Statement of method

What TrueCite does for England and Wales, from which source, without any language model, and where the judgment is left to the practitioner. Written so that a partner, an insurer or a judge can check it.

Deliberately not intelligent

TrueCite uses no language model to decide anything — not to read a citation, not to look it up, not to judge what counts as a match. The whole check is a lookup built from the citation itself, and the same citation against the same register produces the same finding. This is the point, not a limitation.

The courts and the profession have already written the specification, and their own words are set out below. A tool that generates cannot audit itself, and a second generative tool cannot audit the first. Whatever verifies an authority has to sit outside the thing that produced it. That is a structural requirement rather than a commercial one, and it is why the requirement cannot be met by the system that drafted the document: an auditor does not audit their own accounts.

What the courts and the profession require

The Divisional Court has set out what counts as an authoritative source, and has named the register this service checks against. In Ayinde v London Borough of Haringey; Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin), handed down on 6 June 2025, Dame Victoria Sharp P and Johnson J put it this way at §7: “Authoritative sources include the Government's database of legislation, the National Archives database of court judgments, the official Law Reports published by the Incorporated Council of Law Reporting for England and Wales and the databases of reputable legal publishers.” The same judgment states at §8 that “This duty rests on lawyers who use artificial intelligence to conduct research themselves or rely on the work of others who have done so.”, and at §9 requires heads of chambers and managing partners to take practical and effective measures. At §23–29 it sets out what follows for those who do not: public admonishment, a costs order, a wasted costs order, striking out, referral to a regulator, contempt proceedings, and referral to the police.

The Judiciary of England and Wales, in its Artificial Intelligence (AI) Guidance for Judicial Office Holders of 31 October 2025, records that AI tools may “hallucinate”, which includes making up fictitious cases, citations or quotes, or referring to legislation, articles or legal texts that do not exist, and notes that a judge may “remind individual lawyers of their obligations and confirm that they have independently verified the accuracy of any research or case citations that have been generated with the assistance of an AI chatbot.” The Bar Council, in Considerations when using ChatGPT and generative AI software based on large language models of 30 January 2024, states that “it is important for barristers to verify the output of LLM software and maintain proper procedures for checking generative outputs.” The Law Society, in Generative AI – the essentials, updated June 2026, asks practitioners to “carefully factcheck its products and authenticate the outputs”. None of this is a rule this service invented.

Checked on 12 September 2026 and reviewed each quarter. Where this table and a court's own page differ, the court's page governs. Corrections to info@truecite.uk.

How the check is made

You submit one or more United Kingdom neutral citations. For each one, the server asks Find Case Law, the National Archives database of court judgments, for that single judgment's record. The lookup is made once per citation, at the moment you submit it, and it goes nowhere else. Nothing is traversed, nothing is prefetched, no list is walked, and no copy of any judgment is kept here.

This is where the United Kingdom service differs from the Australian one, and it is stated plainly rather than glossed over. The Australian check is answered from indexes this service holds and makes no outbound request at all. The United Kingdom check makes one outbound request per citation, and the only place it is ever made to is the National Archives. Find Case Law permits single lookups without a licence; the licence that would permit bulk extraction is applied for and has not been granted, so one request never fans out into many.

What comes back is identity only: the case name, the decision date, the judgment number and the address of the judgment. The text of the judgment is not downloaded, not stored and not retained. Nothing in the finding, the cache or a log can contain it.

If you supply a case name, it is compared with the name on the record. A name that plainly belongs to another case produces an identity conflict, and both names are shown. A name that differs from the record without contradicting it is reported as not compared rather than resolved, because a spelling difference is not an accusation. Nothing you send — a document, a link, a claim about the source — can change what the register said.

At most twenty United Kingdom citations are looked up in one request, and consecutive lookups are spaced apart, so a request stays a sequence of individual lookups rather than a burst. Past that limit, further citations are answered as not checked rather than queued. That limit is the source's own policy, enforced before anything leaves this service.

Three outcomes

What the service does not do

TrueCite does not check quoted wording. It does not check paragraph content, relevance to a proposition, or whether an authority is still good law. It does not decide the law and it is not legal advice. Each of these is left to the practitioner in writing rather than passed over in silence.

Coverage, and its limits

Coverage follows what the National Archives holds at the moment of the lookup, and nothing wider. In scope: the courts of England and Wales, the Supreme Court of the United Kingdom, the Privy Council, and United Kingdom tribunals. Out of scope: Scotland and Northern Ireland. Reported series citations — [2020] AC 1, [2019] 1 WLR 123 and the like — are not resolved, and are answered by asking for the neutral citation instead.

There is no local index of United Kingdom judgments and no snapshot behind the answer, so coverage cannot be stated as a number of stored records and is not stated as one. The coverage endpoint carries what the service can actually use, and the same declaration is attached to every result.

A citation the register does not hold is a notice to investigate. It is never a finding that the judgment does not exist. A failure to reach the register is a service failure, not a judgment about a citation, and is reported as such.

Attribution and licence

Contains information licensed under the Open Justice - Licence v2.0. Every United Kingdom finding and every coverage declaration carries that sentence and a link to the Open Justice - Licence v2.0. The licence is the National Archives'. The findings are ours.

What is retained

A verification run retains the citation strings, the compact findings, the identification hash of the source record, the observation time, the coverage version and, if you gave one, a report label. Judgment text is not retained. Card details never reach this service. A report URL is readable by anyone who holds it; avoid confidential information in a report label. A hash detects a change to the stored payload; it is not court endorsement.

Payment and partner pilots

A report costs £2.90, paid once through Revolut's hosted checkout page; there is no subscription. Card details are entered on that page and do not pass through TrueCite. The first ten reports issued to a registered firm are free. A partner API pilot is metered but is not automatically billed; a commercial rate must be agreed in writing before billing begins.

Patent and contact

The verification method is the subject of Australian provisional patent application no. 2026907583, filed 6 September 2026. A provisional application is not a granted patent.

TrueCite (ABN 37 688 027 367) · Sydney, Australia · info@truecite.uk