HomeAI Detector › For Legal Writing

AI Detector for lawyers, paralegals, and in-house counsel, built for briefs, contracts, and court filings.

Pre-scan every brief, memo, contract, demand letter, and motion before it leaves your hands. Sentence-level highlights show exactly which lines react AI on argument sections, recitals, bespoke operative clauses, and newly drafted addenda. Calibrated for the formal register and IRAC structure of legal prose. Aware of ABA Model Rule 1.1 competence and Rule 1.6 confidentiality duties. Adjunct to legal judgment, never a substitute for it. No training on submitted drafts. Free to try. No card.

Talk to sales See pricing
Pro at $19.99/mo, $14.99 yearly Privilege aware 90-day matter history
Who it is for

Built for lawyers, paralegals, and in-house legal teams.

Solo practitioners, litigation associates, transactional attorneys, paralegals running the first cut of memos, and in-house counsel reviewing outside-counsel work product all face the same question by 2026: did this filing, this contract, or this client letter pass through an AI tool, and can the lawyer of record defend what was filed under their name. The pre-submission scan is the cheapest answer.

The legal writing stack runs from the demand letter to the appellate brief. Pre-scanning fits every stage because judges, opposing counsel, partners, and sophisticated clients now read with the AI question in mind, and several federal district judges issued standing orders during 2024 and 2025 requiring disclosure of AI assistance on filings.

Solo practitioners and small firms

Solo attorneys and small-firm associates carry the broadest document mix: briefs and motions in litigation matters, contracts and NDAs in transactional work, client emails and demand letters on every active matter. AI assistance has crept into every stage of that workflow during 2024 and 2025. Pro at $19.99 a month, or $14.99 a month on yearly, unlocks 10,000 character pastes and unlimited scans for the iteration weeks before a filing deadline.

Litigation associates and paralegals

Litigation associates draft motions, briefs, and discovery responses under partner supervision. Paralegals run the first cut of legal memos and research summaries that feed senior writers. Both roles routinely use AI as an outline tool or first-pass drafting aide. The pre-submission scan catches AI-shaped phrasing in argument sections before a partner reads the draft, and before the partner files anything under their signature.

In-house counsel and corporate legal

In-house counsel review outside-counsel work product, draft policy memos, and own the internal client communications that travel across business stakeholders. Sophisticated corporate clients increasingly require disclosure when generative AI assisted any portion of paid legal work product. In-house teams use pre-submission scanning to set internal voice baselines and to give outside counsel a clear bar to clear.

ABA Model Rules and bar ethics

Rule 1.1 competence, Rule 1.6 confidentiality, and the duty to supervise.

State and national bar associations have issued AI guidance documents touching competence, confidentiality, candor to the tribunal, and supervision of non-lawyer assistants. The duty of competence in 2026 is widely read to include understanding the tools used in representation. Pre-scanning is one piece of that understanding; it does not replace any ethical duty.

Rule 1.1: competence and AI tool understanding

ABA Model Rule 1.1 on competence is widely read by state bars to include understanding the tools used in representation, including the propensity of large language models to fabricate citations, to hedge on factual claims, and to produce confidently formatted prose that is structurally indistinguishable from researched legal writing. A lawyer who uses AI without understanding those limitations may not be meeting the competence duty as state bars are now interpreting it.

Rule 1.6: confidentiality and third-party tools

Rule 1.6 confidentiality duties apply when client matter is fed into any third-party system. State bar guidance generally points lawyers toward enterprise tools with appropriate data terms, redaction of identifying details before submission, and informed client consent where applicable. TextSight does not train on submitted drafts, but the lawyer still owns the decision to paste privileged content into any external tool. Redaction of party names, docket numbers, and confidential facts is the safe default.

Candor to the tribunal and citation accuracy

The duty of candor under Rule 3.3 sits behind the citation-verification expectation. A lawyer is responsible for the accuracy of everything filed under their signature regardless of the drafting tool. Pre-scanning surfaces AI-shaped passages but does not verify citations; the two checks are independent and both are required for any filing touching AI-assisted research.

Rule 5.3: supervision of non-lawyer assistants

The duty to supervise non-lawyer assistants under Rule 5.3 extends to anyone using AI on a matter, including paralegals, contract reviewers, and document subscribers. Partners running litigation teams and general counsel running in-house functions are both responsible for what their non-lawyer staff produce. The shared scan history on Business tier gives the supervising attorney a contemporaneous view of who scanned what and when.

Bar association group discounts

State and local bar associations including several practice-section groups have negotiated group discounts for member firms on Business tier. Contact sales for the current list. The compliance posture this supports is the same across every state bar: pre-scan, disclose honestly when required, and never let detection substitute for an attorney's professional judgment on competence, confidentiality, or candor.

Plans & pricing

Pricing for solo attorneys, firms, and in-house legal.

Pro at $19.99 a month standard, $14.99 a month on yearly billing, fits solo practitioners and litigation associates. Business at $39.99 a month, or $29.99 a month on yearly, fits law firms, in-house teams, and corporate legal departments with 5 seats. Bar-association group discounts available on Business. Full details on the pricing page.

Free
$0/forever

 

Sample a short memo, demand letter, or contract recital section.
  • 3 scans / day
  • 5,000 chars per scan
  • Sentence-level highlights
  • 2 lifetime AI rewriter uses
Start free
Starter
$7.49/month

Billed $89.88/year — Save $30

For a paralegal or law clerk running a few documents per week.
  • 20 scans / day
  • 20,000 AI rewriter words/mo
  • Chrome extension
  • Email support
Get Starter
Business
$29.99/month

Billed $359.88/year — Save $120

For law firms, in-house legal teams, and corporate counsel.
  • 100,000 AI rewriter words/mo
  • 5 team seats, shared history
  • Audit log, REST API
  • White-label PDFs
Get Business

Most litigation firms and in-house teams see Pro or Business pay for itself the first time a flagged filing gets caught internally rather than in a sanctions hearing. Bar-association group discounts on Business. View full pricing →

Document genres

Briefs, memos, contracts, demand letters, motions, and client communications.

Every legal document genre carries a distinct AI risk profile and a distinct calibration target. The headline score on a contract is not the headline score on an appellate brief, and the editing posture differs accordingly. Read the sentence map by section type, not the single percentage across the document.

Court filings, briefs, and motions

Motions, briefs, complaints, and appellate filings are the highest-stakes category in the post-Mata era. Hallucinated citations are the headline risk; AI-flavored argumentation around real citations is the secondary tell that often correlates with weaker independent verification. Pre-scan every section before filing. Aim above 75 on argument prose, and remember the sentence map is the diagnostic, not the headline score.

Contracts, NDAs, and addenda

Settled boilerplate scores low on detectors regardless of who drafted it because defined terms and standard clauses recur verbatim by design. The signal worth chasing concentrates in recitals, bespoke operative clauses, and any newly drafted addenda or schedules. AI flavor in those passages reads as lazy lawyering rather than misconduct, but it also raises the quiet question about whether the bespoke language reflects considered judgment.

Legal memos and research summaries

Internal memos summarising case law, deposition transcripts, or discovery production are common AI tasks because input volume is large and output format is consistent. The risk is the memo inheriting the model's hedging cadence and tidy parallel structure, then becoming the input for a more senior writer who relies on the summary. Pre-scan the memo before it leaves the paralegal's hands.

Client emails and demand letters

Client communications and demand letters now travel through AI tools at the first-draft stage for many practitioners. Clients reading a letter that scans as AI lose confidence in the attorney faster than they lose confidence in a sloppy edit, because the AI signal reads as bot-substitute rather than human attention. The pre-submission scan on outgoing client letters is the lowest-cost trust insurance available.

Law school exam answers and moot court briefs

Law students writing under permitted-AI policies use detection to confirm their own voice carries the final draft. Students writing under prohibited-AI policies should not be using AI at all; a detector does not retroactively cure an honor code violation. The pre-submission scan helps an honest writer confirm their work reads as theirs and gives the moot court board a clear external signal.

Pre-submission workflow

Scan before partner review, opposing counsel review, and the bench.

Most legal documents clear three checks before they reach the tribunal or the counterparty: the drafter's own pass, the partner or supervising-attorney review, and the conflict and citation verification before filing. The pre-submission scan fits before the first internal review, so the version your partner reads is already AI-clean.

Step 1: draft normally

Write in your usual editor: Word, Docs, or your firm's document automation system. Using AI for an outline pass, a research synthesis, or to break drafter's block on a difficult argument section is the realistic 2026 default under most firm policies. Compose the prose itself in your own voice or the firm's voice from your research notes, prior filings, and conversations with the supervising partner.

Step 2: verify every citation against an authoritative reporter

Before the AI scan, verify every citation against an authoritative reporter or research platform: Westlaw, Lexis, Bloomberg Law, the underlying federal reporter, or the state equivalent. Detection does not verify citations; that check sits on its own track and is the non-negotiable step before any filing. Mata v. Avianca made the consequence of skipping it stark.

Step 3: paste and scan section by section

Open app.textsight.ai, paste each section, and scan. Free tier handles 5,000 characters in one paste. Pro handles 10,000, which fits a typical motion section or a long contract clause. For a 20-page brief, split by section: statement of facts, argument headings, application sections, conclusion. The scan returns in well under a minute with an Authenticity Score and a sentence-by-sentence colour map.

Step 4: rewrite reds in your own voice

Rewrite flagged sentences in your own voice while preserving legal precision, defined terms, and citation accuracy. A brief or memo typically needs three to eight sentence rewrites to move from a 65 score into the 85 range. The sentence map tells you which lines need attention; the headline score confirms the rewrite landed.

Step 5: hand the package to partner review

Send the consolidated draft to the supervising partner, in-house GC, or court clerk for the substantive review. Pro history keeps every section scan for 90 days. PDF export gives you a contemporaneous record. A typical motion section round-trips in about fifteen minutes; a full appellate brief in about ninety. If your jurisdiction requires disclosure of AI assistance on filings, include it according to the standing order; the scan itself is internal hygiene, not a disclosure document.

Court submission context

Mata v. Avianca, Loomis v. Wisconsin, and the AI disclosure trend.

Several federal district judges issued standing orders during 2024 and 2025 requiring disclosure of AI assistance on filings. State courts followed in California, New York, Texas, and Florida. Pre-scanning every filing and disclosing honestly on the record is the defensible 2026 posture for any litigator.

Mata v. Avianca: the watershed sanctions case

In 2023, in a federal case widely referred to in shorthand as Mata v. Avianca, lawyers were sanctioned after a filing was found to contain case citations generated by ChatGPT that did not correspond to any real opinion. The case crystallised a problem that had been theoretical: a large language model will produce confident, plausibly formatted citations to authorities that simply do not exist. A lawyer who relies on them without independent verification puts their license on the line. Treat any citation that originated in or passed through an AI tool as unverified until you have read the underlying opinion yourself.

Loomis v. Wisconsin and algorithmic risk

Loomis v. Wisconsin and the surrounding 2024 to 2025 discussion of algorithmic tools in criminal sentencing pushed the bench to ask explicit questions about AI provenance on filings. Several state appellate courts now ask AI-assistance questions on the record at oral argument when the underlying briefs read AI-shaped. Disclosure on the record matters more than it did even twelve months ago.

Federal standing orders on AI disclosure

By late 2025, dozens of federal district judges had issued standing orders requiring attorneys to certify on filings that AI tools were either not used or, if used, that every citation was independently verified. The exact phrasing varies by chambers; the substantive expectation is consistent. Check the standing orders for every judge you appear before, and update your firm's filing checklist accordingly.

State court trends and bar disclosure expectations

State bar disclosure expectations are tightening alongside the federal standing orders. New York, California, Florida, Texas, and the District of Columbia have all issued bar guidance touching AI disclosure during 2024 and 2025. Smaller state bars are following. The defensible posture across every jurisdiction is the same: pre-scan, disclose honestly when required, and never let detection substitute for the lawyer's verification step or for professional judgment on the substance.

Client trust and firm voice

Clients reading "looks AI" letters lose confidence fast.

Sophisticated corporate clients increasingly require disclosure when generative AI assisted any portion of paid legal work product. Beyond the formal disclosure, clients reading a demand letter or a closing memo that scans as AI lose confidence in the attorney faster than they lose confidence in a sloppy edit. The AI signal reads as bot-substitute rather than human attention.

Why AI-shaped letters damage client relationships

Clients pay for an attorney's professional judgment, not for prose generation. A demand letter or closing memo that reads as AI-shaped sends the implicit signal that the lawyer did not engage with the matter, regardless of how much substantive work happened upstream of the writing. The signal is wrong in most cases but the perception sticks. Pre-scanning outgoing client communications is the lowest-cost client-retention investment available.

Calibrating firm voice across multiple attorneys

Litigation matters and large transactions routinely ship from three to seven attorneys across associates, partners, and co-counsel. Each contributor enters AI assistance at a different point and the combined draft can read uneven, which itself reads as AI to a careful judge or sophisticated counterparty. Business tier with 5 seats and shared scan history lets the team see the same Authenticity Score and flagged sentences. The supervising partner runs the consolidated scan before the filing or the closing.

Sophisticated client AI disclosure requirements

Several large corporate legal departments now require outside-counsel firms to disclose AI use on any work product billed at the partner rate. The exact form varies: some clients ask for a line item on the engagement letter, others ask for a per-matter disclosure. The pre-submission scan history on TextSight gives the firm a contemporaneous evidence trail supporting whatever disclosure the client requires.

Shared scan threshold on Business tier

Business tier with 5 seats and shared history lets the supervising partner, the lead associate, the contract reviewer, and the paralegal all see the same scan and flagged sentences. The Authenticity Score becomes the team threshold rather than a per-author judgment. The audit log records who scanned what and when, useful when a junior associate delivers a section that flags higher than the rest of the brief.

Your client matter stays yours

Privilege-aware: no training on submitted text, redaction is the safe default.

Legal work is privileged and Rule 1.6 confidentiality duties apply to every third-party tool in the workflow. TextSight is designed to honour those rules out of the box, not as a paid setting you have to find. For privileged matters, the firm IT and the bar guidance for every jurisdiction where you practice are the binding context, not this page.

No training on submitted drafts

Briefs, memos, contracts, client letters, and demand letters submitted for scanning are never used to train the classifier or any other model. This is a contract clause, not a configuration toggle. It applies on the free tier the same way it applies on Pro and Business. The scan runs against the existing model and returns a score and sentence highlights without feeding the training pipeline.

Redaction is the safe default for privileged content

Many practitioners scan redacted drafts where party names, docket numbers, addresses, dollar figures, and confidential facts are replaced with placeholders before pasting into any external tool. The detector reads prose cadence, not party identity, so redaction does not affect the score. For privileged matters and for any matter under a protective order, consult firm IT before pasting any content into a third-party system.

Retention you control

Free tier scans are not stored beyond the result session. Pro tier history is retained for 90 days; individual scans are deletable on request. Business tier offers configurable retention with team controls and a standard DPA on request for in-house procurement. Match the tier to your firm's external-tool retention policy, and align retention settings with the bar guidance for every jurisdiction where you practice.

Consult firm IT and your jurisdiction

None of this page is legal advice and none of it substitutes for the bar guidance issued by the jurisdictions where you are admitted. Before using any third-party detector on client matter, confirm with firm risk and IT teams. The professional judgment on competence under Rule 1.1, confidentiality under Rule 1.6, candor under Rule 3.3, and supervision under Rule 5.3 remains with the attorney of record. TextSight is an adjunct to legal judgment, never a substitute.

FAQ

Legal writers frequently ask.

Does TextSight replace independent citation verification?
No. TextSight reads prose cadence and returns an AI-likelihood score with sentence-level highlights. It does not verify that case names, reporters, and pinpoint cites resolve to real opinions. Citation verification against an authoritative reporter or research platform is a separate, non-negotiable step before any filing, and the Mata v. Avianca sanctions made that point in stark terms. Use TextSight as an adjunct to legal judgment, never as a substitute for the verification step or for an attorney's professional review.
How does TextSight handle privilege and confidentiality on submitted drafts?
Drafts submitted for scanning are never used to train the classifier or any other model. This is a contract clause, not a configuration toggle. That said, ABA Model Rule 1.6 confidentiality duties still apply to any third-party tool in the workflow. Many practitioners scan redacted drafts with party names, docket numbers, and confidential facts replaced by placeholders before pasting into any external system. For privileged matters, consult firm IT and the bar guidance for every jurisdiction where you practice, and align your retention settings with firm policy on Business and Enterprise tiers.
Why do fully human briefs and contracts sometimes score as AI?
Legal writing carries a formal register, a prescribed IRAC or CRAC structure, defined terms that recur from one document to the next, and settled boilerplate clauses. All three traits overlap with the patterns detectors associate with machine-generated prose. A fully human brief can land in the 55 to 70 band simply because of the genre. The sentence-level highlights are the diagnostic that matters: you see which sentences flag, decide whether they reflect required legal phrasing such as defined terms or boilerplate, and rewrite only the lines that read AI-shaped.
What do ABA Model Rule 1.1 and Rule 1.6 require of an attorney using AI?
ABA Model Rule 1.1 on competence is widely read to include understanding the tools used in representation, including the propensity of large language models to fabricate citations and hedge on factual claims. Rule 1.6 on confidentiality applies when client matter is fed into any third-party system. Many state bars, including New York, California, Florida, Texas, and the District of Columbia, have issued AI guidance reinforcing both duties and adding candor and supervision threads. TextSight supports compliance by providing pre-submission visibility; the professional judgments under both rules remain the attorney's responsibility.
How does the Mata v. Avianca and Loomis v. Wisconsin context apply to my practice?
Mata v. Avianca, the 2023 federal sanctions case, established that lawyers are responsible for the accuracy of every citation in a filing regardless of whether the citation originated in an AI tool. Loomis v. Wisconsin and the 2024 to 2025 wave of standing orders from federal district judges on AI disclosure have pushed many courts to ask AI-assistance questions on the record. Several jurisdictions now require disclosure of AI use in filings. Pre-scanning every filing and disclosing honestly on the record is the defensible 2026 posture for any litigator.
Which tier fits a solo attorney versus a litigation firm?
Pro at $19.99 a month, or $14.99 a month on yearly billing, fits solo practitioners and small firms running 20 to 50 filings, memos, or contracts a week, with 10,000 character pastes, unlimited scans, 90-day history, and PDF export for matter files. Business at $39.99 a month, or $29.99 a month on yearly, fits litigation firms, in-house legal teams, and corporate counsel with 5 seats, shared history, audit log, REST API, and white-label PDFs. One avoided sanctions hearing or client-relationship problem covers either tier indefinitely. Bar-association group discounts are available on Business; contact sales.
Can I use TextSight on a law school exam, memo, or moot court brief?
Yes, provided your law school's academic policy permits AI use in the workflow that produced the draft. If your school prohibits AI assistance on exams, journals, or moot court submissions, do not use AI in drafting; a detector cannot retroactively cure an honor code violation. If AI is permitted as an aide for outlining or research, pre-scanning helps you confirm your own voice carries the final draft before submission. Consult your honor code, the specific course policy, and your moot court rules before relying on any AI workflow.
Is contract drafting language particularly hard to score cleanly?
Yes. Contracts repeat defined terms verbatim by design, reuse settled boilerplate across instruments, and follow predictable structural patterns. All three push detector scores into the 55 to 70 band even on entirely human-drafted contracts. The signal worth chasing concentrates in the recitals, the bespoke operative clauses, and any newly drafted addenda or schedules. Sentence-level highlights surface AI-shaped phrasing in exactly those passages, so you can rewrite the bespoke language rather than re-engineer the settled boilerplate that scores low for legitimate genre reasons.
Related

More for lawyers and legal teams.

Pre-scan every filing, memo, and contract. Submit clean. Disclose honestly.

Free to try. No card. Pro at $19.99/mo, or $14.99/mo on yearly. Business with 5 seats for litigation firms and in-house legal teams.

Talk to sales See pricing
Privilege aware · No training on submitted drafts · ABA Model Rule 1.1 and 1.6 context · Adjunct to legal judgment, never a substitute