UniCourt is a litigation data platform that collects, structures and delivers US court records, mainly through APIs. It covers more than 3,000 state and federal courts in over 40 states, with more than a billion dockets and documents. Machine learning extracts data from dockets and filings, then standardizes it and removes duplicates. DEEP is the extraction and enrichment platform, and DART is the app for docket analytics, research and tracking.
Customers can take data through the web app, APIs, bulk downloads, webhooks or data shares into their own warehouses, structured to SALI standards and fed into matter management, CRM, document management and docketing systems. Buyers are law firms, for docket research, strategy, tracking and experience management, and insurers, for litigation risk and data management. Three self serve plans are priced publicly, with a custom enterprise plan that includes API access.
Court fees pass through at cost. Documents any user downloads go into a shared CrowdSourced Library, free to all other users. UniCourt says its records come from government or reputable private suppliers, and that sealed or restricted documents are not available.
Side by side on the same 15 capability axes and 12 legal signals: where the two part company, and what neither one publishes.
Vendors in the same categories as UniCourt whose graded profiles look most alike, with a line on what each product does.
Similar is not the same, so check that each one does the job you need. No vendor pays for placement.
Ordered by how closely the graded profiles match across the 15 capability axes, inside the same categories. See how each axis is graded in the methodology.
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
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.
UniCourt's machine learning turns scattered court filings into a usable dataset, on a delivery platform that would still work without it. UniCourt describes AI data extraction that pulls data from deep in dockets and documents, with standardized, deduplicated output, and names DEEP as its AI extraction and enrichment platform. Without the models, what remains is collection and delivery of raw court records through APIs, a real product but a much weaker one. The AI drives the core capability on top of a data pipeline.
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.
Every record UniCourt returns is the actual court document or docket entry, but the AI's accuracy is not measured. A user can open and read the source behind anything the platform shows. UniCourt also publishes a real limitation: its public records policy says records may contain errors, be incomplete or inaccurate, or not be the most current, and are provided as is. What is missing is any measure of what the AI does.
Extraction, standardization and deduplication are called unmatched, with no accuracy or error rate, benchmark, test set or evaluation found. Citator checks do not apply to a platform that delivers records rather than legal arguments.
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.
The extraction runs unattended and no oversight model is published. Machine learning classifies, normalizes and deduplicates docket data before a customer ever sees it, and for API and customers that output flows straight into their own systems with no review surface in between. Searched the home page, pricing page, enterprise API page, PACER API page, general disclaimer, terms of service and public records policy: nothing describes what the models decide alone, no confidence or exception handling is documented, and no correction or reporting route for mis-extracted data is published.
The mitigation is structural rather than stated, in that every extracted field sits alongside a retrievable source document, so an attentive user can check any value against the original.
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.
Logos stand in for evidence. The pricing page carries a customer strip including DLA Piper, Paul Hastings, Morrison Foerster, Alston & Bird, Akerman, Pillsbury, Manatt, Stikeman Elliott and Seyfarth Shaw, alongside CNN, The Wall Street Journal, Financial Times, NBC News, RBC, Colliers and Coindesk, which is a serious roster spanning large firms, media and financial services. Nothing sits behind it. Searched the same surfaces and located no named individual, no attributed testimonial, no case study, no figure for what any customer achieved, and no deployment scope or date.
Two names in the strip are misspelled, as Seyearth Saw and Stikeman Elliot, which is a small quality signal on the one page carrying the vendor's customer evidence.
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.
UniCourt discloses its central confidentiality issue plainly and then leaves it unaddressed. Its CrowdSourced Library pools retrieval: whenever a user downloads a court document or docket, it is added to the library and made free to all users, which is why no court fee applies to material already pulled. The material is public court record. Sealed and restricted documents are excluded and nonpublic information is not distributed, so privileged content is not exposed.
But the pattern is. The PACER API page confirms that search results show which cases are already held and when they were added and updated, so a subscriber can infer someone retrieved a docket and roughly when. Nothing says whether a firm's search history, tracked cases or saved work are visible to anyone else, and no training, retention or matter separation position was found.
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
The advice line is addressed directly in UniCourt's disclaimer. The general disclaimer says information on the site is not legal advice and that others may reach different conclusions. UniCourt disclaims all liability for actions taken or not taken based on the information, and creates no attorney client relationship by sending it. It also gives no opinion or endorsement on the quality, competence or character of any person or entity on the site.
That last point matters, since the platform sells analytics on named attorneys and firms, and UniCourt declines to stand behind any conclusion a user draws about them. Missing: no bar or ethics guidance, including ABA Formal Opinion 512, and no jurisdiction or practice limits.
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.
Model behavior at UniCourt has no stated governance. Searched the home page, pricing page, enterprise API page, PACER API page, general disclaimer, terms of service, public records policy and privacy policy on 31 Aug 2026, plus two targeted searches for a security, trust or responsible AI page. None exists. Nothing names an accountable owner, describes testing before release, sets out a responsible AI framework, or addresses uneven output.
The risk is quieter than for a product that predicts rulings, but real. Extraction decides which attorneys, firms and parties are matched to which cases, and matching errors flow into the analytics customers buy and the systems the data is pushed to.
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.
Beyond a privacy policy, UniCourt publishes no security practice. The policy covers what is collected, the CCPA framework and user rights, and offers a way to request redaction of a court record that appears in search engines, a real remedy. Nothing else was found. Searched the home page, pricing page, API pages and the full published document set on 31 Aug 2026, plus two targeted searches for a security or trust page.
There is no encryption standard, access control model, retention period, named subprocessor, breach notice practice or business continuity position. The privacy policy could only be read in part, so a full reading may change what is recorded here as missing.
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.
UniCourt publishes its allocation of loss before signing, and is explicit about the risk its product creates. The terms of service say the subscription agreement includes the general terms, the privacy policy, the public records policy and the applicable rate schedule, which ties the published prices into the contract. The limitation clause excludes indirect, incidental, special, punitive and consequential damages in contract, tort or strict liability.
It excludes any claim about the customer failing to do research or other work properly or completely, even with UniCourt's help. And it excludes any decision or action taken in reliance on information from the platform. The general disclaimer provides the service as is and disclaims all warranties, including merchantability, fitness for purpose and noninfringement. Two things could not be established from the text available on 31 Aug 2026: the liability cap, and whether any indemnity runs to the customer.
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.
For UniCourt, integration is the product, and it is documented to the level an implementer needs. Delivery routes are named: REST APIs, bulk downloads, webhooks, data shares into a customer's warehouse or data lake, and prebuilt integrations. Target systems are named: matter management, CRM, document management, internal legal databases, docketing and experience management systems. Output follows SALI standards, a named industry data standard.
Implementers get API documentation, sample code and a Python library. Billable calls are listed as Search, Case, Download Document, Order Document Confirm and Update Cases, with all other APIs and webhooks included, so a buyer can size an integration in advance. A PACER API also shows whether a case is already held and when it was last updated, so a customer can avoid paying twice.
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.
Nothing published on where the software runs or where customer data sits. Searched the home page, pricing page, enterprise API page, PACER API page and the full published document set on 31 Aug 2026, plus two targeted searches for a security or trust page. No hosting provider is named, no tenancy model is stated, no region or residency option is offered, and nothing distinguishes processing from storage. The gap is more consequential here than on a self-contained application, because the delivery model actively moves data into customer-controlled warehouses and lakes, so a buyer needs to know where UniCourt's side sits in order to reason about the whole pipeline, and no answer is published.
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.
No independent security attestation is published, and no security page exists. No certification is named, no auditor identified, no scope or coverage period given, no penetration testing referenced, no badge displayed and no trust center or portal exists. The absence stands out against how much else this vendor publishes: full pricing, a five-document legal set and API documentation are all open, which makes security the one area where nothing is offered.
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.
Nothing published about the models behind the extraction. The vendor describes AI-powered data extraction, sophisticated data normalization and an AI-driven analytics layer, and names DEEP as its AI extraction platform, but searched across the home page, pricing page, enterprise API page, PACER API page and the full document set no model provider is identified, no model or version is named, no architecture is described, no processing location is given, no subprocessor list exists and nothing commits the vendor to notifying customers when any of it changes.
For a platform whose output flows automatically into customer data warehouses, the inability to tell which model produced a field, or when that changed, is a live operational question rather than a documentation nicety.
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.
The most complete pricing disclosure located in either pull. Rates are published for every self-serve tier: Personal at 49 dollars a month, Professional at 149 and Premium at 299, with annual billing shown as two months free at an effective 40.83, 124.17 and 249.17. The unit of charge is defined and so are the limits attached to it, covering searches at 50, 200 and unlimited, users at 1, 3 and 7, case tracking at 3, 15 and 50, automated searches, and case export at 1,000, with enterprise custom above that.
API billing is itemized rather than gestured at, with billable calls named as Search, Case, Download Document, Order Document Confirm and Update Cases and all other APIs and webhooks stated to be included. Pass-through costs are addressed head on: court fees apply where a court charges and not where it does not, and no fee applies to material already in the CrowdSourced Library. Contract length, cancellation, plan switching, upgrade and downgrade timing and accepted payment methods are all answered, and the absence of a free trial is stated with the reason rather than left blank.
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 are documented by role and use case, and coverage is quantified with two real limits stated. Law firms are addressed through four named applications, covering docket research, litigation strategy, docket tracking and experience management, and insurance through litigation risk and strategy and master data management, with the buying question named in each case. Coverage is stated as more than 3,000 state and federal courts across over 40 states with more than a billion dockets and documents, all federal courts including district, bankruptcy and appellate, and a dedicated court coverage page is published for detail.
Two boundaries are stated plainly: 40-plus states is not 50, and sealed or restricted documents not publicly available online are explicitly excluded. What is missing is depth within the covered states, since nothing published indicates which state systems are complete and which are partial, and the coverage page itself was not opened.
The public pages on file for UniCourt, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.
Ethical Walls and Matter Segregation, Billing and Fee Posture
Read Aug 31, 2026
Court Disclosure Support
Read Aug 31, 2026
Bar Guidance Alignment
Read Aug 31, 2026
Primary Law Corpus Provenance
Read Aug 31, 2026
Fabricated Citation Record
Read Aug 31, 2026
Rates, units and limits are all published on the pricing page, read 31 Aug 2026. There are three self serve tiers: Personal at 49 dollars a month, Professional at 149 and Premium at 299, with a custom Enterprise tier above them. Annual billing gives two months free, for effective monthly rates of 40.83, 124.17 and 249.17. Limits are set per tier. Case research searches are 50, 200 and unlimited, with every search, change or saved search run counted.
Users are 1, 3 and 7. Tracked cases are 3, 15 and 50. Automated searches are 1 and 3, and case export is 1,000. API access is Enterprise only, and API billing is itemized: billable calls are Search, Case, Download Document, Order Document Confirm and Update Cases, with all other APIs and webhooks included. Court document fees pass through on a stated rule. Where a court charges, UniCourt charges, where a court is free, UniCourt is free, and nothing is charged for material already in the CrowdSourced Library.
Contracts are monthly with no long term commitment and can be canceled any time, with discounted annual plans available. Upgrades apply at once and downgrades at renewal. There is no free trial, and UniCourt says that is because plans are month to month. Payment is by major credit card, with ACH on Enterprise plans.
Note: The entry price is $49, the lowest monthly paid rate on the default view. Annual billing brings the same tier to an effective 40.83 a month, recorded in the pricing basis rather than the headline so the figure shown matches what a buyer sees first. There is no free tier, so zero would be wrong. UniCourt publishes three useful things: the limits on each tier, how API calls are billed, and how court fees pass through. Not published: the enterprise rate, which is custom and set through sales, and any figure for API volume pricing beyond the list of which calls are billed.
UniCourt released a legal research plugin for ChatGPT. Users can search dockets and documents across more than 5,000 state and federal courts from inside ChatGPT, drawing on UniCourt's structured litigation data.
A signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Can material a lawyer puts into this product be used to train a model?
A published agreement or policy exists and none of it addresses the question either way, or the document that would answer it could not be read and the summary names the retrieval limit. The summary states which shape the silence takes: an improvement right granted that never names training, or no improvement right granted at all.
Customer data at UniCourt comes with no training position in its terms, privacy policy, pricing or API pages. Nothing says whether customer activity, search history or uploads are used to train or refine the extraction models, either way. The risk differs from a document platform, since customers retrieve public court records rather than submitting client files. But search patterns and tracked case lists are still customer specific data, and neither is addressed. The terms of service and privacy policy could only be read in part, so a full reading of either could change this.
How long does the product keep what a lawyer typed, and can that be set to zero?
No located public material states how long prompts and outputs are retained.
No retention period is set for search history, saved searches, tracked case lists or exported data, and customers get no setting of their own to configure retention. One retention position is published but runs the other way, being about permanence rather than deletion: court documents and dockets retrieved by any user are added to the CrowdSourced Library and made available to all users, with no stated mechanism for removing them.
A redaction request route exists in the privacy policy for a court record appearing in online search engines, which is the closest published deletion remedy located. Both legal documents were read in fragments.
Does retrieval respect the firm’s ethical walls, or can the model read across them?
No located public material addresses walls or matter level segregation.
Customer separation runs against UniCourt's core design, and no separation model is set out. Any document or docket a user downloads is added to the CrowdSourced Library and made free to all users, so retrieved material is pooled across customers by design. That material is public court record and sealed or restricted documents are excluded, so no confidential content is exposed. But there is no document level separation between customers, because the product is not built for it.
Nothing covers access control within a firm on plans with up to seven users, or whether one user's saved searches or tracked cases are visible to colleagues. Checked 31 Aug 2026.
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
No located term or policy addresses third party requests for customer data.
Compelled disclosure of customer data goes unaddressed in UniCourt's terms of service, privacy policy, records policy and disclaimer. Nothing addresses what happens if a third party, law enforcement agency or court requests customer data from UniCourt, and no commitment to notify the customer was found. No transparency report exists. The terms of service and privacy policy could only be read in part, and a compelled disclosure clause would usually sit in one of them, so a full reading could change this.
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the license or rights basis.
UniCourt describes where its records come from and the legal basis, but does not name its suppliers. Its public records policy says all records come from government or reputable private suppliers, that it does not distribute nonpublic information, and that records are provided as is and may contain errors or be incomplete, inaccurate or out of date. It sets out the legal basis, citing support for open court records from the Supreme Court, federal courts and state courts, and noting that public records are usually exempt from privacy laws including the CCPA.
Federal and state data are updated daily. The private suppliers are described by type, not named, so a buyer cannot trace a record to its direct source.
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Good law status belongs to research tools, and UniCourt's product is court data rather than legal authority. No citator, treatment signal or currency check is claimed, and none applies: UniCourt delivers dockets, filings and court documents rather than legal propositions requiring validation, and its analytics concern parties, counsel and case outcomes rather than the standing of authority. This is a question the product class does not raise rather than one the vendor has neglected.
The vendor does publish an adjacent and more relevant currency statement, that records may not be the most currently available information.
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot ground an answer.
Extracted fields arrive without any marker of how reliable the AI extraction was. No confidence score, extraction certainty indicator or error handling is documented for the AI extraction. Nothing says what the platform returns when a field cannot be reliably extracted or a court's records are unavailable. UniCourt says in general that records may be incomplete or inaccurate, but gives no signal on individual outputs.
That matters more for API and data share customers, because extracted values arrive in their own systems with no indication of reliability.
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.
No court order, opinion or disciplinary record naming this product has been found as of 31 Aug 2026. Searched the AI Hallucination Cases database maintained by Damien Charlotin, which tracks court decisions on hallucinated AI content worldwide and records the tool where known, along with 2026 sanctions trackers and trade press summaries. This describes the public record on that date, within what the database covers, and is not a clearance.
UniCourt delivers actual court documents and docket entries that can be traced to source rather than generating citations, so a fabricated citation would have to come from elsewhere.
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
A clear professional-responsibility position is published without reference to any bar authority. The general disclaimer states that information on the site should not be construed as legal advice, that others may reach contrary conclusions on the same information, that no attorney-client relationship is created, and that UniCourt offers no opinion or endorsement as to the quality, competency or character of any person or entity on the site.
Searched the same surfaces on 31 Aug 2026 and located no reference to ABA Formal Opinion 512, no state bar guidance, and no mapping of the product to any professional conduct obligation. The disclaimer is the vendor protecting its own position rather than helping a lawyer meet theirs, which is why this records as not addressed despite the substantive statement.
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
UniCourt's cost material is about its own charges, not a firm's billing to clients, and on that it is clear. Court fees pass through at cost where a court charges, and are waived where a court is free or the material is already in the CrowdSourced Library. PACER costs are framed as something the platform reduces. Searched the home page, pricing page, API pages and the published documents on 31 Aug 2026. No per matter record of AI work for fee purposes and no guidance on billing, fees or client disclosure was found. Case export exists and could support cost allocation, but nothing presents it that way.
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
No located public material supports a client side disclosure obligation.
Outside counsel guideline questions about who sits in the processing chain go unanswered, and UniCourt names no security or trust page that could answer them. Nothing that would support a client-side disclosure obligation was located: no subprocessor list, no model provider named, no trust center or portal, no named certification, no data processing agreement and no client-facing consent or notification material. The published legal set is genuinely open and covers the commercial relationship well, but none of its five documents addresses the third parties in the processing chain, so a firm asked by a client which vendors see its research activity would find nothing to answer with.
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
UniCourt ties every value to its source, but says nothing about the AI's role. Each value sits beside the docket entry or court document it came from, which a user can open. A lawyer relying on UniCourt data can produce the primary record, and case export moves that material out of the app. The AI layer is invisible. No model is named or versioned, nothing separates a machine extracted field from one taken directly from the court record, and no export built for a court disclosure or AI use certification was found on 31 Aug 2026. For customers pushing extracted data into a warehouse, that missing distinction is the harder problem.
Every grade and every signal on this index is drawn from public sources and dated. If a record is wrong, out of date, or missing an artifact the index did not locate, send the source and it will be reviewed and the record redated. Vendors are welcome to submit documentation. Nothing on this index is for sale, including a listing, a placement, or a grade.