TL;DR

AI-assisted billing narratives are now common in family law, but the ethics rules that govern them are not new. ABA Model Rule 1.5 (reasonable fees), Rule 1.1 (competence, including technological competence per Comment 8), Rule 1.6 (confidentiality), and Rule 5.3 (supervision of nonlawyer assistance) all apply to AI-assisted billing the same way they apply to paralegal-drafted billing. The compliance question is not whether AI can write a time entry — it can — but whether the resulting bill accurately reflects work actually performed, at a reasonable fee, with client confidentiality preserved.

This article covers the four ethics rules that matter, the specific workflow patterns that keep AI-assisted billing defensible, three categories of mistakes that produce real bar complaints, and the disclosure practices that protect both the firm and the client.

Why this matters more in family law than elsewhere

Every practice area is figuring out AI billing. Family law is figuring it out under heightened scrutiny.

Three reasons:

1. Family law clients dispute bills at the highest rate of any consumer-pay practice. When a client is already inclined to question the invoice, an AI-generated narrative that reads as generic or templated becomes ammunition. "This doesn't even sound like my lawyer wrote it" is a phrase that has shown up in fee arbitration filings.

2. State bars have noticed. California, New York, Florida, and several other state bars have issued ethics guidance on AI use in legal practice between 2023 and 2025. The guidance is generally permissive but conditional — and the conditions are exactly the ones that get violated in unstructured AI billing workflows.

3. Family law involves protected information at unusual density. Financial disclosures, custody evaluations, mental health records, allegations of abuse. The confidentiality stakes for sending client data to an AI tool are higher than in many other practice areas.

This is not an argument against AI billing. It is an argument for disciplined AI billing.

The four rules that govern AI billing

ABA Model Rule 1.5 — Fees

Rule 1.5(a) prohibits "unreasonable" fees. Rule 1.5(b) requires that the scope of representation, fee basis, and expenses be communicated to the client. Rule 1.5(c) requires contingent fee agreements to be in writing.

For AI-assisted billing, the operative concern is Rule 1.5(a). A fee is unreasonable if it bills for work not actually performed, or if it bills at a rate inconsistent with the level of skill the work required.

This translates to two specific requirements for AI-assisted billing:

• Time entries must describe work that actually occurred. An AI narrative that hallucinates a task, embellishes a description, or rounds up time is a Rule 1.5(a) violation regardless of who or what generated the language.

• The rate billed must match the work performed. If AI drafted the time entry, the firm cannot bill that drafting time as attorney work at attorney rates. (More on this below.)

ABA Model Rule 1.1 — Competence

Rule 1.1 requires competent representation. Comment 8, adopted by most state bars, expands this to include "the benefits and risks associated with relevant technology."

This means a family law attorney using AI for billing cannot plausibly claim ignorance about how the AI works, what data it processes, or what its failure modes are. "I don't know, the software did it" is not a defense.

In practice, Rule 1.1 requires attorneys using AI billing tools to understand:

• What client data the tool processes and where that data goes

• How the tool generates narratives (template-based, generative, or hybrid)

• The tool's accuracy rate and failure modes

• The verification workflow required before bills are issued

ABA Model Rule 1.6 — Confidentiality

Rule 1.6(a) prohibits revealing information relating to representation without informed consent. Rule 1.6(c) requires reasonable efforts to prevent inadvertent disclosure.

For AI billing, the question is whether sending client matter information to an AI tool — particularly a cloud-based tool — constitutes a disclosure requiring consent.

The current consensus among state bar opinions is:

• Use of AI tools that process client data is generally permissible if the tool has appropriate data protection terms

• Tools that use client data to train their underlying models without explicit opt-out are problematic

• Informed consent is required when client data is being sent to a third party in a manner that creates non-trivial confidentiality risk

• Engagement letters should disclose the firm's use of AI tools in client work, including billing

State-specific opinions worth knowing:

• California State Bar Practical Guidance on the Use of Generative AI in the Practice of Law (issued November 2023, updated 2024) — provides the most detailed framework, including specific guidance on confidentiality

• Florida Bar Ethics Opinion 24-1 (January 2024) — addresses generative AI use, including the disclosure question

• New York State Bar Association Report on the Use of AI by Legal Professionals (April 2024) — broader analysis with billing-specific commentary

ABA Model Rule 5.3 — Responsibilities Regarding Nonlawyer Assistance

Rule 5.3 requires attorneys to make reasonable efforts to ensure that nonlawyer assistance is compatible with the lawyer's professional obligations.

The 2025 ABA Resolution 604 and several state bar opinions have clarified that AI tools fall under Rule 5.3 for the purposes of supervision. This means an AI billing tool is treated, ethically, as a nonlawyer assistant. The supervising attorney is responsible for its output.

In practical terms: if an AI drafts a time entry and an attorney issues it without review, and the entry turns out to be inaccurate, the attorney bears the same responsibility as if a paralegal had drafted it.

What "AI-assisted billing narrative" actually means

The phrase covers at least four distinct workflow patterns. They are not equally risky.

Pattern 1: AI-generated from timekeeper notes

The attorney records brief notes during work ("draft response to opposing counsel re: §4320 factors"). At billing time, AI expands these into full narrative time entries ("Drafted response to opposing counsel's settlement proposal addressing California Family Code §4320 spousal support factors, including analysis of duration of marriage, marketable skills of supported party, and standard of living established during marriage. Reviewed prior correspondence for consistency.").

Risk level: Low to moderate. The factual basis (the timekeeper's notes) is human-generated. The AI is acting as a writing assistant. Verification before issuance keeps this defensible.

Pattern 2: AI-generated from activity logs

The practice management software logs activity automatically (documents opened, emails sent, calendar entries). AI generates time entries from these logs without explicit timekeeper input.

Risk level: Moderate to high. The activity log is not the same as billable work. Opening a document is not 0.3 hours of substantive review. Without human verification, this pattern routinely overstates billable time.

Pattern 3: AI-suggested from matter context

The AI looks at the matter, what stage it's in, and what work is typical at that stage, and suggests time entries the attorney "probably" performed.

Risk level: Very high. This pattern generates entries from inference rather than from record. It is the pattern most likely to produce Rule 1.5(a) violations and the one most likely to surface in fee arbitration.

Pattern 4: AI-cleaned timekeeper drafts

The timekeeper writes a draft time entry. AI improves the prose — fixes grammar, expands abbreviations, ensures consistent formatting — but does not add factual content.

Risk level: Very low. This is the equivalent of using spell-check. Provided the AI is not adding facts, this pattern is generally uncontroversial.

The defensible workflow uses Pattern 1 and Pattern 4. Pattern 2 requires aggressive human verification. Pattern 3 should be avoided entirely.

For broader context on how these patterns fit into family law billing economics, see Aparti's family law collections playbook.

The three categories of mistakes that produce real bar complaints

State bar complaint data through 2025 suggests AI-related billing issues cluster in three categories.

Mistake 1: Time entries describing work that did not happen

The most common AI billing problem. The AI generates a plausible-sounding time entry — for instance, "Research regarding move-away custody factors under In re Marriage of LaMusga" — that describes work the attorney did not actually perform.

The client (or opposing counsel during a fee motion) asks for the underlying work product, and there is none.

This is a Rule 1.5(a) violation, and it is the single fastest way for an AI billing workflow to produce a bar complaint.

Prevention: Every AI-generated time entry must be reviewable against an underlying work product (a draft, an email, a research memo, a court document). If no work product exists, the entry must be removed before the invoice is issued.

Mistake 2: Block billing disguised as itemized billing

AI tools can generate the appearance of detailed billing while actually obscuring what happened. For example, a single 4-hour block of work gets broken into eight 0.5-hour entries, each with a plausible-sounding description, but the entries don't actually correspond to discrete tasks.

This is Rule 1.5(a) violation territory because the apparent itemization is misleading.

Prevention: Each time entry must correspond to a discrete task with a defined deliverable. If the underlying work was a single 4-hour block, bill it as a single 4-hour block with a substantive description, not as eight fragmented entries.

Mistake 3: Confidentiality leakage through AI tools

The firm uses a consumer AI tool (ChatGPT free tier, public Claude.ai) to draft time entries by pasting in matter information. The AI vendor's terms of service allow them to use the input for training. Client information has now left the firm's control.

This is a Rule 1.6 violation, and depending on the jurisdiction, may also implicate Rule 1.1.

Prevention: AI billing tools used in legal practice must have explicit terms preventing the use of client data for model training, must have appropriate security certifications (SOC 2 at minimum), and should ideally be legal-specific tools rather than general-purpose AI assistants.

The defensible workflow

A defensible AI-assisted billing workflow has six characteristics.

1. Human-originated factual basis. Every time entry traces back to a human-generated note, a documented activity, or a verifiable work product. The AI never originates facts.

2. Attorney-level review before issuance. Every AI-assisted invoice is reviewed by the responsible attorney before it goes to the client. Review is documented, even if just by timestamp and initial.

3. Verifiable against work product. For any time entry an opposing party might challenge in fee motion practice, the firm can produce the underlying work product or activity record.

4. Confidentiality-preserving tool selection. The AI tool used has appropriate data protection terms, does not train on client data, and meets the firm's security standards.

5. Disclosed in engagement letter. The firm's use of AI in billing (and other client work) is disclosed in the engagement letter, with sufficient specificity that the client gives genuinely informed consent.

6. Audited periodically. The firm runs periodic audits of AI-assisted time entries — sampling, say, 5% of monthly entries — to verify that the entries accurately describe work performed.

A firm that runs all six characteristics has a defensible AI billing workflow. A firm missing any one of them has an exposure point.

The rate question: who is doing the work, and at what rate?

This is the most underdiscussed issue in AI billing ethics.

When AI drafts a time entry, the AI is doing some of the labor that historically would have been attorney or paralegal labor. The question is: at what rate is that labor billed?

Three positions in current practice:

Position A: Bill at attorney rate. The attorney reviews and approves the AI output, so the work product is attorney work, billed at attorney rate.

Position B: Bill at paralegal rate. The AI is functionally a paralegal, so the time should be billed at paralegal rate.

Position C: Don't bill at all. The AI's marginal contribution is a tool, like spell-check, and tools aren't billable.

The emerging consensus, reflected in California's 2024 guidance and several other state opinions: the time billed should reflect the time the attorney actually spent, not the time the work would have taken without AI assistance.

This is consistent with Rule 1.5(a)'s reasonableness requirement. If AI lets an attorney complete in 30 minutes what would have taken 90 minutes, billing 90 minutes is unreasonable. Billing 30 minutes is reasonable.

The practical implication: firms that adopt AI billing tools should generally see their billable hours per matter go down, not up. A firm where billable hours stay flat or rise after AI adoption should examine whether AI is being used as an efficiency tool or as a billing inflation tool.

Disclosure: what to tell clients

The disclosure question has two parts: what to disclose, and where.

What to disclose

A reasonable AI disclosure for a family law engagement letter covers:

• The tools used. Identify, at least by category, the AI tools the firm uses in client work. "We use AI-assisted tools for time entry drafting, form preparation, and document review" is more specific than "We use AI."

• The data those tools process. Client information including financial disclosures, communications, and matter facts may be processed by AI tools.

• The vendors' data terms. The firm confirms that vendors have appropriate confidentiality and non-training terms.

• The supervision standard. All AI-assisted work product is reviewed by the responsible attorney before being issued or filed.

• Client choice, where applicable. Some firms offer clients the option to opt out of AI-assisted workflows. This is not required but is increasingly common.

Where to disclose

Disclosure typically appears in two places in the engagement letter:

1. A dedicated "Use of Technology" or "AI-Assisted Work" section in the body of the engagement letter

2. A reference in the section on billing practices, specifically noting that AI may be used in time entry preparation

Some firms also include a standing notice on invoices themselves — a one-line footer indicating that the invoice was prepared with AI assistance and reviewed by the responsible attorney.

The competitive landscape: which billing AI tools are in use

A short overview of the tools family law firms are actually using for billing narratives in 2026.

Practice management built-in AI. Clio Duo, MyCase IQ, and similar features built into existing practice management platforms. These offer AI-assisted time entry drafting as part of the existing subscription. The advantage is integration. The limitation is generic AI without family-law-specific knowledge.

Standalone legal AI. Harvey, CoCounsel, and similar enterprise legal AI tools can be configured to assist with billing narratives. Used primarily by larger firms.

Jurisdiction-native tools. AI tools that ship with knowledge of specific jurisdictional workflows. For California family law, this includes Aparti, which generates time entries from the underlying form and intake workflows the firm has already performed — eliminating the inference problem at the heart of Mistake 1 above. See Aparti's case intake and evaluation page and the broader overview of the best AI software for family law firms.

Consumer AI tools used informally. A substantial number of solo practitioners use ChatGPT or Claude directly to clean up time entries. This is the highest-risk pattern from a Rule 1.6 perspective unless the firm has paid plans with appropriate data terms.

The choice of tool matters less than the workflow it sits inside. A firm with rigorous review and verification can use a generic AI tool defensibly. A firm without those workflows will produce ethics issues no matter how sophisticated the tool.

For the broader tech stack context, see Aparti's family law invoice automation tech stack guide.

What state bars are actually saying

A concise summary of the most-cited state bar guidance on AI in legal practice as of early 2026.

California State Bar (November 2023, updated 2024). Practical Guidance on the Use of Generative AI in the Practice of Law. Permissive on AI use; emphasizes Rule 1.1 competence, Rule 1.6 confidentiality, and Rule 1.5 reasonable fees. Specifically addresses billing: "A lawyer must not charge for time saved by using generative AI."

Florida Bar (January 2024). Ethics Opinion 24-1. Permissive with conditions. Requires informed consent for use of AI tools that process client data with non-trivial confidentiality risk.

New York State Bar Association (April 2024). Report on the Use of Artificial Intelligence by New York Lawyers. Comprehensive analysis; emphasizes attorney supervision under Rule 5.3 and the importance of verifying AI output.

Pennsylvania Bar Association (May 2024). Joint Formal Opinion 2024-200. Echoes ABA framework; addresses billing specifically.

Texas State Bar (2024). Less formal guidance, but the Texas Disciplinary Rules Committee has issued informal advisories consistent with the California framework.

American Bar Association (July 2024). Formal Opinion 512 on Generative AI. The most authoritative national framework. Covers competence, confidentiality, supervision, candor to the tribunal, and reasonable fees specifically.

A family law firm using AI for billing should be familiar with both ABA Formal Opinion 512 and the guidance from their specific state bar. Both are short — Opinion 512 is roughly 14 pages — and both reward direct reading rather than secondary summary.

Three questions to ask before adopting an AI billing workflow

1. Can every AI-generated time entry be verified against a human-generated source?

If the answer is no — if the AI is inferring entries from activity logs without human input — the workflow is on shaky ethics ground regardless of how good the AI is.

2. What happens to client data sent to the AI tool?

Read the vendor's data processing terms. Confirm: client data is not used for training, is encrypted in transit and at rest, is retained only as needed, and the vendor has appropriate security certifications. If the answer to any of these is unsatisfactory, the tool is not appropriate for client billing work.

3. Does the firm's engagement letter disclose AI use with sufficient specificity?

A boilerplate "we may use technology" disclosure is not sufficient. Disclosure should identify the categories of AI use (time entry drafting, form preparation, etc.) and confirm that AI-assisted work is reviewed by the responsible attorney.

The trajectory: where this is going

Three predictions for AI billing in family law over the next 24 months:

1. State bar guidance will tighten on the inference problem. The biggest source of complaints is Pattern 3 (AI inferring entries from matter context without underlying record). Expect explicit guidance prohibiting this pattern, likely starting in California and New York.

2. Disclosure requirements will become more specific. Generic "we use AI" disclosures will become insufficient. Engagement letters will need to identify specific tools, specific data flows, and specific supervision practices.

3. The rate question will get more attention. Position C (don't bill for AI-saved time) will become the increasingly explicit expectation. Firms billing the same hours after AI adoption as before will face scrutiny.

The firms that adapt early — by adopting Pattern 1/Pattern 4 workflows, by selecting tools with strong data terms, and by writing genuinely specific engagement letter disclosures — will not just avoid ethics issues. They will have a meaningful competitive advantage when the rules tighten.

Frequently asked questions

Is it ethical for a family law attorney to use AI to draft billing narratives?
Yes, when done correctly. ABA Formal Opinion 512 and state bar opinions from California, Florida, New York, and others confirm that AI-assisted billing is permissible. The conditions are accuracy (Rule 1.5), competence in tool use (Rule 1.1), confidentiality (Rule 1.6), and supervision (Rule 5.3).

Can I use ChatGPT to write time entries for family law matters?
Only with care. Free-tier consumer AI tools generally have terms that allow training on user input, which creates Rule 1.6 confidentiality issues. Paid enterprise tiers with appropriate data terms are more defensible. Legal-specific tools designed with attorney workflows in mind are safer still.

Do I have to disclose AI use to my family law clients?
The emerging consensus is yes. ABA Formal Opinion 512 and several state bar opinions require informed consent when AI tools process client data in ways that create non-trivial confidentiality risk. Disclosure typically appears in the engagement letter.

Can I bill the same number of hours after adopting AI as before?
This is increasingly being treated as a Rule 1.5(a) reasonableness issue. The California State Bar's 2023 guidance specifically states that a lawyer "must not charge for time saved by using generative AI." If AI makes work faster, the time billed should reflect actual time spent, not pre-AI time estimates.

What is ABA Formal Opinion 512?
ABA Formal Opinion 512, issued in July 2024, is the most authoritative national ethics guidance on attorney use of generative AI. It covers competence, confidentiality, supervision, candor to the tribunal, communication with clients, and reasonable fees. It is required reading for any attorney using AI in practice.

What's the riskiest AI billing pattern?
AI generating time entries from inference about what the attorney "probably" did, without a human-generated source record. This pattern is the most likely to produce time entries describing work that did not actually happen, which is a Rule 1.5(a) violation.

Should AI-generated time entries be billed at a paralegal rate?
The emerging consensus is no — the question is not what rate to apply to AI work, but how much time the attorney actually spent. If an attorney spent 30 minutes on work that AI helped complete, billing 30 minutes at the attorney's rate is appropriate. Billing 90 minutes (the pre-AI duration) is not.

Can opposing counsel challenge AI-generated time entries in a fee motion?
Yes, and increasingly do. The standard defenses — underlying work product, contemporaneous records, verifiable activity logs — apply to AI-assisted entries the same way they apply to traditional time entries. The firm must be able to produce the work that the entry describes.

What's the most defensible AI billing workflow?
A workflow with six characteristics: (1) human-originated factual basis for every entry, (2) attorney review before issuance, (3) verifiable against underlying work product, (4) confidentiality-preserving tool selection, (5) disclosure in the engagement letter, and (6) periodic audit of entries. Firms running all six have defensible AI billing.

Does using AI to draft time entries violate the rule against block billing?
Not inherently. AI can produce well-itemized time entries that comply with anti-block-billing rules. The risk is the opposite: AI can produce false itemization, breaking a single block of work into fragmented entries that look itemized but don't correspond to discrete tasks. That is its own Rule 1.5(a) problem.

CTA

Aparti's family law platform was built to avoid the inference problem that drives most AI billing ethics issues. Time entries trace back to the actual intake and form workflows the firm has performed — not to AI guesses about what "probably" happened.

See how Aparti handles case intake and evaluation →

Important note: This article describes general practices and ethics frameworks. It is not legal advice. Attorneys should consult their state bar's current guidance and, where appropriate, an ethics counsel before adopting any AI workflow. State rules vary and evolve rapidly.