IPRally vs PatSnap: how they compare in 2026

The machine learning is what the buyer pays for. IPRally's proprietary Graph AI, described as trained on millions of patent examiner citations, runs the search: a user enters a plain text description, a publication number, an image or a hand drawn search graph, and the model returns ranked prior art without Boolean construction. The company says this replaces query building rather than speeding it up, and the Agent page argues that generic large language models cannot search the patent index and that an agent is only as good as the search technology underneath. Without the models, what remains is a full text patent database with the Boolean search and filters IPRally offers as a secondary option, which is not what anyone buys.

A data platform that existed for over a decade before the models, with a substantial AI layer built on top. Founded in 2007, PatSnap's underlying asset is its data: more than 2 billion structured data points and over 210 million patent, science and technology records, with search, landscaping, dashboards and analytics that predate generative AI and work without it. The vendor itself describes Eureka as a capability layer rather than a separate platform, which is an accurate self assessment. The AI layer is substantial rather than decorative, spanning agents for prior art, freedom to operate, drafting, translation and life sciences work, and the vendor describes a four stage data pipeline built to feed it. A mature platform hosts the model layer rather than depending on it. Remove the models and PatSnap remains a working patent analytics business.

Grounding is documented and inspectable, and measurement stops short of a test an outsider could run. The Agent page says every search case can be opened, reasoning and citations are given per feature, and every claim assessment, feature mapping and relevance score traces to a source document in IPRally's patent index. The retrieval method is described: semantic search over graph representations, patent families, and following examiner citations outward from the strongest documents. One figure is published, a 14 percent improvement in search recall over a full text single search, but the baseline is IPRally's own, no test set is described, and no failure mode is named. The Generative AI Additional Terms of 14 March 2025 disclaim any warranty that outputs are correct and recommend checking them against a primary source. The site gives two corpus sizes, 120 million patents on the pricing page and 130 million on the Agent page. The product retrieves patent documents, not legal authority, so citator checks do not apply.

PatentBench is published as a benchmark built for real patent work, measuring specific tasks against structured data, legal context and source linked evidence, with a full methodology the vendor links to. The published results are a 77 percent hit rate on design freedom to operate and a claim of being 83 times more accurate than ChatGPT on that measure. Naming the comparator and the task means an outsider can find the methodology and disagree with it without contacting the vendor. Grounding is also described in the architecture: retrieval augmented generation over a governed corpus, a four stage pipeline the vendor says reduces hallucinations, and source linked evidence as a property of the benchmark. Three limits apply. The benchmark is the vendor's own rather than independent. The 83 times figure compares against a general purpose chatbot on a task it was never built for, which flatters the result. The methodology page itself was not opened and may say more.

The control structure is well described, with two gaps. What runs unattended is stated: Agent analyzes an uploaded disclosure, identifies claims and features, searches and builds the feature chart. Where it stops is stated too: Agent verifies its findings with the user before searching, and depth is a user setting, from quick triage to litigation grade digging. Every search case can be opened, reasoning and citations are given per feature, and output hands off into a full IPRally project for a professional to continue, described as designed for professional review, not blind trust. The Generative AI Additional Terms add that generative features are always marked before use, are enabled by an administrator and can be disabled at any time. Nothing published says what happens after the system is wrong, and no abstention path is described for Ask AI.

Agent language is used throughout and no oversight model is published. The platform is described as agentic across IP and life sciences, with named agents including an FTO Agent, a lead compound agent, an Antibody Target Predictor and a Markush Claim Drafting agent. A published customer account describes completing infringement analysis for new products in hours rather than weeks. Data governance is described at each stage of the pipeline, which governs inputs rather than agent behavior. Nothing published states what an agent may complete without human review, whether any output can be relied on without verification, what confidence signal accompanies an agent result, or where a human checkpoint sits in a workflow that runs from query to report. The gap matters because the outputs are decision grade: a freedom to operate assessment that clears a product for launch, and a novelty judgment that decides whether a molecule is pursued.

One named customer has a figure, with no date or method behind it. The TLB case study names TLB GmbH, the central technology transfer office for universities in Baden-Wurttemberg, quotes Dr Frank Schlotter, Authorized Officer, and reports about a 50 percent cut in time spent on patent searches, linked to Smart search for query preparation and Ask AI during review. The PDF describes the earlier process of extensive keyword planning, manual Boolean queries and another AI tool the team was not satisfied with. Five more case studies name Unilever, RPX Corporation, Ossur, Metsa Group and Perl IP Consulting, and a testimonial on the Agent page names Josh Walling, Senior Patent Agent at Milwaukee Tool. No case study is dated, and the headline figure is the customer's own estimate of time saved. Only the TLB study was opened; the others may say more.

Named customers, a named executive and independent coverage are all dated and checkable. Named customers include NASA, Tesla, Vodafone, MIT and General Electric, reported in independent trade coverage and not only on vendor pages. NCH Corporation is also named, with Michael Schuster, VP of Product Development and Innovation, quoted by name and title. Corporate history is public: founded in 2007, a unicorn valuation in 2021 after a $300m Series E led by SoftBank Vision Fund 2 and Tencent, and Jeffrey Tiong as chief executive. Independent evidence goes beyond press releases: a Gartner Peer Insights listing, a SoftwareOne marketplace listing, and a hands on review by a named trade publication that took a trial and published its assessment. An outsider can verify the company, the customers, the funding and the product without contacting the vendor. The outcome claims, productivity up to 60 percent higher and workflows 90 percent faster, carry no method, and one customer quote about hundreds of thousands of dollars saved names no organization.

The commitments a buyer most wants are in the agreement, and segregation is unanswered. The Terms of Service of 5 April 2024 say IPRally will not own user generated content, will not share, distribute or disclose it to any third party, and will use it only to perform the Service, and give the customer the right to delete all of it permanently. The Generative AI Additional Terms require any outside provider not to train on input or output or claim ownership, and say user generated content is never sent outside IPRally's cloud unless the customer approves. Privilege and work product are not mentioned anywhere read, and nothing addresses separation between users or matters within an account beyond company specific encryption keys and user level feature rights. That may suffice for an in house IP department, but IPRally also sells to law firms, and for them the matter level question is open.

General enterprise confidentiality is evidenced and the legal professional dimension is absent. Published: strict data isolation, ISO certified security policies, encryption in transit and at rest, consent management, GDPR and CCPA compliance, and positioning as a privacy first partner for enterprises. Those are real controls over the confidential research and portfolio material customers upload. Nothing mentions attorney client privilege or work product, or acknowledges that a law firm user's material carries professional obligations distinct from a corporate research team's. The buyer profile explains the gap without excusing it. PatSnap sells across research and development, strategy, life sciences and legal, and treats confidentiality as an enterprise data question throughout. That is coherent for its market, and it leaves the practitioner question unanswered.

What is published concerns the reliability of AI output, not professional responsibility. The Generative AI Additional Terms make users solely responsible for generated content, tell them to verify it independently, and place the entire risk on them, and the Agent page says Agent searches while the user decides. Nothing says IPRally is not a law firm or does not give legal advice, sets a jurisdiction limit, addresses a practitioner's supervision or competence duties, or engages with ABA Formal Opinion 512, USPTO practitioner guidance or any EPO or national equivalent. This matters because the Agent page markets a should we file signal to R&D and innovation teams described as needing no patent expertise, putting a patentability assessment in front of nonpractitioners with only a general verification disclaimer.

Not located. The platform generates freedom to operate assessments, novelty and design patent risk analyses, and drafts patent claims including Markush structures, all outputs that a practitioner signs off or a business relies on. The published customer accounts describe research and development staff who are not lawyers using them for risk decisions. Nothing published addresses whether output is legal advice, what role qualified counsel plays where an engineer or chemist runs a freedom to operate agent directly, or any professional conduct framework. The question is pressing for this vendor because its named buyers include scientists and engineers, not only attorneys.

Principles are published, with no mechanism behind them. The about page states three company values: transparency, defined as clear, explainable AI so a user always knows how decisions are made; simplicity; and reliability. The features page repeats explainability as a product property. That is a principle specific to AI. There is no responsible AI or AI governance page, no ISO 42001 or EU AI Act position, no one named as accountable for model behavior, no testing regime before release, and no evaluation of uneven output for a system that classifies technology and scores relevance. The marking of generative features and the customer's ability to switch them off are covered under oversight.

Data governance is described in operational detail and model governance is not. The vendor publishes that a four stage pipeline processes over 2 billion data points across 20 domains with rigorous governance at every stage, says this is what reduces hallucinations and improves reliability, and names its retrieval architecture. PatentBench is a published evaluation instrument, and its existence shows that measurement happens. Governance of the models themselves is absent. There is no AI policy, no model card, no bias or fairness testing, no accuracy monitoring beyond the benchmark, no drift statement, no named governance body, no ISO 42001 and no EU AI Act positioning. The vendor documents the governance of its data thoroughly and the governance of its system not at all.

Retention, deletion, access control, subprocessors and incident practice are all published and specific. The privacy policy, updated 12 December 2025, keeps personal data for at most twelve months plus a normal backup period after the reason for holding it ends. The Terms of Service let the customer permanently delete all user generated content, and the Data Processing Agreement requires deletion or return of personal data on termination. Access is limited to authorized staff bound by confidentiality, with SSO, optional adaptive multifactor authentication, user level feature rights and company specific encryption keys. DPA Annex B names nine subprocessors, Okta, Google Cloud EMEA, HubSpot, Intercom, Planhat, Mailgun, Zapier, Slack and Pendo, each with address, purpose, processing location and data categories. DPA clause 9 requires breach notice without undue delay and sets out four required contents, alongside a Vulnerability Disclosure Policy last modified 17 February 2026 and a public status page. Retention is framed around personal data, with no separate period for patent content or AI inputs, and the subprocessor list is dated January 2025.

Isolation and compliance are stated, and the training question is unanswered. Published: strict data isolation, ISO certified security policies, encryption in transit and at rest with named protocol and cipher, GDPR and CCPA compliance with stated consent management, and a self description as privacy first. Nothing is published on whether documents customers upload, search histories, proprietary formulations or research materials are used to train or improve the domain models, and no retention period or deletion right is stated. The gap is material because customers upload their own novel compounds, formulations and unpublished technical material, which is competitively sensitive in a way patent search over public data is not.

The allocation of loss is published and readable before signing, and runs entirely against the buyer. The Terms of Service cap liability at the amount paid in the twelve months before the event, and say IPRally is not responsible for damage caused by correct or incorrect data in the Service. The Generative AI Additional Terms provide outputs as is with no warranty of correctness, place the entire risk on the user, and disclaim liability for outside model services and for security failures at third party providers. The Data Processing Agreement has the same cap but excludes willful misconduct and gross negligence from it. No indemnity runs to the customer, no warranty covers output, and no insurance position is stated. A buyer can read the whole allocation before signing.

No published position is published on liability for AI output, warranty, service levels or remedy. The exposure here falls on business decisions rather than filings. A published customer account describes the FTO Agent completing infringement analysis in hours, and another describes lead compound screening as the first gate in a research process, so a false clearance sends a product to market and a false negative on novelty kills a viable molecule. The vendor publishes a 77 percent hit rate on design freedom to operate, which by its own arithmetic means roughly one in four is missed, and no published position says who carries that.

The product largely stands alone. What is documented is export rather than integration: custom export to Excel, PDF and Word, sharing of searches, collections and monitoring within the account, optional single sign on, and an unspecified reference to tool interoperability on the features page. No connection to an IP management or docketing system, document management system or patent office filing system was found, and there is no integrations page. A company blog post dated 23 July 2026 says a new IPRally API has shipped and MCP support is coming; the future half is not evidence today. A help center collection on product news, data coverage and interoperability, with three articles, was not opened and could change this. Checked the home page, features, pricing, the product navigation, the legal hub and the help center collection list.

The Patsnap Open Platform is a dedicated developer product rather than a connector list, with four integration surfaces: REST APIs, MCP servers, UI widgets and agent skills, backed by a published Developer Center and documentation. Coverage is counted: more than 20 patent APIs, more than 25 corporate APIs and more than 16 infringement APIs. The MCP servers expose the data and its agents as tools for outside AI systems, and the vendor names the frameworks it targets, including LangChain, AutoGen and function calling workflows. Agent skills are also published as a surface. A buyer can get an API key from the site, so a developer can assess, obtain and build against the platform without contacting sales. No named IP management or docketing system integration is published, which is the connection a prosecution practice would ask for first.

Residency is answered plainly and tenancy is not. The privacy policy says IPRally's servers are in the European Union, and the subprocessor annex names Google Cloud EMEA Limited of Dublin as the main cloud provider processing in the EU, with a processing location for each of the other eight subprocessors. DPA clause 4 bars transfer of personal data outside the EU or EEA without the customer's prior written authorization and requires standard contractual clauses for any transfer. The AI question is answered too: the Generative AI Additional Terms separate In-Cloud Services, running inside IPRally's cloud, from External Services reached by API, say core models are hosted by IPRally, and commit that user generated content is not sent outside that environment without customer approval. Nothing says whether the platform is single tenant or multitenant, and the EU is the only location offered.

Nothing located. No hosting provider is named, no region or residency commitment is published, and no single tenant, dedicated or on premises option is described. The absence matters here. The company is headquartered in Singapore, names SoftBank and Tencent among its investors and serves customers including NASA, MIT and General Electric, so where research material and unpublished technical data are processed is a question those customers' compliance functions would ask directly, and public material does not answer it. Stated GDPR and CCPA compliance shows that those regimes are addressed and says nothing about where data sits.

The certification is real and stated on the home and features pages as ISO 27001 certification of the information security management system. No scope, certificate date or period, or auditor is published on any readable page, so a buyer cannot confirm it is current. A trust center at trust.iprally.com, linked from the home page and footer and hosted on Vanta, returned an empty page frame on every attempt, and a search did not surface its contents. It may say more. On the pages that loaded, a named standard appears without scope, date or a route to the report.

A published trust center names standards and annual audits, and the SOC report is Type 1. The trust center lists certification by the AICPA, ISO 27001:2022 with a statement that systems undergo annual external audits, TLS 1.3 and AES-256 encryption in transit and at rest, GDPR and CCPA compliance, and consent management. Naming the ISO revision year and committing to annual external audit addresses currency. The SOC attestation is announced and described as SOC 2 Type 1. Type 1 assesses the design of controls at a single point in time, while Type 2 tests whether those controls operated effectively over a period, and the difference is not cosmetic. No auditing firm is named, and no certificate date is given for the ISO certification.

The model supply chain is disclosed in a contract, not a marketing page. The Generative AI Additional Terms of 14 March 2025 name Anthropic PBC and OpenAI LLC as External Services reached by API, and Google LLC as an In-Cloud Service, covering Gemini models and Model Garden models that meet stated requirements, with links to each provider's terms. External Services host models outside IPRally's cloud and In-Cloud Services run within it, and the features page adds that IPRally hosts its core AI models itself. Every provider must meet minimum requirements: no training on input or output and no ownership claim over either. Providers may be added or removed at IPRally's sole discretion, but the list sits in a published agreement whose changes are posted, with material changes notified by email or at login under the Terms of Service, so customers can see when it changes.

Nothing located. No foundation model provider, model family or version is named, and no subprocessor list was found. The vendor describes its architecture in technical terms, naming retrieval augmented generation and what it calls retrieval augmented thinking, and describes domain specific and domain trained models, all of which characterize method rather than provenance. Independent product review notes an evolving AI backbone, which implies underlying models change over time and makes identification more rather than less relevant for a buyer assessing consistency. For a platform whose Open Platform exposes agents to external systems through MCP servers, the identity of the models behind those agents is a question a developer integrating them would ask.

The unit and structure are published without a figure. Two plans are named, Individual for one user and Team for two to unlimited users, so the unit is the seat. A comparison table of about sixty items covers search, Boolean, review, AI assistants, monitoring, collections, collaboration, classification, coverage, security and support, each marked included, optional or coming soon. Paid extras are named: custom taxonomy AI classifiers, AI classification of monitoring results, simplified R&D licenses, SSO and adaptive multifactor authentication. A three day free trial is self serve. No rate, band, floor or currency appears, the Team plan routes to Contact Sales, and Agent is described as value based and agreed per customer. No volume bands or currencies are given, and nothing says what implementation adds.

The pricing structure is disclosed and no figure is. An independent analyst platform, consistent with vendor material, describes a subscription with multiple tiers that vary by features, user numbers and access level, with details given on request, and a named Enterprise plan that includes agents such as the Antibody Target Predictor and Markush Claim Drafting. Knowing that the meter runs on seats and feature tier, and which capabilities sit in the highest plan, tells a buyer how the bill will grow and which plan they would need. Two self serve entry points lower the barrier: a free trial of Eureka and an API key from the Open Platform. No price, range or entry figure is published at any level, and every commercial route ends in a quote request. That structure comes from the analyst listing because no vendor pricing page is published.

Coverage is described in substance, and the outer boundary is left open. Five role pages address the patent searcher, analyst, manager, attorney and head of IP, and the Agent page explains what IP leaders, patent professionals, R&D and innovation teams, and law firms each get. Five use cases are supported: novelty and patentability, invalidity, state of the art, freedom to operate and classification. Coverage is 58 jurisdictions and 120 million or more patents with machine translation. Customers range from corporate IP departments such as Unilever, Bosch and Nestle to the Danish Patent Office, technology transfer offices such as TLB, and firms such as Perl IP Consulting and Laine IP. Some limits are stated per feature: Agent supports novelty and invalidity, with freedom to operate on the roadmap and Boolean monitoring coming soon. Nothing states limits on team or portfolio size, or that the platform covers patents rather than the wider IP estate.

Coverage is stated with counts and on the dimensions a researcher checks. Scale: more than 2 billion structured data points and over 210 million patent, science and technology records, spanning patents, scientific literature, litigation records, clinical data and technical knowledge across 20 domains. Coverage extends beyond patents, which matters because prior art is not confined to patents. Industries are listed: agriculture and chemicals, consumer goods, food and drink, life sciences, automotive, oil and gas, professional services, aviation and aerospace, and education. Depth in life sciences is shown through named capabilities for sequences, Markush structures, structure activity relationship extraction, clinical trial comparison and druggability. Patent translation trained on global patent data addresses work across jurisdictions. A buyer can test these counts, domains and named document types against their own work. No update lag or refresh frequency is published, and no list of patent offices is given.