighing the question should establish, for each provider, where processing happens and whether the model suppliers cache content.