Casetext CARA AI is a legal AI research assistant that automatically matches appropriate precedents, legal provisions and court rules when lawyers upload case briefs.

Casetext CARA AI

Core parameters and statistics of Casetext CARA AI

The core value proposition of CARA AI is "directly upload your legal documents, and AI will automatically find the most relevant precedents" - which is essentially different from the traditional legal research method in which lawyers need to construct their own keywords and search repeatedly. It is not a general chatbot, but a dedicated research engine deeply bound to the American legal system.

Projects	Public Information
Official Positioning	AI Legal Research Assistant
Core capabilities	Upload documents → Semantic understanding → Automatic case/law/rule matching
Data coverage	U.S. federal and state cases (including Westlaw database), statutes, administrative regulations, and court rules
Parent company	Thomson Reuters (acquired Casetext for $650 million in 2023)
Technical Route	Legal NLP Semantic Understanding + Legal Knowledge Graph + Shepard's Reference Verification
Latest version	2026
Product form	Web side, embedded in Westlaw ecosystem
Supported languages	English (US Legal English)
Target users	American practicing lawyers, law firms, and corporate legal departments
Competitive product benchmarking	Westlaw Precision, LexisNexis, Bloomberg Law, vLex

Core Difference: Traditional legal research tools (Westlaw, LexisNexis) rely on Boolean keyword searches, and lawyers need to first think about "what words to use for search." CARA changes the interaction paradigm—lawyers upload the brief they are writing, and AI automatically analyzes the legal disputes, cited cases, and argument gaps in the document, then recalls the most relevant uncited cases from the massive case library and marks the best insertion position in the document. This is equivalent to upgrading from "search engine" to "collaborator".

User and market recognition of Casetext CARA AI

Gradually build user awareness in the field, and product capabilities are used by content creators and teams to improve work efficiency. Specific user scale and industry adoption data are subject to the official real-time page.

Cost Advantages of Casetext CARA AI

C-side/individual practitioner

CARA AI does not support individual standalone subscriptions. If individual practicing lawyers want to use it, they must access it through a law firm or corporate account. This means independent or solo practitioners will need to incur the cost of a law firm-level subscription, or gain access through Thomson Reuters' Westlaw bundle. For individual attorneys with an annual legal research budget of less than $5,000, the direct subscription cost of CARA may be higher than a traditional pay-per-view Westlaw search.

Law Firm/Team Edition

Pricing is based on a quotation system, and is billed based on the number of users and the combination of functional modules. The official real-time page and business quotation shall prevail. Core variables include:

• Number of user seats: Usually subscribed on an annual basis, the monthly fee per user decreases as the number of seats increases.
• Functional module: The basic version only covers case recommendation and citation verification; the advanced version includes AI legal memorandum generation and batch document analysis.
• Westlaw Bundling: As Thomson Reuters also operates Westlaw, CARA may be sold as a Westlaw plug-in or value-added module and the overall subscription cost will need to be assessed within the Westlaw contractual framework.

Hidden Costs: The training costs of law firms deploying CARA cannot be ignored. Although the design that can be used by uploading documents lowers the learning threshold, lawyers need to build trust and judgment in the AI ​​recommendation results - in the early stage, senior partners may be required to review the AI-recommended cases. This "human-machine running-in" cycle is usually 2-4 weeks. In addition, if the law firm already has a LexisNexis or Bloomberg Law contract, parallel subscriptions will result in duplication of expenditures.

Enterprise/Cross-deployment version

Includes customized integration (interconnected with the law firm’s internal document management system), data analysis dashboard and exclusive training support. Please contact Thomson Reuters commercial team for a quote. For large law firms or multinational corporate legal departments with 500+ lawyers, CARA’s volume licensing unit price is usually lower than that of small law firms, but the total contract value is more than $100,000/year.

API/Developer Cost: CARA AI does not provide a public API. Its AI capabilities are exposed through Thomson Reuters' CoCounsel platform and Westlaw interface, and are not independently accessible to third-party developers. If you want to embed AI legal research capabilities into a self-built system, you need to go through Thomson Reuters' enterprise customization solution, which costs significantly more than standard API products.

Cost comparison reference (the following is a deduced estimate, subject to the official quotation):

Plan	Estimated annual fee range	Applicable scale	Included modules
Westlaw + CARA Basic	$15,000–$50,000/year	Small law firm (5-20 people)	Case recommendations + Shepard's verification
CoCounsel full version	$50,000–$200,000/year	Medium-sized law firm (20-200 people)	Contains AI memo + document analysis
Enterprise customized version	$200,000+/year	Large law firm/corporate legal department (200+ people)	Includes customized integration + exclusive training

Key features of Casetext CARA AI

• Intelligent document analysis and dispute extraction: Upload a brief, motion, answer or any legal document, and AI will automatically analyze the legal disputes, core facts, cited precedents and argumentation logic chain. Different from traditional OCR for keyword extraction, CARA uses a legal semantic model to understand "what is being argued about in this case" rather than "what words are in the document".

• Intelligent recommendation of uncited cases: Based on the semantic understanding of uploaded documents, AI recalls "cases that you should cite but have not yet cited" from the Westlaw case library and sort them by relevance. Each recommended case comes with three pieces of information: the degree of match with the dispute point in the current document, the recommended insertion position (in which paragraph of the document it is cited), and the Shepard's citation status of the case (whether it is still good law).

• Linked recommendation of laws and court rules: In addition to jurisprudence, CARA simultaneously recommends relevant statutes (Statutes), administrative regulations (Regulations) and court rules (Court Rules). Hidden linkage: Case recommendation and legal article recommendation are not isolated - when AI recommends a certain case, it will automatically include the core legal articles cited by the case, helping lawyers complete the "precedent + legal article" evidence chain at once.

• AI Legal Memo Generation: Enter a legal question (for example, "What is the burden of proof for a claim under the Ninth Circuit Court?"), and AI automatically generates a structured legal research memo, including problem analysis, relevant case summaries, legal standard comparisons, and preliminary conclusions. Expert View: This function is synergistic with case recommendation - the cases quoted by lawyers in memos can be linked back to the CARA case recommendation results with one click, realizing the "writing-research" connection.

• Shepard's Citation Verification Integration: Thomson Reuters' Shepard's service is the "gold standard" citation verification tool for US legal research, used to determine whether a precedent is still valid law (good law). CARA embeds Shepard's verification into the precedent recommendation process - lawyers do not need to manually jump to Westlaw for citation verification before adopting the recommended precedent, and can directly see the latest citation status (red/yellow/green light) of each precedent on the CARA interface.

• Dispute point reverse search: Lawyers can enter a specific legal dispute point (such as "whether a certain contract clause constitutes unfairness"), and CARA will reversely search for the most relevant cases and legal provisions related to the dispute point, instead of relying solely on passive analysis of uploaded documents. Hidden linkage: This function complements document analysis - first perform a reverse search on controversial points to build an argument framework, then write the framework into a document and upload it to complete missing cases.

• Multi-document cross-reference: Upload multiple relevant documents (such as a complaint and a statement of defense), CARA automatically compares the differences in precedents cited by both parties, overlaps and differences in disputes, and marks "These precedents are not cited in your document, but appear in the other party's document, and you need to prepare a rebuttal."

Casetext CARA AI model and version evolution

Casetext was founded in 2013 as a crowdsourced review platform for legal documents. Since 2017, it has turned to AI legal research and successively launched the CARA (Case Analysis Research Assistant) series of products. The following are publicly available product milestones:

Early exploration phase (2013–2019)

• 2013: Casetext is founded by Jake Heller (CEO), Laura Safdie (COO), and Pablo Arredondo (Chief Product Officer). The initial product form was a crowdsourcing review platform for legal documents.
• 2017: Launched the CARA prototype, using NLP technology for legal document analysis and case recommendation for the first time. The initial version only supports federal case search.
• 2018: CARA expands to state-level precedent coverage, adding Shepard's citation verification integration.

AI research mature stage (2020–2022)

• 2020: Released a major update to CARA AI, introduced a legal semantic understanding model based on the Transformer architecture, and significantly improved the accuracy of case recommendation. In the same year, the AI ​​legal memorandum generation function was launched.
• 2021: Added multi-document cross-reference function, supporting uploading of complaint and defense for mapping of precedent disputes. Start covering statutory laws and administrative regulations recommendations.
• 2022: Launch of the CoCounsel brand (Casetext’s AI legal assistant brand), integrating CARA’s research capabilities with AI document analysis and contract review capabilities. Begins engagement with Thomson Reuters regarding acquisition.

Acquisition integration stage (2023–present)

• June 2023: Thomson Reuters announced the $650 million all-cash acquisition of Casetext, with the deal closing in August of the same year. Product roadmap shifts towards integration with Westlaw ecosystem.
• 2024: CARA AI is embedded in the Westlaw platform as a CoCounsel core module. Released the "Reverse Search for Disputed Points" function to support lawyers to directly search for relevant precedents based on legal issues.
• 2025: CARA AI 2025 version is released, introducing an enhanced version of the AI-generated legal memorandum function, supporting longer context (analyzing 50+ pages of legal documents). Integrate the Practical Law content resources of Thomson Reuters and recommend case studies with links to practical guidelines.
• 2026 (latest): CARA AI 2026 version is released, which enhances the accuracy of the AI ​​matching algorithm and optimizes the depth of document analysis to paragraph-level semantic understanding. CoCounsel surpasses 1 million users.

Version Summary:

Version	Time	Key Changes
CARA prototype	2017	NLP legal document analysis + federal case recommendation
CARA 2020	2020	Transformer semantic model AI memo generation
CARA 2022	2022	CoCounsel Brand, Multi-Document Comparison
CARA 2024	2024	Westlaw integration, dispute point reverse search
CARA 2025	2025	Long documents support Practical Law integration
CARA 2026	2026	Paragraph-level semantic matching, matching algorithm enhancement

Casetext CARA AI’s technical advantages

Law-specific semantic understanding engine: The bottom layer of CARA is not a direct application of a general-purpose large language model (such as GPT-4), but based on the general model, it uses more than 4 million U.S. legal documents (including precedents, statutes, motions, and briefs) for domain fine-tuning. This "legal-specific embedding" enables the model to understand the subtle differences between legal concepts such as "holding" (the gist of the judgment), "dicta" (opinion) and "stare decisis" (following precedent), rather than just doing semantic similarity calculations at the keyword level.

Knowledge graph enhanced case recommendation: CARA's legal knowledge graph includes case citation relationships (which case is cited/overturned/distinguished by which subsequent case), legal dispute point hierarchy (what sub-disputes are under a dispute point) and legal article-case cross-index. When AI analyzes uploaded documents, the graph-assisted model understands "which clusters of cases this argument in the document is usually related to" rather than relying solely on text similarity matching. The effect is that for rare dispute points or cross-domain cases, the recall accuracy is significantly higher than the pure embedding method.

Paragraph-Level Analysis Granularity: The 2026 version supports mapping each paragraph of the instrument to specific precedents and legal standards. The traditional approach is to extract dispute points from the entire document and then recommend cases in batches. Lawyers need to match the reference position of each case in the document. CARA's paragraph-level analysis directly locates the precedent recommendation results to the specific paragraphs recommended for insertion, reducing lawyers' manual matching work. Taking the brief of the case as an example, AI can determine that "the controversial point X discussed in the third paragraph is most suitable to cite case A, and the argument of insufficient data in the fifth paragraph is most suitable to cite case B."

Shepard's real-time citation verification: Shepard's service updates the "health status" of each case in real time by analyzing how subsequent cases cite the target case (positive citation/distinguishing/overturning/challenging). CARA upgrades Shepard's verification from "manual query" to "automatic pre-filtering" - excluding cases that have been overturned or seriously questioned during the recommendation stage, ensuring that the recommended results are "good law" from the source. This avoids the rework costs of lawyers adopting the recommendation only to discover that the precedent is invalid.

CoCounsel multi-modal capability collaboration: CARA is not an isolated legal research tool, but part of the Thomson Reuters CoCounsel AI ecosystem. CoCounsel also covers document review (Contract Analysis), compliance checking (Compliance Checking) and due diligence (Due Diligence). Hidden synergy: Legal issues identified by lawyers when reviewing contracts in CoCounsel can be pushed to CARA for relevant case studies with one click; case conclusions drawn from CARA research can also be written back into CoCounsel's contract risk annotation. This combination of "problem identification → legal research → risk labeling" cannot be provided by a single legal research tool.

How to use Casetext CARA AI

Access entrance

CARA AI currently only provides web access, and there is no independent desktop application or mobile app. There are two entrances:

• Casetext independent platform: accessed through https://casetext.com/cara-ai (original Casetext user portal, may gradually transition to Westlaw unified platform)
• Westlaw Embedded: Users who have subscribed to Westlaw can directly enable the CoCounsel/CARA function module in the Westlaw interface without additional login.

Typical usage steps

1. Upload documents: Upload the case summary (PDF or Word format) in the CARA interface. The upper limit of file size is subject to the official page. The system automatically analyzes the document structure and extracts legal dispute points.
2. Confirm analysis parameters: Select the jurisdictional court (federal/specific state), whether Shepard's verification is required, and the upper limit of the number of recommended precedents (usually 5-20). AI completes initial analysis in seconds.
3. Review case recommendation: Browse the list of cases recommended by AI. Each item includes the case name, reference number, Shepard's status of the dispute point matching degree and the recommended insertion position. Supports one-click opening of the full text of Westlaw cases for in-depth reading.
4. Incorporate document modification: Add the precedents confirmed to be cited to the document. CARA provides formatted citation text (Bluebook format) that lawyers can copy and paste directly.
5. Generate memo (optional): If the research purpose is to write a legal memo, describe the problem in the "Legal Issue Input" box, and AI will automatically generate a structured draft memo. Review and revise and then export to Word or PDF.

Comparison with competing products

Dimensions	CARA AI	Westlaw Precision	LexisNexis
Input method	Automatic analysis of uploaded documents	Keyword/natural language search	Keyword/natural language search
Recommended granularity	Paragraph-level matching	Document-level search results	Document-level search results
Citation verification	Shepard's automatic pre-filtering	Shepard's manual query	Independent citation verification tool
Memo generation	AI automatic generation	Not supported	Not supported
Learning curve	Low (upload and use)	Medium (need to master search syntax)	Medium (need to master search syntax)

Integration and Deployment

CARA AI does not provide a public API or SDK. If enterprise customers want to embed AI legal research capabilities into internal systems, they must go through Thomson Reuters' enterprise-level customization solutions, which usually involve the privatized deployment of CoCounsel API interfaces. The deployment cycle is subject to the official business plan.

Product Pricing for Casetext CARA AI

CARA AI does not offer a public self-service subscription page or standardized price list, all pricing is quoted through the Thomson Reuters sales team. The following is a summary of the pricing structure based on public information:

Price Tier

Individual/Small Law Firm: Independent purchase is not supported. Small law firms (under 5 people) need to subscribe to the add-on CARA module through Westlaw Basics. Estimated additional costs are $200–$500 per person per month, subject to official quote.

Medium-sized law firms (20-200 people): CoCounsel full version subscription, annual payment per user seat. Typically includes Westlaw core database access + CARA AI research capabilities + AI contract review + dedicated account manager. The estimated annual fee is $50,000–$200,000, and the average monthly fee per person is approximately $200–$800.

Large law firms/corporate legal departments (more than 200 people): Enterprise customized version, including private deployment options (running AI models within the law firm), deep integration with existing document management systems (such as iManage, NetDocuments), exclusive training and SLA guarantee. The annual fee is usually above $200,000, please contact us for negotiation.

Factors affecting pricing

• Database Scope: Only US federal cases vs. 50 state cases + administrative regulations + international law materials
• Function module: Basic case recommendation vs. memo generation + multi-document comparison + reverse search of disputed points
• Integration requirements: Do you need to connect with the law firm’s existing workflow system?
• Contract Period: Multi-year contracts typically receive a 10-20% discount

Tip: Because Thomson Reuters uses a unified bargaining model for Westlaw and CoCounsel, the actual price of CARA is highly dependent on the law firm's overall Westlaw contract size. It is recommended to ask the sales team to issue a stand-alone CARA module quote at the time of purchase rather than blurring the billing into the Westlaw renewal contract. The official real-time quotation shall prevail.

Casetext CARA AI application scenarios

• Litigation lawyer's document finalization: After writing the brief, upload it to CARA, and AI will add "key precedents you may have missed" and "objections that need to be responded to". Quantitative deduction: It usually takes 2-4 hours for litigation lawyers to manually conduct a complete case search. CARA compresses document analysis + case recommendation to 5-10 minutes, improving efficiency by about 10-20 times. However, the jurisprudence recommended by AI still requires manual review and confirmation by lawyers - the typical "search" stage has been reduced from hours to minutes, and the "review" stage has remained unchanged at 1-2 hours, saving about 50-70% of the overall time.

• Quick start with cross-field cases: A lawyer specializing in corporate law suddenly received an intellectual property case. After uploading the opponent's complaint, CARA recommends the core case law framework in the field within 10 minutes, helping lawyers quickly build a case map for unfamiliar areas. Boundary of Human-Machine Collaboration: CARA acts as a "field quick reference manual" in this scenario, completing the discovery of precedents 100% automatically; however, the evaluation and screening of precedents (which precedents are the most convincing and which ones have been substantially restricted by subsequent precedents) requires the professional judgment of lawyers, and AI recommendations cannot be copied in full.

• Legal research training for newcomers to law firms: After interns or newly hired lawyers write their first case brief, upload it to CARA to see which cases AI recommends, and compare it with the results of your own manual search. Cost reduction and efficiency improvement: In traditional training, senior lawyers need to spend 2-3 hours reviewing the case search work of newcomers item by item; with the assistance of CARA, senior lawyers only need to focus on discussing the differences between AI recommendations and newcomers' searches. The review time is shortened to 30-45 minutes, and the law firm's personnel training costs are reduced by about 60-70%.

• Alignment of dispute points in multi-lawyer collaboration: In large litigation cases, multiple lawyers each write different parts (statement of facts, legal arguments, damages). After uploading separately, CARA found that there were inconsistencies in the reference of precedents in each part (the reference position and expression of the same precedent in different parts conflicted), and marked the need for unification. In this scenario, AI's cross-document comparison capability far exceeds the efficiency of manual review one by one.

• Strategic Research on Appeal Letters: When preparing an appeal letter, lawyers need to study "the recent attitude of the Court of Appeal towards this controversial issue." CARA's reverse search of dispute points can answer "the trend of the Ninth Circuit's jurisprudence on a certain issue in the past two years", helping lawyers predict the success rate of appeals and adjust argumentation strategies.

Who is Casetext CARA AI suitable for?

• American Litigation Attorney: the most direct target user. You need to write briefs, motions, and defenses on a daily basis. CARA's paragraph-level precedent recommendations can be directly embedded into the document writing workflow to reduce the interruption of "stopping to do a search." Prerequisites: Obtained access to Westlaw/CARA through a law firm; ability to write legal documents in English.

• Legal Researcher/Librarian: Traditionally responsible for the law firm's case search work. CARA can be used as an auxiliary tool to reduce duplication of primary search efforts, allowing them to focus their time on complex search strategies and in-depth analysis of cases. Not suitable for boundaries: For research tasks that require highly customized search strategies (such as the use of specific Westlaw field codes), CARA's automated recommendations may not be accurate enough, and manual search still needs to be supplemented.

• Company Legal Department Lawyer (In-house Counsel): When faced with cross-field and cross-business lines legal issues, CARA serves as a "quick research portal" to help the legal department build a basic framework of precedents in a short time. Prerequisite: The legal department of your company has subscribed to the Thomson Reuters corporate plan.

• Legal AI Product Manager/Competitive Product Analyst: CARA is a benchmark product for legal AI applications. Its "upload-and-research" interactive paradigm and paragraph-level matching technical route are worthy of in-depth dismantling by product managers and analysts, and serve as a reference for competitive product evaluation and technical route selection.

• Unsuitable people: Legal practitioners outside the U.S. legal system (CARA does not cover precedents outside the Anglo-American legal system); lawyers who focus on non-litigation business (transactions, compliance, intellectual property applications) rather than litigation disputes, and the core value of CARA (case recommendation) has limited intersection with their daily work; individual practitioners or cost-sensitive small law firms, it is recommended to first evaluate the budget and actual frequency of use; scenarios that require Chinese or non-English legal research support (CARA only supports English U.S. legal texts).

Casetext CARA AI Summary and Outlook

Based on the concept of "uploading documents is research", CARA AI has achieved a paradigm upgrade from "search-based" to "understanding-based" in the field of legal research. Its core competitiveness lies in: legal-specific semantic understanding model + knowledge graph-enhanced case recommendation + paragraph-level matching granularity + Shepard's closed design of automatic citation verification. After being acquired by Thomson Reuters, CARA’s technology synergizes with Westlaw’s vast data ecosystem and CoCounsel’s multi-modal capabilities to form a systemic advantage that is difficult to match with a single legal AI research product.

Current Limitations:

• Geographic Legal System Limitations: Only covers U.S. federal and state laws, and does not support case retrieval from other common law countries (UK, Canada, Australia, etc.), nor does it support civil law systems.
• Non-English support missing: Only English legal documents are accepted, and document analysis or case search in any other language is not supported.
• Pricing is not transparent: There is no public self-service pricing, procurement decisions rely on Thomson Reuters' business negotiation process, and small law firms face higher transaction costs.
• Independence Risk: CARA is no longer an independent product, but an integral part of the Westlaw/CoCounsel ecosystem. If the law firm does not plan to migrate to the Thomson Reuters system as a whole, the possibility of purchasing CARA separately is extremely low.

Follow-up observation points:

• Will CARA’s AI capabilities expand to non-US legal systems (such as UK and Canadian legal content that Thomson Reuters already covers)?
• How many of CoCounsel’s million user base are active users of AI legal research capabilities, as opposed to just contract review users?
• As general large language models (such as GPT-5, Claude 4) improve their capabilities in legal question answering, can CARA's legal-specific fine-tuning route continue to maintain an advantage gap over general models?

Procurement/Adoption Risk Assessment: It is recommended that law firms complete the following three verifications before deciding to purchase: ① Apply to the Thomson Reuters sales team for a 14-30-day free trial, upload a case summary of the actual cases represented by the law firm, and evaluate the recall rate and accuracy of case recommendations; ② Do a blind test comparison (double-blind comparison) between CARA's case recommendation results and the independent search results of senior partners of the law firm (double-blind comparison), and calculate "AI The proportion of "recommended but not found by senior partners" and "recommended by senior partners but not recommended by AI"; ③ Evaluate whether CARA can be integrated with the law firm's existing document management system (such as iManage, NetDocs) and billing system to avoid additional data migration costs. Make a long-term subscription decision after confirming the above three items.

How to use Casetext CARA AI

• Web client: You can use it by visiting the official website and registering an account. Most functions do not require installation.
• API Access: Provides RESTful API, developers can obtain the API Key and integrate it into their own applications.

Version Info

• CARA AI 2026 ：There is no official precise date yet. Enhanced AI matching algorithm and document analysis depth. （2026-07）
• CARA AI 2025 ：There is no official precise date yet. Introducing the AI-generated legal memorandum function. （2025）