Check these 5 things before AI text enters a filing | Legalnaut A Markdown version of this page is available at https://legalnaut.com/blog/check-these-things-before-ai-text-enters-a-filing.md for AI and LLM tools. [AI IN PRACTICE](https://legalnaut.com/blog/category/ai-in-practice) 2026-09-04 2 MIN READ Check these 5 things before AI text enters a filing =================================================== Courts have sanctioned lawyers over citations a model invented. The defence is procedural, and it takes about four minutes per passage. Language models produce fluent, confident, well-structured text that is sometimes about a world that does not exist. In most work that is an annoyance. In a filing it is a professional problem, and it has already ended badly for lawyers in several jurisdictions. The risk is specific, so the check can be too --------------------------------------------- The failure mode is not that a model is bad at law. It is that a model has no way to distinguish between recalling something and constructing something that fits the shape of the answer. Both come out in the same confident register. So the check is not "is this good". It is "can each factual claim here be traced to a document I can open". Five passes, in order, and they get faster with practice. 1. Every claim names a source ----------------------------- Take each sentence that asserts a fact and ask what it points at. A passage that cites nothing is not a draft, it is a hypothesis. Send it back or delete it. This is why it matters that the tool you use attaches sources as it works rather than offering to find them afterwards. Retrofitted citations get attached to whatever is closest, which is a different thing from where the claim came from. 2. Open the source, do not trust the label ------------------------------------------ A reference that looks right is not the same as a reference that is right. Open the document. Confirm it exists, confirm it says what the passage says it says, and confirm it is the document you think it is rather than a similar one. For file material a checksum makes this cheap: the same document, verifiably unaltered since it entered the file. For cited authorities there is no shortcut. Look it up. 3. Quotations are verbatim, or they are not quotations ------------------------------------------------------ Check quoted text character by character against the original. Models paraphrase while keeping quotation marks, which is the most dangerous error on this list because it survives casual review. A related trap in multilingual files: a quotation translated into the language of the filing has stopped being a quotation. Quote the original and translate alongside it. 4. Dates and numbers, individually ---------------------------------- Dates drift. Figures get rounded, transposed, or attached to the wrong party. These are exactly the errors an opponent enjoys finding, and they are quick to check against the source you already have open from step two. 5. Cut what nobody asked for ---------------------------- Models fill space. Ask for a summary of three documents and you may get a fourth point that reads well and rests on nothing. Anything that does not trace back to a document in step one comes out, however good it sounds. Build the check into the tool, not into your discipline ------------------------------------------------------- A process that depends on remembering to be careful at eleven at night is a process that fails at eleven at night. Legalnaut attaches the source to every generated statement and marks AI drafts as drafts requiring review, so an unsupported claim has nowhere to hide and nothing leaves the system labelled as finished work. The checking still belongs to a lawyer. What changes is that the material needed to check is already attached. [contradictions](https://legalnaut.com/blog/tag/contradictions) [ai-review](https://legalnaut.com/blog/tag/ai-review) [for-solo](https://legalnaut.com/blog/tag/for-solo) Legalnaut does this to your own case file: a chronology built from the documents, every finding showing the source it came from. [See the plans](https://legalnaut.com/pricing). Keep reading ------------ [ How to find contradictions in witness statements Two witnesses describing the same meeting differently is normal and rarely worth much. A witness describing a meeting one way while an email sent that afternoon describes it another way is a different thing, and it is the thing worth the search time. ](https://legalnaut.com/blog/find-contradictions-between-statements-and-documents) [ Why your served bundle is invisible to search You search the file for a word you are certain appears in it, and get nothing back. The document is there. The word is on the page. What is missing is any text for the search to match, because the page is a photograph of a page. ](https://legalnaut.com/blog/why-your-served-bundle-is-invisible-to-search) [ How to build a case chronology from 900 emails A commercial dispute is usually an argument about what was agreed and when. The answer sits in correspondence written years ago by people who have since left both companies, and the side that can lay out the sequence cleanly is the side that looks credible. ](https://legalnaut.com/blog/build-a-case-chronology-from-900-emails) ON THIS PAGE SEARCH TOPICS - [ AI in practice 1 ](https://legalnaut.com/blog/category/ai-in-practice) - [ Case preparation 2 ](https://legalnaut.com/blog/category/case-preparation) - [ Evidence 1 ](https://legalnaut.com/blog/category/evidence) TAGS [ for-litigators ](https://legalnaut.com/blog/tag/for-litigators) [ chronology ](https://legalnaut.com/blog/tag/chronology) [ disclosure ](https://legalnaut.com/blog/tag/disclosure) [ contradictions ](https://legalnaut.com/blog/tag/contradictions) [ ocr ](https://legalnaut.com/blog/tag/ocr) [ ai-review ](https://legalnaut.com/blog/tag/ai-review) [ for-solo ](https://legalnaut.com/blog/tag/for-solo) [ exhibits ](https://legalnaut.com/blog/tag/exhibits) [ALL TAGS →](https://legalnaut.com/blog/tags) SCREENSHOT ![]()