DocJuris vs LegalSifter: how they compare in 2026
DocJuris and LegalSifter both review incoming contracts against a company's own playbook and return redlines in Word. Each is stronger in a different area. DocJuris is stronger on evidence and isolation. It names eight customers with a figure for each, including Flex cutting review from eight days to five minutes, and runs each customer in its own single tenant deployment. Its terms commit to notice before a government compelled disclosure. LegalSifter is stronger on what it tells a buyer about the AI itself. It publishes an issue spotting accuracy of 95 percent or more, prices each contract review from $80, and lets customers choose OpenAI or Google Gemini to process their contracts. It also commits to breach notice within 24 hours. DocJuris's terms grant it a perpetual license to customer data and output to improve its services. LegalSifter's contractual commitments sit in its separate terms and conditions.
Which one fits
Choose DocJuris if
- You want an isolated deployment for your legal team. DocJuris runs each customer as a single tenant on its own subdomain with its own service stack, single sign on through your identity provider, and permission groups scoped down to individual repositories and applications.
- You want more than contract review from one vendor. DocJuris offers about two dozen applications, from vendor screening and document generation on your own form to lease abstraction, outside counsel invoice review and security questionnaires, and builds a new one for a fee when none fits.
- You want to choose how closely you review each position. DocJuris offers a Quick Pass for speed and a Guided Review that stops at each position for confirmation, returns every edit as a tracked change, and names Siemens, Flex and Purolator with measured results.
Choose LegalSifter if
- You want to price each review before you buy. LegalSifter publishes $80 per contract on its Basic plan, $72 on Team and $65 on Enterprise, for contracts up to ten pages, with standard and custom playbooks included on every tier.
- You want to choose which AI provider reads your contracts. LegalSifter lets an organization configure whether OpenAI or Google Gemini processes its contracts, and its Google Docs listing documents what the add in reads, creates and sends.
- You want deletion and breach notice commitments. LegalSifter states that clients decide when their data is deleted, can separate content for scenarios such as litigation hold, and will be told within 24 hours of a confirmed data breach.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
DocJuris's catalog is roughly two dozen applications, each delivering a finished AI work product: a redline with tracked changes, a deviation memo on the customer's template, a restated contract, an Excel lease abstract, a completed security questionnaire, a research brief with citations. Pricing follows the same logic, sized to contract volume and the applications enabled rather than headcount. On the custom route, the vendor scopes a described workflow and ships a new application in weeks. The catalog does include an operations layer, with intake forms, matter boards, approvals, a repository and a counterparty deal room, and that kind of workspace could run without models. DocJuris describes each of these as run by models: the repository is searchable by meaning, obligations are extracted, and invoice verdicts are computed rather than guessed. It also positions itself against workflow products, saying it returns a finished redline rather than a workflow ticket.
ReviewPro uses models to interpret contract language, detect playbook issues, propose edits and answer questions through a contract assistant. The playbook provides the organization's positions and fallback instructions, so interpretation and generation are essential to the review service. The current product also includes a versioned contract repository, intake ticketing and controlled playbook management. Those organizational functions still have a use when an AI review is not running. LegalSifter's human services can customize playbooks or handle redlining and extraction, while Contract Logix is a separate contract management offering within the company. The broader workflow now extends beyond the original automated review engine.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
A Legal Research application in the DocJuris catalog writes briefs with citations from primary law across eight jurisdictions, so the product generates authority and not just clauses. Markups come from the customer's own playbook and precedent, generated documents put deal terms onto the customer's form with nothing silently invented, and security questionnaires are answered from the customer's own policies with nothing invented. No accuracy figure, test set or error rate is published, and no retrieval method is described. Nothing says what the system does when it finds no support, or that a user can open the primary authority behind a research brief. Subscription Terms section 5.03(b) states that, given the probabilistic nature of machine learning, the Services may produce incorrect Output that does not accurately reflect real people, places, facts or laws, and directs the Client to evaluate accuracy, including through human review.
ReviewPro 2.0 evaluates clauses against a selected playbook and returns proposed edits with rationale for the reviewer. Its July 2026 study used 30 NDAs, 16 positions and three runs per agreement, comparing results with attorney approved edits. With the engine and agreements fixed, richer playbook instructions raised issue detection from 91% to 95%, selection of the right action from 77% to 93%, and edit fidelity from 55% to 93%. These are distinct measures of contract review, not a general guarantee of legal accuracy. The benchmark does not cover citation treatment, an abstention threshold or performance across other agreement types.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
Review comes in two named modes, Quick Pass where speed matters and Guided Review where the user confirms each position. The review surface is the document itself. Every proposed edit arrives as a tracked change in Word that a lawyer accepts or rejects, and internal comments are kept from the counterparty. DocJuris describes the product as recommending focused changes through a checklist with a human in the loop, with the user in control of strategy, and Subscription Terms section 5.03(b) makes the Client responsible for evaluating Output and names human review as the means. Other marketing describes an operations layer that runs itself, invoice verdicts computed rather than guessed, and agents that solve any workflow. Nothing published sets a threshold, confidence boundary or class of work at which the system goes ahead without a reviewer.
ReviewPro runs a first pass against the selected playbook and returns redlines and explanations in Word or Google Docs. The customer must vet standard or custom playbooks before using them, and reviewers remain responsible for checking and accepting the result. The terms impose an additional categorical boundary on Trial Features: they are solely for evaluation and testing, and cannot be used in production. Named seat plans also require interactive human use rather than unattended batch processing. Routine business review can follow the legal team's approved standards, but the terms recommend legal counsel and leave responsibility for the final agreement with the customer.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Eight customers are named, each with a figure. Flex: contract review down from eight days to five minutes, described as 99 percent less review time. Siemens: ten times more consistent review. GEODIS: 75 percent faster approvals from remote teams. Ancestry: execution twice as fast across procurement workflows. Purolator: over $300,000 saved and 50 percent lower legal technology spend. Careington: five times more consistent review against standard terms. CMA CGM America: 90 percent faster turnaround. ECS Limited: twice the review speed, including on locked and scanned files. Three come with a named executive quoted: the General Counsel of Siemens APAC, the Deputy General Counsel of CMA CGM America and the Head of Legal Operations at Purolator. No date appears against any of the eight, and no method is published for any figure, so how 99 percent or ten times was worked out is not stated. The full case study library is published.
A July 2026 study describes a field observation involving 15 complex agreements, where an unnamed legal team reached 90% alignment with its intended edits and reduced a roughly 90 minute review to about five minutes. That account is separate from the controlled 30 NDA benchmark used to test playbook instruction quality. The field result indicates a concrete workload and reported change, but gives neither the customer's name nor a detailed timing protocol. The internal benchmark explains its agreements, positions and repeat runs; it does not turn the field observation into a reproducible customer deployment study across different contract portfolios.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Subscription Terms section 4.01(e) says DocJuris will hold all Client documents and playbooks confidential with precautions as stringent as it uses for its own confidential information, and will not disclose them in a way that identifies the Client, for the term and five years after. If a government authority compels disclosure, it will notify the Client where permissible and cooperate to obtain a protective order. Separation between customers is architectural: a single tenant deployment on the customer's own subdomain with its own service stack, permission groups scoped to individual repositories and applications, and audit logs of logins and permission changes. Privilege and work product are not addressed. There is no contractual position on training: the marketing says models are never trained on customer data, but section 4.02(f) grants a perpetual, irrevocable, sublicensable license over Client Data and Output to improve or enhance the Services, and the agreement governs. Nothing is published on what a third party model provider may retain, and no provider is named. Retention and deletion are absent from the agreement.
The ReviewPro agreement treats customer contract content as confidential and expressly intends disclosure to preserve attorney client privilege and work product protection. Access by service providers is limited to those who need the information and owe confidentiality duties. Tenant records are isolated from other customers, with role and feature controls inside the service. Training third party AI models is prohibited, but the terms grant LegalSifter a perpetual right to train and improve its own models with anonymized customer content. The security FAQ describes training as a choice made in the agreement. This default permission and the absence of a stated internal matter wall limit the confidentiality position.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
The footer of every DocJuris page says DocJuris is not a law firm or a substitute for an attorney or law firm, and cannot provide any kind of advice, explanation, opinion, or recommendation about possible legal rights, remedies, defenses, options, selection of forms or strategies. The disclaimer sits on the marketing pages as well as in the terms, and Subscription Terms section 5.03(b) directs human review of Output. The audience is named, and the marketing does not describe the product in advice terms. The platform is sold to people who are not lawyers: procurement and sales operations work in the same tenant, are told it is safe to use on their own, and generate redlines and negotiating positions on legal documents. Nothing published explains how a supervising lawyer meets competence and oversight duties over that work, who may use which application, or any jurisdictional limit. A separate Website Disclaimer page is linked in the footer.
LegalSifter identifies ReviewPro as software, disclaims legal advice and an attorney client relationship, and strongly encourages review by legal counsel. Customers must vet their playbooks and verify the resulting work. Business personnel can use standards approved by their legal team, while the agreement leaves legal decisions with the customer. Subscription scope also matters: named seat plans are restricted to internal contracts and matters, while legal services for external clients require express authorization under a consumption plan. The product supports supervision through explained redlines and reusable standards, but does not set jurisdiction specific practice limits or determine whether a particular user may give advice.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
A privacy first posture runs across the DocJuris site, with a statement that customer data never trains AI models and SOC 2 Type II certification, which is security governance rather than AI governance. Two contract provisions also work as governance. Section 4.03(c) states that the Services are not designed for medical records, US Munitions List data, defense articles or ITAR related data, and bars the Client from putting any of it through the product. Section 5.03(b) acknowledges that output is probabilistic. There is no responsible AI page, no framework and no certification such as ISO 42001. Nobody is named as accountable for AI, nothing describes testing before an application ships, and nothing addresses uneven output across contract types, counterparty language or drafting conventions, for a product that applies one company's playbook to everyone else's paper.
LegalSifter separates issue detection, selection of the appropriate action and edit fidelity when evaluating ReviewPro. Its July 2026 study holds the review engine and NDA set constant while varying the specificity of playbook instructions, with attorney approved edits as the comparison. It also compares releases and describes recurring assessment of thoroughness, accuracy and readability. These mechanisms connect product changes to defined review tasks rather than a general responsible AI promise. The terms prohibit discriminatory consequential decisions and malicious attempts to manipulate models. LegalSifter names no accountable AI governance owner and gives no results on uneven behavior across populations or contract categories.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
DocJuris publishes single tenant isolation and single sign on through the customer's identity provider over OpenID Connect, with invite only access when it is off. It also publishes permission groups scoped down to individual repositories and applications, audit logging of logins and permission changes, and usage metering for customer administrators. The Subscription Terms contain no retention period, no deletion or return of Client Data on termination, no incident notification commitment and no security program owed to the customer. Section 4.03(a) and (b) make the Client responsible for its own systems, credentials and uploads, and section 3.02(g) gives the only audit right in the agreement to the vendor, not the customer. There is no subprocessor list and no data processing addendum; the site footer lists every legal document, and neither appears. The website privacy policy is a document hosted by iubenda.
ReviewPro uses encrypted contract storage, AWS key management, role and feature controls, and tenant isolation at the record level. Auth0 supports SAML and OIDC single sign on, with the customer's identity provider enforcing multifactor policies. Contracts are deleted within 30 days after subscription termination, while no ordinary retention period is stated for the full prompt and output history. OpenAI and Anthropic are listed as model processors in the United States, and the terms prohibit their training on customer content. Confirmed customer data or model integrity incidents trigger notice without undue delay, including incidents involving poisoning or model extraction. Retained anonymized material has separate contractual treatment.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
Section 5.01(b) of DocJuris's Subscription Terms caps each side's total liability for direct damages at the greater of fees paid in the prior year or USD 50,000. Section 5.01(a) excludes consequential, incidental, indirect, special and punitive damages, including loss of data and breach of data or system security. Section 5.01(c) shortens the limitation period to two years, and the cap falls away only for fraud or willful misconduct. Section 5.02 is an indemnity covering only third party intellectual property infringement, and its remedies end in a refund of the unearned fee. A bad clause accepted, a missed obligation, a redline that gives away a position, or an invented authority in a research brief all fall on the Client under the agreement. Section 5.03(a) supplies the Services as is and disclaims every warranty, including accuracy and fitness for purpose, and section 5.03(b) makes the Client responsible for evaluating Output. No data breach indemnity, insurance position, service level or warranty of the work product is published.
The ReviewPro terms provide a vendor defense and indemnity for specified claims against the authorized service, but expressly exclude AI output, reviewed contracts and customer playbooks from that protection. LegalSifter disclaims output accuracy and liability for decisions based on the service. The general liability cap is the fees paid during the preceding twelve months, with exceptions including indemnification obligations, gross misconduct and applicable legal duties. Trial Features carry separate warranty and indemnity exclusions and cannot be used in production. These terms allocate platform infringement risk more specifically than a generic disclaimer, while leaving errors in an accepted redline and resulting legal decisions with the customer.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Connectors number more than fifty across CRM, ERP and storage, listed on DocJuris's integrations page. Those named include Microsoft Word, SharePoint, Salesforce, Excel, Google Drive, OneDrive, Dropbox, Box, Asana, Trello, Jira, monday.com, ClickUp, Typeform, Microsoft Forms and Google Forms. Zapier and Power Automate are offered as routes to several thousand more. Contracts go in and come back as Word files with tracked changes, comments and formatting intact and no conversion step. DocJuris also describes intake through Microsoft Forms, redlining inside Word and version control through SharePoint. An API exists with published triggers for turn added, turn completed and internal comment. A documented round trip sends a PDF or Word file from a contract record into DocJuris and returns the redline, using OAuth 2.0 and agent access controls. The API documentation is not public and is obtained by emailing support. The connector list is presented as a sample, and the direction of sync is described for Word but not for CLM and ERP connections.
ReviewPro runs as a Microsoft Word add in applying tracked redlines in the document, and as a Google Docs add in whose marketplace listing documents exactly what it does: reads the current document to extract contract text, creates a redlined copy in Drive, transmits content to LegalSifter's servers and stores nothing locally, with the Google permission scopes listed. The product page states connection to existing apps and CLM software without naming them beyond the company's own Contract Logix.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Each DocJuris customer gets an isolated single tenant deployment on its own subdomain, in the form company.docjuris.ai, with its own service stack. DocJuris describes this as never a shared multitenant black box and as isolation by architecture rather than by policy alone. Deployment follows three published steps: the dedicated tenant, identity provider integration, and administration through audit logs and usage meters. Nothing is published on residency. No region is named, no choice of regions or in country option is offered, no cloud provider is identified, and nothing separates where data is processed from where it is stored. The model providers behind the applications are also unnamed. The hosted privacy policy gives a place of processing for website visitors, which concerns tracker data rather than customer contracts.
ReviewPro operates on AWS with multiple availability zones, encrypted storage and customer separation enforced at the record level. OpenAI and Anthropic appear in the model processor register with United States locations. The security description covers resilience, identity and tenant controls, but does not name an available application hosting region or a customer residency selector. A model provider's listed country is not a complete account of where every processing step takes place. Contract Logix belongs to the same company but is a separate platform, so its Azure hosting arrangement has no bearing on ReviewPro's architecture or a deployment option for this review service.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
SOC 2 Type II is named on at least four DocJuris pages, including the Privacy and Security page and the banner across the site, and is stated plainly in answer to a direct question. Administrative controls are documented one by one: OpenID Connect single sign on, permission groups scoped to repositories and applications, audit logging of logins and permission changes, invite only access when single sign on is off, and usage metering. The evidence sits behind a sales conversation: the published route is to book a demo and bring the security team, and the vendor will walk through the architecture and share the documentation a review needs. No auditor, report period, certificate or report number is published, and there is no trust portal. No second standard such as ISO 27001 is claimed, and nothing describes penetration testing, vulnerability remediation targets or incident response.
LegalSifter states that ReviewPro's security controls undergo independent SOC 2 Type II auditing. Its security material describes AES 256 encryption, TLS 1.2 or later, and AWS key management using FIPS 140 2 Level 3 validated hardware. Staff background checks, security training, monitoring and identity controls accompany the attestation claim. The certification description does not identify an audit coverage period, named assessor or freely accessible report with a scope statement. Encryption standards and a cloud provider's validated hardware support the operating controls but are separate from the scope of LegalSifter's own independent audit.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
Each task runs on the best frontier model, the DocJuris pricing FAQ says, while the catalog describes agents as frontier models solving any workflow and the platform is called proprietary AI. No model or provider is named, no architecture is described, and no commitment to notify customers of changes is published. The same FAQ offers to run inside the customer's own approved AI deployment and mentions Claude and Copilot, but only as tools a customer may already own, not as anything DocJuris uses. There is no subprocessor list and no data processing addendum, and no offer to share the providers on request is published.
The current processor register names OpenAI and Anthropic as large language model and API suppliers for ReviewPro, both in the United States. The security explanation discusses OpenAI's business terms, while the service agreement requires third party AI components not to train on customer content and to follow stated retention controls. LegalSifter separately reserves the right to develop and improve its own models from anonymized customer material. LegalSifter gives no exact model versions, no mapping of review tasks to each provider and no commitment to notify customers of every model substitution. Earlier marketplace descriptions of provider choices do not fully describe this current supplier list.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
No price is published. There are two ways to buy DocJuris. One is an annual platform subscription covering the full catalog, with applications, agents and enterprise integrations included, sized to contract volume and the applications enabled rather than headcount. The other is a services route: a one time build fee, then hosting and usage credits, with change orders. The vendor says it charges nothing per seat, per document or per question. It also publishes two feature comparison tables, gives a reason for withholding the number, and offers 20 percent off a first scope of work engaged within thirty days of a demo. The dollar amounts on the pricing page describe competitors, not DocJuris. The pricing page and the agreement disagree about the unit of charge, and the agreement governs. The page promises unlimited users and no seat fees, while Subscription Terms section 3.02(a) sells a set number of Authorized Users stated in the Order and charges any extra at the rate per Authorized User. The agreement also makes fees payable in advance, irrevocable and nonrefundable, requires ninety days' notice to terminate for convenience, and allows amendment by public posting, with thirty days of silence treated as acceptance.
ReviewPro offers annual volume subscriptions with unlimited seats or annual seat subscriptions with unlimited interactive contract reviews. Volume plans cost $2,750 for 50 reviews, $7,500 for 150, $18,000 for 400 and $40,000 for 1,000. Seat pricing starts at $6,000 a year for one user; larger minimum seat commitments reduce the rate per user. Standard implementation is included at no extra charge. Custom playbook architecture and ongoing custom maintenance are separately scoped optional services. Named seat subscriptions are limited to internal contracts, while legal client work requires express permission under a consumption subscription. The plan choice therefore changes both the charging unit and permitted use.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
DocJuris sells to in house legal departments first, with procurement and sales operations named as users in the same tenant and given their own pages. The stated size range runs from a department of twenty to forty lawyers up to a global enterprise. The vendor positions itself against law firm economics, saying the product is built around in house workflows rather than law firm billing models. Practice coverage falls into five named groups: contracts and commercial, legal operations, compliance and regulatory, real estate and facilities, and finance and treasury. Roughly two dozen applications sit beneath them, with industry pages for healthcare contracts, SaaS providers and vendor contracts. The contract sets one limit: section 4.03(c) says the Services are not designed for medical records, US Munitions List data, defense articles or ITAR related data. The eight jurisdictions covered by legal research are never named, and no practice area is marked unsupported. Nothing says which applications a nonlawyer may use, and the site gives the catalog size as more than 25 in one place and 22 in another.
ReviewPro supports legal departments and business teams handling recurring commercial agreements against approved positions. Standard playbooks cover common forms, and custom architecture lets a legal team encode its own preferred and fallback language. Industry offerings include healthcare, insurance, manufacturing, technology, higher education, staffing and law firms. A firm's plan matters because named seat subscriptions cover only its internal contracts, not an unlimited external client review service. Client work requires express authorization under a consumption subscription. The terms also restrict protected health information to an executed BAA and eligible features. The full set of supported jurisdictions and excluded agreement types is not defined.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The banner across DocJuris's site says customer data stays the customer's and never trains AI models. A product page qualifies this: the vendor refrains from using contract data to train general models and keeps contracts and playbooks private with enterprise grade encryption. The Subscription Terms contain no matching prohibition. Instead, section 4.02(f) has the Client grant DocJuris a royalty free, fully paid, worldwide, transferable, sublicensable, irrevocable and perpetual license to Client Data and Output as necessary or useful to enforce the Agreement, perform, improve or enhance the Services, and exercise its rights.
Section 4.02(a)(iv) makes data about the Client's use of the Services, used in aggregate and anonymized form, DocJuris's own intellectual property. Section 4.02(f) never names training or machine learning, and the license runs to the Services rather than the models, so the contract does not expressly reserve a right to train. The improvement right is perpetual and survives termination, and one page limits the no training statement to general models, leaving training specific to a customer unaddressed either way.
The ReviewPro terms grant LegalSifter a perpetual right to use anonymized customer content to train, tune and improve its own models and services. They prohibit using that content to train third party AI components. The security FAQ describes training as an agreement choice, but the standard agreement itself supplies the permission. This is a contractual right to reuse content, not evidence that any particular customer's contracts have already trained a model.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Uploaded contracts, prompts and generated output have no stated retention period. The Subscription Terms contain no retention period, no deletion commitment, no return of Client Data on termination and no export window; section 2.03(d) covers only payment of amounts owed on termination. No data processing addendum exists to fill the gap. The footer lists every legal document on the site (the privacy policy, these terms, a website disclaimer, a CCPA notice, an antibribery policy and an antislavery policy), and there is no DPA, subprocessor page or trust center.
The Privacy and Security page covers access control, single tenant isolation and audit logging and says nothing about how long anything is kept. The privacy policy, hosted at iubenda.com, covers website visitors. Section 4.02(f) grants a perpetual and irrevocable license over Client Data and Output, while nothing says when the underlying material is deleted.
Contracts are permanently deleted within 30 days after the subscription ends. Third party model handling is subject to separate retention restrictions, while the terms reserve continuing rights over anonymized training material and derived improvements. Neither the termination deadline nor processor handling establishes an ordinary lifetime for the customer's prompt and output history in ReviewPro. The current security position describes no configurable in service retention period and no zero history setting.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product runs its own documented permission model, which the customer has to keep aligned. Separation between customers is architectural: each gets an isolated single tenant deployment on its own subdomain with its own service stack. Inside the tenant, access runs through DocJuris's own permission groups, scoped down to individual repositories and applications, with invite only access when single sign on is off, worked examples of business users limited to intake forms and restricted workspaces, and audit logs of logins and permission changes.
The product does not inherit a document management system's access rules at query time, so the customer's administrators carry the work of keeping groups correct. Nothing describes separation between matters or deals inside one tenant, even though a legal department, its procurement team and its sales operations may all work in the same instance on opposing sides of related negotiations, and nothing says whether the permission model limits what the models can retrieve, as distinct from what a person can open.
ReviewPro isolates customer records and supports role and feature based access inside the application. Its single sign on connection delegates authentication to the organization's identity provider. Those controls protect a tenant and govern which functions a user can access. The current security explanation does not describe internal client matter walls or a retrieval connector that inherits per user permissions from a legal document management system.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
DocJuris's agreement commits to notice where lawfully permitted, and there is no transparency report. Subscription Terms section 4.01(e) says DocJuris will hold all Client documents and playbooks confidential and will not disclose them in a way that identifies the Client, for the term and five years after. If a government authority requests or requires disclosure, DocJuris will notify the Client where permissible and will cooperate with the Client to seek a protective order or other assurance.
Section 4.01(d) puts the same obligation on the Client for the vendor's confidential information. The commitment covers Client documents and playbooks, not all Client Data as defined in section 4.02(f), so what triggers notice is narrower than what the vendor holds. Section 2.03(b)(i) lets DocJuris suspend or end access, on prior written notice, when it receives a judicial or government demand, order, subpoena or law enforcement request. No transparency report, reporting schedule or aggregate figure is published.
The agreement requires notice before legally compelled disclosure unless the law prohibits it, allowing the customer to seek protection and participate in the response. Confidential information includes customer contract content, and the parties intend disclosure through the service not to waive privilege or work product protection. This is a protection for legal demands, separate from the promise to notify confirmed security or model integrity incidents. LegalSifter does not mention a government request transparency report.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Coverage is described by jurisdiction, and the sources are never named. Most of the product works from material the customer supplies: its own playbooks, fallback positions, precedent, templates and policies. The Legal Research application produces briefs with citations from primary law across eight jurisdictions; neither the jurisdictions nor the sources are named, so the extent of coverage is stated but not what is read or under what rights.
DocJuris also publishes more than fifty ready to use playbook templates for NDAs, MSAs, DPAs, procurement, employment and other agreements, each said to include fallback language and negotiation guidance built by legal experts, with no statement of where that drafting came from. Subscription Terms section 4.02(a)(ii) excludes standard or publicly available contract playbooks, terms and clauses from the vendor's own intellectual property, which implies the library draws on public material without saying so. No license, subscription or public domain basis is published for any source.
ReviewPro interprets a customer's contract against standard or custom playbooks containing legal positions and editing instructions. The July 2026 study identifies attorney approved edits as the reference for its NDA benchmark and varies the detail in those playbooks. These are concrete sources for reviewing an agreement. The product does not identify a licensed primary law collection or a complete account of the foundational training material behind every model it uses.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
Cited authority gets no check for later history in anything DocJuris ships. DocJuris ships a Legal Research application that returns briefs with citations drawn from primary law across eight jurisdictions, and a Trademark Clearance application that produces a DuPont risk rating from USPTO records. Nothing published describes a citator, a treatment signal, a good law check, how often the underlying law is updated, or even a prompt to verify what a brief cites.
The closest material is the general warning in Subscription Terms section 5.03(b) that Output may not accurately reflect facts or laws, with a direction to use human review. The Horizon Scanning application tracks regulatory change against the customer's contracts, which is not the same as checking cited authority.
ReviewPro connects review issues to contract language and the selected playbook, with redlines and explanations available in Word or Google Docs. Its accuracy study concerns issue detection and the fidelity of edits to an attorney approved position. A method for checking subsequent judicial treatment, a named licensed citator or a warning that an authority has ceased to be reliable is not part of the stated contract review workflow.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
Two DocJuris applications promise not to fill gaps. Document Generation returns a draft on the customer's own form with nothing silently invented, and Security Questionnaires answers SIG, CAIQ and custom grids from the customer's own policies with nothing invented. The word silently implies a gap is flagged rather than filled, though how it is flagged is not described. Guided Review builds the same behavior into the workflow, stopping at each position for the user to confirm, and is offered as an alternative to Quick Pass.
Subscription Terms section 5.03(b) states the general position: machine learning is probabilistic and may produce incorrect Output, and the Client should evaluate accuracy, including by human review. No confidence or grounding score is shown to the user, no uncertainty marking appears in a redline or brief, no rate of abstention or escalation is published, and no evaluation demonstrates the behavior.
The review process flags issues and proposes edits against a selected playbook. A person remains responsible for vetting that playbook and checking the returned changes, and Trial Features cannot be used in production. The study measures missed issues and incorrect actions rather than a per answer confidence display. A defined abstention response when the playbook provides insufficient support for a conclusion is not described.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
The AI Hallucination Cases database maintained by Damien Charlotin and two independent sanctions trackers record no court order, opinion or disciplinary record about fabricated or hallucinated legal citations produced by DocJuris. This describes the public record on that one subject. It is not a litigation history, and other proceedings involving the vendor would not appear here.
The AI Hallucination Cases database has no tool entry naming LegalSifter or ReviewPro as of October 15, 2026. ReviewPro's stated task is to identify contract issues and generate playbook guided redlines rather than prepare a court brief. That use context is separate from the database result. Neither the absence of a named court entry nor a contract benchmark establishes that every proposed legal edit is correct.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Bar and ethics guidance sits entirely outside DocJuris's product and terms. No bar opinion is named, ABA Formal Opinion 512 is not cited, no state guidance on generative AI is mentioned, and nothing maps any application to a jurisdiction's professional conduct rules. The Subscription Terms never mention privilege, professional conduct obligations or the rules that govern the Client's own use; the only obligations on the Client are commercial and security ones.
The site statement that DocJuris is not a law firm and cannot give advice is about the vendor's own status and engages no guidance. The product is sold to procurement and sales teams who are not lawyers and who build negotiating positions in it.
LegalSifter says it is not a law firm and does not provide legal advice. Its terms require customer verification, encourage legal counsel and distinguish internal seat subscriptions from authorized services for legal clients. Those are concrete responsibilities and scope limits. A named bar ethics opinion or jurisdiction specific professional rule governing the use of ReviewPro is not identified in these product and contractual instructions.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The buyer is a team that bills no client for the work, so no fee between lawyer and client turns on how fast the product runs. The vendor says repeatedly that it sells to in house legal departments and builds for in house workflows rather than law firm billing. It charges an annual platform fee sized to contract volume rather than metered work, and argues that law firm software is metered because law firms bill by the hour.
The savings claims, such as eight days to five minutes at Flex and over $300,000 saved at Purolator, concern the buyer's own costs and reach no client's bill. One application touches lawyer fees without changing this. The eBilling and Outside Counsel Spend application reviews outside counsel invoices against the customer's billing guidelines, with verdicts computed rather than guessed. That gives the client's view of someone else's invoice, not a record of what this vendor's AI did. Nothing published addresses disclosure of AI assisted work in any fee context.
LegalSifter compares the time and cost of automated review with conventional review, including a field account of a complex contract review falling from about 90 minutes to five. Firms may provide client services only under an expressly authorized consumption subscription. The subscription charge leaves open what a lawyer should bill a client. A client invoice disclosure method or a matter time record designed to allocate AI assisted work is not described.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
DocJuris publishes no subprocessor list, statement of model providers, data processing addendum or disclosure material for clients; the footer lists every legal document, and none of these is there. The only offer of documentation is on the Privacy and Security page, where booking a demo with a security team gets the architecture walked through and the documentation a security review needs, which is security material rather than a list of subprocessors and model providers.
DocJuris says only that it runs the best frontier model per task, so which AI providers process a customer's contracts is not stated. The SOC 2 Type II claim and the single tenant design are published but do not answer this question.
The processor list identifies OpenAI and Anthropic and their United States locations, while the agreement covers confidentiality, training permissions, incident response and restrictions on sensitive data. A law firm can use those details in responding to a client's AI requirements. They do not form a ready made client consent pack, and the default own model training permission remains relevant even though third party model training is prohibited.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
Court disclosure and certification of AI use have no dedicated support in the product. The product records some related facts: every proposed edit is a tracked change in a Word file that a named person accepts or rejects, internal comments are kept apart from what the counterparty sees, incoming markups are standardized across versions, and a screening report comes with each review. That shows who changed what in a negotiation, but it is not an AI disclosure record.
Nothing says that output identifies which model wrote a passage, which parts were machine generated or corrected by a person, which sources were retrieved, or who reviewed them. No disclosure template, guidance for courts or export built for that purpose is published. The Legal Research application's briefs across eight jurisdictions are the output most likely to reach a filing.
ReviewPro returns tracked edits and explanatory comments in the working agreement, preserving part of the reviewer's interaction with proposed changes. Its versioned repository and playbook management add context about contract work. LegalSifter describes no export combining the model version, all retrieved material and human verification for each generated passage. The editing record is therefore useful context for review but is not a complete court AI disclosure certificate.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- Good Law Verification
- Bar Guidance Alignment
Product updates
The latest logged changes from each vendor, newest first, each linked to its source.
DocJuris
-
Integration / interoperability
DocJuris released DocJuris in Word, a Microsoft Word add in side pane that connects the open document to the customer's DocJuris Library of playbooks, checklists, policies, templates and past agreements, with Ask and Edit modes for questions and requested changes.
Source
In summary
DocJuris
DocJuris, from DocJuris, Inc. of Bellaire, Texas, is a contract review and negotiation platform for in house legal departments. It screens incoming paper against a company's playbook and returns tracked changes in Word, alongside about two dozen task applications for procurement, legal operations, compliance, real estate and finance. According to the AI Legal Index, it runs each customer in a single tenant deployment, states SOC 2 Type II, and names Siemens, Flex, Ancestry and Purolator with results. It is sold as a yearly platform fee covering every app, or as a one time build of a single app with hosting and usage paid as you go.
LegalSifter
LegalSifter, of Pittsburgh, Pennsylvania, has operated since 2013. It sells ReviewPro, which reviews contracts inside Word or Google Docs against a playbook using more than 2,000 trained classifiers and applies generative redlines with one click. It also sells managed services and owns the Contract Logix platform. According to the AI Legal Index, LegalSifter publishes per contract pricing from $80, states issue spotting accuracy of 95 percent or better, and lets customers choose OpenAI or Google Gemini. It commits to breach notice within 24 hours. ReviewPro runs on AWS, and LegalSifter states its controls undergo independent SOC 2 Type II auditing.
Questions buyers ask
DocJuris vs LegalSifter: which is better for playbook contract review?
They are close, and each publishes more in a different area. DocJuris publishes more about isolation, named customer results and professional responsibility. LegalSifter publishes more about the AI itself, including an accuracy figure, its model providers and per contract prices. Teams that want to price a tool before talking to sales have more to read from LegalSifter. From the AI Legal Index, based on each vendor's own published materials as of September 25, 2026. No vendor pays for placement.
Does DocJuris train AI on customer contracts?
Its site says customer data never trains AI models, and one page limits that to general models. Its subscription terms contain no training prohibition, and grant DocJuris a perpetual and irrevocable license, which it may sublicense, to customer data and output to perform, improve or enhance its services, without naming training. LegalSifter's public pages do not address training, and its terms are a separate document. From the AI Legal Index, based on each vendor's own published materials as of September 25, 2026. No vendor pays for placement.
How much does LegalSifter cost?
LegalSifter prices ReviewPro per contract: $80 on Basic with one user and 30 or more reviews a year, $72 on Team with three users and 100 or more, and $65 on Enterprise with unlimited users, 240 or more reviews and single sign on. A contract counts as up to ten pages; longer ones cost more. DocJuris publishes no figure and sizes its fee to contract volume and applications. From the AI Legal Index, based on each vendor's own published materials as of September 25, 2026. No vendor pays for placement.
Which AI providers do DocJuris and LegalSifter use?
LegalSifter's Google Workspace listing says organizations choose whether OpenAI or Google Gemini processes their contracts, and its subprocessors page shows OpenAI. DocJuris says it picks a frontier model for each task. From the AI Legal Index, based on each vendor's own published materials as of September 25, 2026. No vendor pays for placement.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything on it comes from public material on the dates shown. How the index grades.
Three points to weigh. DocJuris's subscription terms grant it a perpetual, irrevocable license to customer data and output to improve its services. They also sell a set number of named users with a charge for more, and the terms are the binding text. LegalSifter's liability position sits in its separate terms and conditions. Its 95 percent accuracy figure comes without a test set or method. Neither vendor reviewed this page.