LinkSquares vs OnitX CLM: how they compare in 2026

LinkAI launched on 5 May 2026 as an agentic architecture that replaced the earlier approach. Independent review material describes it as a rebuild on a new architecture rather than a chat sidebar attached to the existing product. Extraction has been the core of the product from the start. Pulling payment deadlines, governing law, liability, renewal and commercial obligation terms out of signed agreements is a model task. The analytics layer the company is known for is built on that extracted output, so the models generate the data the rest of the product works with. The repository, workflow, approval routing and reporting still stand without the model layer. The agentic design is recent, dating from the May 2026 rebuild rather than from the product's origins.

ReviewAI carries the contract review work inside OnitX CLM. It reads a draft against the legal team's playbook, flags clauses that are missing or off position, proposes redlines and fallback language, and ends each review with a risk profile and recommendations. Generative redlining arrived in the first quarter of 2026, with rules written in plain language that add, replace or delete text at the word or phrase level. Smart Mapping, released in November 2025, lets customers write their own prompts to pull text fields and clauses, and a Chat feature answers questions about clauses and risks in the open document. The rest of the platform, from intake forms and approval routing to the repository, alerts, eSignature and Tableau reporting, runs without the models. Onit's release notes say AI features may need a license fee and services to enable.

LinkSquares publishes buyer guidance on evaluating AI contract management. It tells buyers to demand precision and recall benchmarks on their own documents and to keep review workflows for exceptions. LinkSquares publishes no precision or recall figure of its own, for any clause type, on any set of documents. Its comparison content claims extraction accuracy as a strength without a published measurement. It gives no error rate by clause type, no statement of behavior on nonstandard or ambiguous language, no confidence indicator on extracted values and no evaluation method. Extraction failures in this product are silent. A missed renewal date or auto renewal clause does not announce itself. It simply never appears in the report the legal team relies on.

ReviewAI grounds its review in the customer's own playbook and the document in front of it. Playbook rules fire on whether a concept or detail is present or absent. Each issue in the 2026 review panel links to the part of the document it concerns, so a reviewer can read the clause behind every flag. When a required clause is missing, ReviewAI says so and offers the playbook's language, as in Onit's June 2026 walkthrough, where a missing statement is detected and inserted from the playbook. Onit publishes no accuracy rate, error rate or evaluation for review, extraction or Chat. Its Product Specific Terms say results may contain material inaccuracies even if they appear accurate, and tell customers not to rely on them without confirming them independently. Section 12.3 of the services agreement says output may be incomplete, inaccurate or biased.

LinkSquares positions itself as agentic CLM and states that the platform handles the bulk of standard contracts automatically. Its own buyer guidance recommends maintaining review workflows for exceptions. For its own product, it does not publish which contract types or value thresholds can complete without human review, or what makes a contract standard enough to be handled automatically. It does not say whether an agent can send or approve a contract unattended, or what confidence threshold routes an exception to a person. Nor does it say what the audit record shows when an agent rather than a lawyer made the call.

ReviewAI proposes and the user decides. Suggested redlines appear as comments in Word, and a change enters the document only when the reviewer ticks it and chooses Action Selected Changes. Playbook rules can also raise an alert, run a pre check, add a comment or escalate a contract through the OnitX CLM workflow, so a deviation can route to an approver instead of being fixed in place. Onit's guidance on wrapping up a review tells users to switch off highlights before a document goes to the other side. The Product Specific Terms tell customers not to rely on ReviewAI results without confirming them and to take their own legal advice. Section 12.3 of the services agreement makes the customer solely responsible for review of and reliance on output. Onit does not publish which rules act without a person, or what follows when a wrong redline is accepted.

LinkSquares names DraftKings, TIME, ProPharma and Asurion as customers and states more than 1,200 customers in all. G2 has named it a Leader in contract lifecycle management for five years running, including the Winter 2026 Grid Report. On G2, 98 percent of users are reported as saying the product is moving in the right direction. A case study with Softonic, a named customer, reports outside counsel costs cut by 40 percent. The case study says NDAs used to take three to four business days, and that sales reps using templates now spend on average five times less time executing them. A LinkSquares comparison guide describes the same result as NDA processing time cut by nearly 400 percent. The vendor also states a 360 percent return on investment over three years, without a method, sample or baseline.

Wesfarmers uses ReviewAI on SaaS agreements and plans to extend it to NDAs, service agreements and consulting contracts. Its case study quotes Sheldon Renkema, who says the product sells itself, and gives no figures or date. Medtronic runs Onit ELM and CLM on separate instances, and VMware runs contract workflows through Onit's Unity platform, both without figures. The ReviewAI page claims reviews 60 to 70 percent faster, a 51.5 percent productivity gain per lawyer, more than two hours saved a day and NDAs returned in two minutes, with no customer named. Those figures trace to an Onit study described in a June 2020 blog, in which 19 lawyers did the work of 28 and the cost per contract fell from $592 to $395.50, six years before generative redlining shipped. The OnitX CLM page shows logos for Airbnb, Experian, Sonoco and Delaware North without describing their use.

A dedicated security page publishes SOC 2 Type II, ISO 27001 and GDPR compliance and encryption of all contracts and data. The vendor says customer data stays private within the platform. Independent review material, a practitioner view rather than vendor material, describes the residency and retention controls as mature enough that a public company legal operations lead need not negotiate addenda. Nothing published addresses legal professional privilege or attorney work product. Negotiation playbooks and clause libraries are the closest thing to work product in this product. The security material does not cover their confidentiality, though the vendor says elsewhere that they are isolated. Legal, sales, procurement, finance and HR work in the same instance, and nothing published describes the line between legal's own analysis and the wider business.

Onit's Subscription and Services Agreement, last revised 19 August 2026, governs OnitX CLM as it does Onit's other products. The customer owns its data (5.1), and Onit treats it as confidential, with reasonable advance notice before a disclosure the law compels (8.5). Section 12.1 bars third party model providers from training on customer input, except in deidentified or aggregated form needed to run the service, and lets Onit use deidentified, aggregated usage data to improve its AI. Section 5.2 allows aggregated and anonymized customer data to be used for product improvement, analysis and benchmarking. A 2023 help center FAQ says the contract review engine comes pre trained and does not learn from customer interactions. The agreement, data processing addendum and privacy policy do not mention privilege or work product, and no statement covers what Anthropic keeps from a review.

The product drafts, reviews and approves contracts and is sold for use by sales, procurement, HR, marketing and finance as well as legal. Staff who are not lawyers act on machine generated contract analysis by design, and the vendor presents that as a feature. LinkSquares publishes nothing on the line between the tool and legal advice. No statement says output is not legal advice, and no guidance says when a contract should reach a lawyer. Nothing covers in house counsel's supervision of business users who write contracts themselves, and nothing engages with bar guidance.

Section 4.14 of Onit's services agreement states that Onit is not a law firm and does not offer legal advice, recommendations, opinions, representation, referrals or counseling. The ReviewAI section of the Product Specific Terms tells the customer to obtain its own legal counsel to understand ReviewAI output, and section 12.7 of the agreement restricts uses classed as high risk under applicable law. ReviewAI applies a playbook that the customer's own lawyers write and can update in house without Onit's services, and it presents its output as proposed redlines and recommendations for a reviewer. OnitX CLM is sold to legal departments and to the business users who raise and track contracts through its intake forms. Nothing published ties the product to a lawyer's competence or supervision duties when using AI in negotiation, or sets limits by jurisdiction.

The LinkSquares buyer guide recommends quarterly AI governance reviews to monitor bias, validate accuracy and enforce compliance. It lists model governance covering audit trails, PII handling and zero retention options as things to expect from a vendor. That guidance is written for customers. For LinkAI itself, the published material says only that it is designed with enterprise level data security and governance built in. The vendor publishes no AI policy, model card, bias or fairness testing, evaluation method or result, accuracy monitoring, drift statement or named governance body. The company holds ISO 27001 but has no ISO 42001 certification.

Onit publishes no AI governance material for OnitX CLM or ReviewAI: no responsible AI page, named owner for model behavior, release testing process, model card or AI management certification. The AI terms in the services agreement allocate risk instead. Section 12.3 warns that output may be biased. Section 12.4 lets Onit improve, replace or deprecate models or prompts as long as core functionality is not materially reduced. Section 12.7 restricts high risk uses, training other AI, entering sensitive data without approval and benchmarking for publication. Onit's SOC 2 Type 2 audit covers system controls rather than model behavior. Nothing addresses whether review quality differs across contract types, playbooks or the languages Chat translates, translation that Onit's terms call uncertified.

The vendor states that contract data is never shared with or used to train any third party LLM providers. It also says contract data and proprietary playbooks are completely isolated and never used to train public AI models, and that data stays completely private within the LinkSquares platform. Those statements cover the main risk for a contract lifecycle product, which is customer contract text reaching an outside model provider. Both are limited to third party and public models. Neither says LinkSquares does not use customer contract data to train or tune its own models inside its own platform. Data staying completely private within the platform is consistent with internal training rather than ruling it out. No retention position for AI processed content is published. The vendor's own buyer guide lists zero retention options as something to demand from a vendor, without saying whether this product offers one.

Onit's subprocessor list, updated 17 August 2026, gives OnitX CLM and ReviewAI their own table. It names Anthropic for generative AI in ReviewAI and Catalyst for Contracts, AWS for hosting, Adobe for PDF processing, and Google Analytics, Sentry and Segment for analytics and error reporting, each with a location. Customers subscribe to email notice of changes and have 15 days to object under section 4.5 of the services agreement, with termination of the affected service as the remedy. The data processing addendum, revised 20 April 2026, commits Onit to notify a customer promptly of a security breach, without a period in hours or days. After termination a customer has 30 days to export its data, which is then deleted under a retention policy Onit does not publish (6.5). The security page describes encryption in transit, daily backups and logged system access, and the product page lists role based permissions, audit trails and retention policies.

LinkSquares publishes no terms page. Its enterprise agreements govern liability and are not public. Nothing published covers liability for AI output, warranties, service levels or remedies when an extraction is wrong, a clause is missed, or an agentic action moves forward a contract that should have been escalated. The vendor says the platform handles the bulk of standard contracts automatically. A missed auto renewal or liability cap in a signed agreement is a loss the customer finds later, and the vendor publishes no position on it.

Onit's Subscription and Services Agreement, last revised 19 August 2026, warrants that the services operate substantially as described in the documentation and that material functionality will not decrease during the term (9.2). Onit defends and indemnifies the customer against claims that the services infringe intellectual property rights (11.1). Each party's liability is capped at the fees paid or payable under the order in the 12 months before the event (10.2). Indemnities and payment obligations sit outside the cap (10.3), and there is no separate carve out for a data breach. AI features and output are provided as is and as available (12.5), implied warranties of accuracy are disclaimed (9.5), and no service levels or uptime commitments apply to AI features unless an order says so (12.8). The Product Specific Terms separately exclude ReviewAI and the Word add ins from availability service levels. No insurance is published.

LinkSquares integrates with Microsoft Word, Google Docs, Microsoft 365, Salesforce, Slack, DocuSign and HubSpot. Drafting runs natively in Word and Google Docs. The vendor says clause libraries and playbooks are available inside the drafting environment, so users do not switch tabs or upload files. The CRM and electronic signature connections cover the path to signature end to end. Independent review material says there is no generally available API for developers, which limits moving contract data into systems the vendor has no connector for. No document management system such as iManage or NetDocuments is named.

OnitX CLM ships integrations with Salesforce, DocuSign and Adobe Sign, native Tableau dashboards, and APIs that Onit describes as supporting two way data flows, licensed under section 4.4 of the services agreement. Onit announced Salesforce AppExchange and SAP Ariba connections in December 2022. The Word add in carries the most detail. It handles redlining, versioning and clause insertion, runs ReviewAI on contracts started in the CLM, and writes ReviewAI findings back to the CLM record as structured data. Onit's help center says the add in needs desktop Word 2016 Pro Plus, Word 2019 or Microsoft 365 and does not work in Word online, with a separate manifest file install for Windows and Mac. In June 2026 Onit added a two way ReviewAI connection to ContractWorks Premium. No OnitX CLM page names iManage, NetDocuments or SharePoint, or describes what the Salesforce connection syncs.

Independent review material says LinkSquares' data residency and retention controls are mature enough that a public company legal operations lead does not have to negotiate addenda. That is a practitioner's account, not vendor material. No vendor page names a hosting provider, lists regions, states a residency commitment or describes single tenant or private deployment options.

OnitX CLM is a cloud service. Onit's subprocessor list, updated 17 August 2026, shows OnitX CLM and ReviewAI hosted on AWS in the United States or Germany, with backups in Ireland, so a European customer can keep its contracts in Germany. Anthropic, which runs ReviewAI's generative features, is listed in the United States and various other locations. OpenAI, listed for all Onit platforms, runs in the United States and worldwide, so review content may be processed outside the hosting region. Onit's firewall guidance lists separate European endpoints for its drafting service. The data processing addendum covers transfers under standard contractual clauses, with a 10 day notice period for new subprocessors in those clauses. No tenancy model, private cloud or on premises option is published, and the Word add in requires a desktop installation.

A dedicated security page, open without a login, states SOC 2 Type II and ISO 27001 alongside GDPR compliance and encryption of all contracts and data. Independent review material confirms both certifications and describes them as the two that clear most enterprise procurement checks. The page names no auditing firm for either. It gives no certificate date, examination period or expiry. LinkSquares publishes no trust center or route to request documentation, so there is no stated path to the reports themselves.

Onit's security page says the company audits to SSAE 18 SOC 2 Type 2, SOC 1 Type 2 and HIPAA standards, verified by an independent auditing firm, and names no auditor, report period or scope. Onit's data processing addendum lets customers rely on SSAE 18 reports issued in the prior 12 months in place of an on site audit, which makes the reports available to customers once they sign. Annex 2 of the addendum commits to ISO 27001 certified hosting providers rather than stating an ISO certificate for Onit itself. Onit also runs a trust portal at trust.onit.com. The OnitX CLM page lists single sign on, multifactor authentication, role based permissions and audit trails, and the security page describes 256 bit encryption in transit, daily backups and logged system access.

LinkSquares commits that contract data is never shared with third party LLM providers, which implies such providers are part of the architecture. It names none of them. It publishes no provider, model family or version and no subprocessor list. Nor does it say what runs on its own infrastructure and what runs on an external model. The product was rebuilt on an agentic architecture, and no published material identifies the models underneath it.

Onit's subprocessor list, updated 17 August 2026, names Anthropic as the generative AI provider for ReviewAI and Catalyst for Contracts, in the United States and other places. It also lists OpenAI across all Onit platforms and Google Gemini for Onit's spend and agent products. A 2023 help article records that Precedent ReviewAI was renamed Catalyst Contracts Review, and the review engine traces back to McCarthyFinch, which Onit acquired in November 2020. Customers who subscribe to the list get email notice of a new provider and 15 days to object (4.5). Section 12.4 of the services agreement lets Onit improve, replace or deprecate models or prompts as long as core functionality is not materially reduced. Section 12.9 commits Onit to commercially reasonable efforts to substitute comparable services when a provider fails. No model or model version is named for ReviewAI or Chat.

LinkSquares publishes no price, range, tier structure or unit of charge. Its own buyer guidance tells readers to compare user based, usage based and record based pricing models without saying which one LinkSquares uses. A third party estimate puts the median at around $31,000 a year, with a starter tier near $10,000 and enterprise above $75,000. Those figures are independent estimates, not vendor disclosure. If the estimate is accurate, the product is out of reach for small teams.

Onit publishes no price for OnitX CLM or ReviewAI, and every commercial route on its site is a demo request. Its Product Specific Terms, last updated 6 May 2026, set out part of the structure. The Onit and OnitX section includes 1 TB of storage and charges $600 a month for each additional TB, with limits of 250 MB per attachment and 80,000 Excel rows. Catalyst Contract Review is sold in three volume tiers: Silver at 1,000 contracts and 25,000 chats, Gold at 5,000 and 125,000, and Platinum at 10,000 and 250,000, with no tier prices. Chat use is metered against the tier, Onit says usage averages 25 chats per contract, and heavier use can bring additional fees. Bespoke ReviewAI configuration or training is out of scope unless ordered, and release notes say AI features may need a license fee and services to enable.

LinkSquares describes its coverage by clause and lifecycle scope rather than by jurisdiction. The named terms are payment deadlines, governing law, liability clauses, renewal language and commercial obligations. Batch analysis runs across large legacy portfolios for M&A due diligence and repository audits. The lifecycle runs from before signature to after it in one platform. Users extend beyond legal to sales, procurement, finance, HR and marketing. The vendor states more than 1,200 customers in the in house market. It publishes no list of supported contract types or languages and no indication of extraction depth by agreement type. Governing law extraction is a named capability, but no jurisdictional scope is given for it. Independent review notes there is no primary law research capability, which is typical of the product category.

OnitX CLM is sold to enterprise legal departments and the business users who request, approve and track contracts, and Onit pictures legal operations, general counsel and revenue operations as its users. Onit's ReviewAI material lists dozens of contract types handled out of the box, including NDAs, master services agreements, statements of work, and purchase, lease, employment and construction agreements. Onit's terms set limits. Chat performance is reduced on contracts over 75 pages or 75,000 words, translation is uncertified, and review works on Word documents, with PDFs converted to Word since 2025. A 2023 FAQ listed German, French and Portuguese in beta for one review skillset. Law firm use is not addressed and no industry focus is stated. The case studies come from a diversified conglomerate, a medical device maker, a software company and an agribusiness.