CaseBlink vs Parley: how they compare in 2026
CaseBlink and Parley both began as AI platforms for immigration lawyers, drafting petitions, support letters and RFE responses from uploaded evidence and completing USCIS forms. CaseBlink sits in the top two bands on thirteen of fifteen axes and Parley on ten of fifteen. CaseBlink's lead is proof about where client files go. It names OpenAI and Google as its processors, keeps processing in the United States on Google Cloud, and reports SOC 2 Type 1 and Type 2 audits in 2024 and 2025. Parley's home page shows a SOC 2 Type II label, while its privacy policy says only that its security program aligns with SOC 2 controls, and it names no model provider on any readable page. Parley's counterweight is where it works and what it tells lawyers. Its agent runs inside Outlook and Word, builds a client and matter record from email and calendar, and its published guide applies ABA Formal Opinion 512 to AI use. CaseBlink names no bar guidance.
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.
The models do the core work, which is the A band. CaseBlink's product is AI case preparation: models label and organise client documents, extract data into USCIS forms, summarise evidence, run immigration research and draft support letters, briefs and RFE responses, and the Autopilot agent runs the whole sequence from upload to assembled packet. Only the entry Lite plan (packet assembly and exhibit lists) works without the AI features. Verified 18 September 2026.
The models drive what the vendor leads with, on a workflow layer that would still run without them, which is the B band. The Terms of Service define the platform as an AI-native system that ingests and stores content and applies AI to extract information, generate drafts, populate records and execute agentic tasks, and the home page describes an agent that drafts, fills forms and researches from a firm's own templates and past work, over a client and matter database that fills itself from email and calendar. Around that sits conventional practice software: case tracking with USCIS status updates, exhibit assembly into a combined, paginated PDF, form completion screens and LCA and PAF management. Strip out the models and a firm keeps a case tracker and an exhibit tool, which is why this is B rather than A. Verified 18 September 2026.
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.
Grounding is real and described, short of any accuracy measurement, which is the B band. CaseBlink says every output cites directly to the client's uploaded evidence so the attorney can verify it, that its research draws on AAO decisions, the USCIS Policy Manual and case law, and that it reviews uploaded documents for missing context and quality gaps. No accuracy or error rates are published, and the Service Agreement warns that AI output may contain errors or misleading information. Verified 18 September 2026.
Grounding to the user's evidence is claimed without a described method or measurement, which is the C band. The vendor's content subdomain says drafts are generated directly from uploaded evidence, tie statements to verifiable facts and insert direct quotations from reference letters; the main site says the agent works from a firm's own templates and past samples. No accuracy figure, test set or evaluation is published, no failure mode is named, and nothing describes how a draft links a statement back to the document it came from or what the research function does when it finds no support. The canonical documents run the other way: the privacy policy says AI output may be inaccurate or incomplete and must be reviewed by a qualified human, and the terms disclaim any warranty that generated content is accurate. The vendor's own practice guide discusses hallucination risk at length as a professional duty, which is credited on the professional rows, not here. Verified 18 September 2026.
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.
Human review before filing is the stated model, with citations as the review surface, short of thresholds or stopping points, which is the B band. Autopilot runs document organisation, research, drafting and packet assembly in one pass and then hands the attorney a complete packet to review before filing; the page for individual applicants says the attorney prepares the case using CaseBlink under their supervision. The Service Agreement says CaseBlink may review output before it is returned but does not guarantee it, and puts the risk of relying on output on the customer. No confidence signal, checkpoint inside the Autopilot run or route for flagged errors is published. Verified 18 September 2026.
A written requirement that a qualified human reviews output, with real review surfaces, short of a control structure, which is the B band. The privacy policy states that a qualified human user must review AI-generated output before relying on or using it, the terms say the services are tools intended to assist and are not a substitute for professional judgment and require the customer to set up independent verification procedures, and the vendor's practice guide lays out a draft, verify and edit, then file sequence. Drafts land in Microsoft Word and Outlook where the lawyer edits them. What the A band asks for is missing, and one gap is specific to this product: the home page says the client and matter database automatically updates and fills out fields from emails, meetings and uploaded documents, and nothing states whether those record changes pass a review point. No mode, threshold or categorical limit on what the agent may do alone is published. Verified 18 September 2026.
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.
Named customers and usage figures, without outcome measurement, which is the B band. CaseBlink shows logos of firms including Dorsey & Whitney, Dickinson Wright, Fakhoury Global Immigration, Clark Lau and Cline Williams, names twelve partner firms by city on its applicant page, and reports more than 100 firms, 1,000 users and 3,000 cases prepared (March 2025: more than 50 firms and 1,000 cases). Testimonials are named customer evidence, not outcome evidence (R122), and the 'five times faster filing' figure has no method behind it. Verified 18 September 2026.
Named customers without figures, which is the B band. The about page names Erickson Immigration Group as relying on the platform, pointing to Business Insider reporting in which a partner there describes using it for EB-1, O-1 and national interest waiver support letters, and it names Boundless as a partner using the platform with its legal team. The home page says the product is trusted by more than 100 legal teams. No figure is published for what changed at any named firm, the 100 teams are not identified, and no case study with a method was located on the main site; productivity benchmarks appear only on the vendor's content subdomain without a named customer. Verified 18 September 2026.
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.
Confidentiality and zero retention with AI providers are in the published agreement, but the licence to use customer data is broad and privilege is not named, which is the B band. The Service Agreement (15 July 2026) says third-party AI tools acting for CaseBlink retain no customer data, requires written notice before any compelled disclosure, and its DPA requires deletion within 90 days of termination and lists OpenAI and Google as subprocessors; the security page adds isolated tenant environments and says client content is never used for model training. Against that, section 1.4 licenses CaseBlink to use customer data to improve and develop its products, and neither the agreement nor the security page addresses privilege or work product. Verified 18 September 2026.
Substantive contractual commitments on training and confidentiality, short of segregation and the model provider position, which is the B band. The Terms of Service limit Parley's licence to customer content to providing the services and testing, diagnostics and troubleshooting, and state that Parley shall not use customer or generated content to train, fine-tune or develop models deployed for or accessible by other customers (3.3). Customer content is defined as the customer's confidential information, protected by a confidentiality clause that requires Parley to inform the customer immediately in writing before any compelled disclosure (6.1). The pricing page lists zero-data retention and no model training on the Starter plan. The vendor's own practice guide discusses attorney-client privilege and work product at length, including United States v. Heppner and Warner v. Gilbarco, but it does not state how Parley's own terms preserve either for a firm using it. Three things keep this off A: no segregation between users or matters is documented, the model providers behind the product are not named in any readable document, and the training bar is scoped to models other customers can use, leaving customer-specific training open. Verified 18 September 2026.
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.
A clear position that the tool is not legal advice, with the attorney's role stated, short of jurisdiction limits, which is the B band. The Service Agreement says content is not legal advice, creates no attorney-client relationship and should not be acted on without a competent professional, and describes the service as a tool for legal professionals. CaseBlink also runs a consumer page that takes applicants' details and matches them with partner immigration firms; it says the attorney prepares the case under their supervision but carries no not-legal-advice statement and does not say whether partner firms pay for referrals. Flagged for sampling. Verified 18 September 2026.
A clear tool statement, a defined audience and a published treatment of competence and supervision, short of jurisdiction limits, which is the B band. The terms state that the services and generated content are tools intended only to assist and are not a substitute for professional judgment, and define authorized users as the customer's employees such as attorneys and paralegals, barring use to serve other firms. The vendor's guide Should you use AI as a legal practitioner?, updated 14 July 2026, walks through ABA Formal Opinion 512 on competence, confidentiality, verification and client disclosure, lists what attorneys remain responsible for including supervision of junior lawyers and non-lawyer staff using AI, and points to state bar and court-order trackers. What is missing for A is the jurisdictional boundary: nothing states which jurisdictions or practice areas the product is built for or where it stops, and the guide tells lawyers to check state rules rather than stating how the product fits them. Verified 18 September 2026.
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.
Limitations are acknowledged without a governance programme, which is the C band. The Service Agreement lists the limitations of AI output, including errors, bias from training data and weak reasoning on complex tasks, and makes the customer responsible for any bias audits, notices or consents required where AI processes personal data. No internal owner, testing before release, bias assessment or evaluation of the immigration models is published. Verified 18 September 2026.
Individual AI commitments are published without a governance framework, which is the C band. The privacy policy has a dedicated section on artificial intelligence and automated processing: it describes what the AI does, states that output must be reviewed by a qualified human, commits not to use customer content to train models available to other customers, and states that Parley makes no decisions with legal or similarly significant effects based solely on automated processing. Those are real commitments. Nothing names who inside the company owns model behaviour, what is tested before a release, or what has been found about uneven output across case types or applicants, which matters on a product drafting immigration petitions for people of many nationalities. A bias-minimisation claim appears only on the content subdomain, without a mechanism. Verified 18 September 2026.
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.
Retention, deletion, access, subprocessors and incident practice are published and specific enough to hold the vendor to, which is the A band. The Service Agreement binds CaseBlink's AI providers to retain no customer data; the DPA requires deletion within 90 days of termination, breach notification without undue delay with remediation and assistance, a named subprocessor list (Google, OpenAI, Slack, all processing in the US) with 15 days' notice of changes and a right to object and terminate, and annual audit rights satisfied by a SOC 2 report. The security page adds AES-256 encryption, isolated tenants, SSO, role-based access and audit logs for sensitive operations. The breach clause sets no fixed hour window. Flagged for sampling. Verified 18 September 2026.
Retention, deletion, security and incident practice are published, with the subprocessor list unreadable to this index, which is the B band. The privacy policy, updated 8 September 2026, states retention periods by category of information, ties customer content retention and deletion to the customer agreement and the product's account-deletion and download functions, and records that customers may request their content for 30 days after a subscription ends, after which Parley may delete it (terms 2.4). It describes encryption in transit and at rest, access controls and security assessments, and has a dedicated incident section committing to prompt investigation and to notify customers under the customer agreement or terms. It states that the current subprocessor list is published at trust.parley.so and that customers will get notice of new subprocessors that touch their content as the agreement requires. That list returns no content to automated retrieval and could not be read on this date, so the top band is not asserted on it; it is the surface most likely to move this row. The Data Processing Addendum the policy refers to is not published on the estate. Verified 18 September 2026.
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.
A real published position with a cap and super-cap, and an express exclusion of AI output, which is the B band. The Service Agreement caps liability at 12 months of fees, raises the cap to twice that for breaches of data security or confidentiality involving customer data, and indemnifies the customer for claims that the service infringes a US copyright or trade secret, but excludes output from that indemnity, disclaims any warranty of the accuracy of output, and requires the customer to indemnify CaseBlink for claims arising from its use of output. Verified 18 September 2026.
A published, specific liability position that stands behind nothing on output accuracy, which is the B band. The Terms of Service give a warranty that services will be performed professionally with reperformance as the remedy (7.1), an indemnity against third-party claims that the services violate law or infringe copyright, trademark or US patent, or arise from Parley's fraud, gross negligence or wilful misconduct (8.1), and a cap at fees paid in the prior twelve months, or one hundred dollars during a free trial (9). Against that, the terms disclaim any warranty that generated content is accurate, complete or compliant, exclude generated content from the indemnity, and state that Parley has no liability for the accuracy of generated content or results obtained from using it. No insurance position is published. Verified 18 September 2026.
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.
Connections are thin or unnamed, which is the C band. CaseBlink exports to Word and PDF, supports single sign-on through SAML 2.0 and OIDC with Okta, Azure AD and Google Workspace, and the Service Agreement mentions an API; the Pro plan (in beta) builds case management and a client portal into CaseBlink itself, and Enterprise offers custom integrations. No integration with an immigration case management system, document management system or other practice tool is named. Verified 18 September 2026.
Named integrations with some description of what they move, short of documented depth, which is the B band. The home page names Microsoft Outlook and Word, where users delegate work to the agent, and Google Drive, Box and Dropbox for documents, and says connecting email and calendar populates the client and matter database from emails and meetings. The terms define a Microsoft Word plug-in and a Parley API, which Parley may make available at its discretion. The content subdomain describes a USCIS case status API in beta. No document management system such as iManage or NetDocuments is named, no immigration case management system is named, and no implementation documentation describing direction of sync or configuration was located. Verified 18 September 2026.
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.
Hosting and options are described in real terms, short of region choices for standard plans, which is the B band. CaseBlink runs on Google Cloud Platform in isolated tenant environments; its subprocessor list puts all processing, including OpenAI, in the United States; and the Enterprise plan offers dedicated or private hosting and data residency. Which regions residency covers is not published. Verified 18 September 2026.
Cloud hosting is stated with neither a tenancy model nor a residency option, which places this at C. The terms describe a cloud-based platform that Parley hosts and gives the customer internet access to, and the privacy policy says Parley is based in the United States and may process personal information in the United States and other countries where it, its affiliates or subprocessors operate, relying on the EU-U.S. Data Privacy Framework for transfers. No tenancy model is stated, no region is offered, and processing location is not separated from storage. A private cloud deployment option is claimed only on the vendor's content subdomain and is not relied on here. Verified 18 September 2026.
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.
A named, dated attestation with reports available on request, short of evidence a reader can open, which is the B band. CaseBlink says it completed a SOC 2 Type 1 audit in 2024 and a SOC 2 Type 2 audit in 2025, audits annually, runs regular third-party penetration tests and a vulnerability disclosure programme, and lists 52 controls; the audit reports and its zero-data-retention agreement with OpenAI are available on request through trust.caseblink.com, which renders only in a browser and could not be read (R128). Verified 18 September 2026.
A SOC 2 Type II label appears on the home page while the governing document describes alignment rather than attestation, which places this at C. The home page security section lists SOC 2 Type II, GDPR and no AI model training. The privacy policy, updated 8 September 2026, says the security program is designed to align with industry standards including SOC 2 Type II controls, which is not a statement that an audit has been completed. The vendor's content subdomain says an auditor-issued report is available under NDA on request and that the auditor and control period are disclosed only with it. The trust centre at trust.parley.so returns no content to automated retrieval and could not be read on this date. No auditor, report period or scope is published on any readable surface. The discrepancy between the label and the policy is the finding. Verified 18 September 2026.
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.
The model provider is named and changes are contractually notified with a right to object, which meets the A band. The DPA's subprocessor list names OpenAI OpCo, LLC as the third-party AI deployment provider, processing in the United States, and requires 15 days' notice before any new subprocessor can access customer data, with a right to object and to terminate if the objection is not resolved; the security page describes a zero-data-retention agreement with OpenAI, available on request. The specific models and versions are not named. Flagged for sampling. Verified 18 September 2026.
Third-party models are acknowledged and a change process is described, with no model or provider named on any readable surface, which places this at C. The terms define third-party materials to include AI models used with the service, and the privacy policy commits to notice when a new subprocessor will touch customer content, as the customer agreement requires. The vendor's data protection guide explains how model providers, subprocessors and cloud hosts handle data under zero-data-retention agreements, and the pricing page lists zero-data retention, but neither says which providers Parley uses. The current subprocessor list at trust.parley.so, where providers would be named, could not be read on this date and is the surface most likely to move this row. Verified 18 September 2026.
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.
Prices are published, which is the A band. The pricing page lists Lite at $100 per seat per month (packet assembly), Core at $200 (AI drafting, case preparation and the Autopilot agent) and Pro at $300 (case management, in beta), each with unlimited cases and a 14-day free trial, with one month free on annual billing; Enterprise is custom. The Service Agreement sets out the per-seat reasonable-use limits, trial cancellation, renewal and fee-change terms. Verified 18 September 2026.
Real pricing is published for the entry plan with the enterprise tier withheld, which is the B band. The pricing page lists Starter at $500 a month for up to ten seats in a shared workspace with a set amount of usage credits, agentic drafting, research, forms and review, Outlook and Word integrations and zero-data retention, and Enterprise on request with unlimited seats, credits sized to the firm, SSO and deployment support. The Terms of Service publish the billing mechanics in detail: monthly billing in advance or annual prepayment, included credits, additional credits at a list price set on the order, rollover of unused credits capped at twice the monthly allotment, automatic renewal, a free trial that converts to a paid plan unless cancelled, and 1.5 per cent monthly late charges. What is missing for A is the credit list price and anything on what implementation adds. A pricing row is written. Verified 18 September 2026.
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.
Segments and matter types are set out in detail, short of stated limits on the AI, which is the B band. CaseBlink covers US employment visas (H-1B, L-1, O-1, TN, E-3), green cards (EB-1A, EB-1B, EB-1C, EB-2 NIW, PERM, EB-3), family-based petitions, humanitarian cases (asylum, DACA, TPS, U and T visas) and naturalisation, for solo practitioners through large firms and corporate immigration departments. It is built only for US immigration law; where drafting or research is weaker by case type is not stated. Verified 18 September 2026.
Segments and immigration workflows are described with substance, short of the boundaries, which is the B band. The pricing page separates solo practitioners and small teams on Starter from firms needing governance and SSO on Enterprise, and the terms define users as attorneys and paralegals. The support page lists immigration workflows: research and analysis, form filling, drafting, case tracking, LCA and PAF management and RFE responses, and invites immigration practitioners to book a demo. The home page has broadened to legal teams generally without naming other practice areas, so the scope beyond immigration is claimed rather than evidenced. The content subdomain lists visa categories, including O-1A, EB-1, EB-2 NIW, H-1B, TN, L-1 and E-2. Nothing states which case types or practice areas the product does not handle. Verified 18 September 2026.
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 security page says client data is never stored or trained on and that there is no model training on client content. No matching prohibition is in the Service Agreement (15 July 2026), whose section 1.4 licenses CaseBlink to use customer data to improve and develop its products and services; the agreement does bind third-party AI providers to retain no customer data.
The published Terms of Service prohibit training on customer content, within a stated scope. Section 3.3 limits Parley's license to providing the services and testing, diagnostics and troubleshooting, and bars using customer or generated content to train, fine-tune or develop AI models deployed for or accessible by other customers. Training a model used only for that customer is not addressed by the clause, and the privacy policy adds that cross-customer use may occur where a customer authorizes it in a separate agreement. The pricing page lists no AI model training on the Starter plan.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Fixed terms are published. The Service Agreement says third-party AI providers acting for CaseBlink retain no customer data, the security page says data is purged after each AI operation, and the DPA requires deletion of customer personal data within 90 days of termination. Case files, including AI output, are kept in the account for the life of the subscription.
The customer controls deletion of stored content through the product, with no zero-retention setting stated for the platform itself. The privacy policy says customer content is retained and deleted under the customer agreement and the product's account-deletion and download functions, and the terms let a customer request its content during the subscription and for 30 days after, after which Parley may delete it. Prompts and activity are listed among service information retained for operating the service.
The pricing page advertises zero-data retention, which the vendor's own guide describes as an arrangement with model providers that process prompts in memory; the platform itself stores content by design.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Segregation is claimed at tenant level without matter-level detail. The security page lists isolated tenant environments and role-based fine-grained access controls, and the Pro plan adds team management; nothing describes restricting access to particular cases or clients within a firm.
Separation is asserted at workspace and role level with no published detail on how it is enforced. The pricing page offers shared workspaces on Starter and team workspaces with centralized administration on Enterprise, and the vendor's content subdomain describes role-based access controls. Nothing addresses walls between matters or clients within a firm, or whether access follows the permissions of a connected document store.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
The Service Agreement commits CaseBlink to give written notice before any disclosure required by subpoena or order, to let the customer intervene, and to cooperate in seeking protection; the DPA commits to notify the customer of government access requests. No transparency report is published.
The Terms of Service commit to notice before compelled disclosure. Section 6.1 permits disclosure of confidential information, which includes customer content, only in response to a valid order or as required by law, requires the receiving party to inform the other immediately in writing of the request and its circumstances, and limits disclosure to what counsel advises is legally required. The duty is mutual and carries no lawful-prohibition qualifier.
The privacy policy separately says Parley may disclose personal information to government authorities where required by law. No transparency report was located.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
The research sources are named without a license or provenance statement. CaseBlink says its immigration research draws on AAO decisions, the USCIS Policy Manual and case law; where the case law comes from and on what terms is not stated.
Sources are named without a license or rights basis, and only on the vendor's content subdomain. Those pages say the product stays current with USCIS guidance and draws on Department of Labor data, including SOC and OEWS wage data, for specialty occupation analysis. The main site describes research across immigration materials without naming any source, and no coverage dates or update cadence are published.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
Searched the product, security, pricing and Autopilot pages, llms.txt and the Service Agreement on 18 September 2026. No check that cited AAO decisions or case law remain good law is described.
Searched the home, about, pricing, terms, privacy and guide pages and the content subdomain on 18 September 2026. Nothing states whether authority the product cites, including precedent decisions, is checked for subsequent history. The vendor's guide tells lawyers to verify all citations before filing.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
Searched the same surfaces on 18 September 2026. CaseBlink flags missing context and quality gaps in uploaded documents, but no behavior for signaling uncertainty in drafts or declining to draft when evidence is missing is described.
Searched the same surfaces on 18 September 2026. No abstention or uncertainty behavior is described for drafting or research. The privacy policy says output may be inaccurate or incomplete and must be reviewed by a qualified human, which is a disclaimer rather than a behavior of the system.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
Searched the AI Hallucination Cases database maintained by Damien Charlotin and trade press reporting on 18 September 2026 for court or agency records addressing fabricated content in output from CaseBlink. None located. This signal does not record litigation history of any other kind.
Searched the AI Hallucination Cases database maintained by Damien Charlotin and trade press reporting on 18 September 2026 for court records addressing fabricated or hallucinated legal citations in output from Parley. None located. This signal does not record litigation history of any other kind.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Searched the product, security and applicant pages, llms.txt, the Service Agreement and a web search of the vendor's blog on 18 September 2026. No ethics opinion, court AI order or rule, or bar guidance is referred to.
The vendor engages named guidance in a published guide, updated 14 July 2026. It quotes and applies ABA Formal Opinion 512 on competence, confidentiality, verification and client disclosure, and Formal Opinion 498 on cloud services, discusses United States v. Heppner and Warner v. Gilbarco on privilege and work product, and links independently maintained trackers of state bar guidance and court AI orders. It does not map its own product to any state opinion.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Speed and efficiency are marketed without billing or disclosure guidance. CaseBlink says cases are filed weeks ahead and five times faster and that firms win more cases with less effort; nothing addresses how firms should reflect AI-assisted preparation in fees or tell clients about it.
Speed and productivity are claimed for work inside a lawyer and client fee relationship, without treatment of billing. The home page says the agent delivers finished work, the about page says firms draft faster, and the content subdomain publishes productivity benchmarks. The vendor's guide notes only that state bars diverge on billing for AI-assisted work; it gives no fee or disclosure guidance, and nothing marks work in the product as AI-assisted for billing purposes.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
Disclosure material is published without an agreement in place. The DPA in the public Service Agreement lists every subprocessor with its role and country, naming OpenAI as the AI model provider and Google as cloud host, with change notice and objection rights, and the security page offers the SOC 2 reports and the zero-data-retention agreement with OpenAI through the Trust Center.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
Searched the same surfaces on 18 September 2026. Output cites to uploaded evidence and the platform keeps audit logs of sensitive operations, but nothing records which parts of a filing were AI-generated or supports disclosure of AI use to USCIS or a court.
Disclosure guidance is published without product support for producing a record. The vendor's guide advises documenting AI involvement where disclosure is required and complying with court certification and disclosure requirements for filings, and explains that some courts require attorneys to certify AI use and human verification. No export or record of which model produced a draft, what sources it drew on or who verified it is described on the main site.
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
- Refusal and Uncertainty Behavior
Which one fits
Choose CaseBlink if
- You need proof that client files stay in the United States. CaseBlink runs on Google Cloud in isolated tenant environments, lists OpenAI, Google and Slack as its only subprocessors, all processing in the United States, and offers dedicated hosting and residency on Enterprise.
- Your procurement team asks for audit reports. CaseBlink reports a SOC 2 Type 1 audit in 2024 and a Type 2 audit in 2025, runs third party penetration tests and a vulnerability disclosure program, and makes its audit reports and zero retention agreement with OpenAI available on request.
- You research AAO decisions and the Policy Manual as part of drafting. CaseBlink researches AAO decisions, the USCIS Policy Manual and case law, cites every output to the uploaded evidence, and reviews uploads for missing context and quality gaps before drafting.
Choose Parley if
- Your firm lives in Outlook and Word. Parley's agent takes delegated work inside Outlook and Word, connects to Google Drive, Box and Dropbox, and builds a client and matter database that fills itself from emails, meetings and uploaded documents.
- You want the vendor's own reading of your ethics duties. Parley publishes a guide applying ABA Formal Opinions 512 and 498, discusses privilege and work product cases, links trackers of state bar guidance and court AI orders, and advises documenting AI use where courts require disclosure.
- You want a team plan with a known monthly price. Parley publishes Starter at $500 a month for up to ten seats with included usage credits and rollover of unused credits up to twice the monthly allotment, offers a free trial, and quotes Enterprise.
In summary
CaseBlink
CaseBlink is a New York company, founded in 2023, whose platform prepares US immigration cases for attorneys across employment, family, humanitarian and naturalization work. It labels client documents, fills USCIS forms, researches AAO decisions and the Policy Manual, drafts support letters, briefs and RFE responses cited to the uploaded evidence, and assembles filing packets through its Autopilot agent. The AI Legal Index grades it in the top two bands on thirteen of fifteen capability axes. It names OpenAI and Google as processors in the United States, reports SOC 2 Type 1 and Type 2 audits, and publishes prices from $100 per seat a month. It says more than 100 firms use it. As of 18 September 2026 the index located no measured accuracy or named bar guidance.
Parley
Parley, from Parley Technologies, Inc. of San Francisco, is an AI platform for legal teams built first for immigration practice. It connects to a firm's email and calendar to build a self updating client and matter record, turns the firm's templates into reusable skills, and drafts petitions, support letters, briefs and RFE responses from uploaded evidence, assembles exhibits and completes USCIS and Department of Labor forms, working inside Outlook and Word. The AI Legal Index grades it in the top two bands on ten of fifteen capability axes. Its terms bar training models other customers can use, and it publishes Starter at $500 a month. As of 18 September 2026 the index located no named model provider and no completed SOC 2 statement in its governing documents.
Questions buyers ask
CaseBlink vs Parley: which is better for immigration drafting?
On published evidence CaseBlink sits in the top two bands on thirteen of fifteen AI Legal Index capability axes and Parley on ten of fifteen, mostly because CaseBlink names its processors, reports its SOC 2 audits and publishes per seat prices. Parley works inside Outlook and Word, builds its client record from email, and publishes guidance on ethics rules. Firms that run from their inbox will find more that fits in Parley.
Does Parley hold SOC 2 certification?
Parley's home page lists SOC 2 Type II, but its privacy policy, updated 8 September 2026, says its security program is designed to align with SOC 2 Type II controls, which is not a statement that an audit was completed. Its trust center could not be read. CaseBlink reports a SOC 2 Type 1 audit in 2024 and a Type 2 audit in 2025, with reports on request. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
Does Parley train AI on client files?
Section 3.3 of Parley's terms bars using customer or generated content to train, fine tune or develop models deployed for or accessible by other customers, and its pricing page lists no model training on the Starter plan. Training a model used only for that customer is not addressed. CaseBlink's security page says client content is never trained on, and its agreement licenses customer data to improve its products. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
Which AI providers do CaseBlink and Parley use?
CaseBlink's data processing addendum names OpenAI as its AI provider and Google as cloud host, both in the United States, with 15 days' notice before any new subprocessor. Parley's documents acknowledge third party AI models and commit to notice of new subprocessors, but its list is held in a trust center that could not be read, so no provider is confirmed. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
What do CaseBlink and Parley both leave unpublished?
Whether their drafts are accurate and whether cited authority is still good. Neither publishes an error rate or test set, neither checks precedent decisions or case law for later treatment, and neither describes what its AI does when the evidence does not support a point. Neither documents walls between matters inside a firm or publishes an AI governance framework. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 25, 2026. No vendor pays for placement.
Three readings to weigh. The outside counsel guideline readiness signal has not been recorded for Parley; that is a gap in the record, not a finding about the vendor. Parley's trust center, where its subprocessor list sits, returned no readable content, so its model providers could not be confirmed. CaseBlink's agreement licenses it to use customer data to improve and develop its products, which is a published term. CaseBlink and Parley were both verified on 18 September 2026. Neither vendor reviewed this page.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything above comes from public material on the dates shown. How the index grades.