Can lawyers bill for using AI in 2025? Ethics opinions, pass‑through costs, and billing best practices

Clients bring up AI on intake calls now, sometimes before they even describe the matter. A few are blunt: “Are we paying for that?”

So can lawyers bill for using AI in 2025? Yes, with guardrails. Most recent ethics guidance says you can bill for legal work done with AI and, in some cases, pass along per‑use AI fees, if the fee is reasonable, you’re transparent, and client data stays protected.

In this guide, you’ll see:

• The ethics basics: Model Rules 1.5 (reasonable fees), 1.1 (tech competence), 1.6 (confidentiality), plus communication duties
• What counts as billable with AI, what belongs in overhead, and when pass‑through costs are okay
• Billing models that fit AI (hourly, flat, hybrid) and time entry examples that show supervision
• How to disclose vendor use and get consent when third parties touch client data
• Practical policies, pitfalls to avoid, and engagement letter language you can copy

Whether you lead a boutique or a growing mid‑size firm, you can bill for AI‑enabled work without surprises, write‑offs, or ethics headaches.

Let’s make this simple and workable.

Quick answer and who this guide is for

Short version: you can bill for lawyer work performed with AI in 2025, so long as you exercise judgment, supervise the output, and keep fees reasonable. Florida Bar Ethics Opinion 24‑1 (2024) and ABA Formal Opinion 93‑379 (expenses and disbursements) track that approach. If you’ve wondered, “can lawyers bill for AI time in 2025,” think of AI as a fast junior who still needs your review.

This guide is for managing partners, billing admins, and practice leads who want clear rules that stand up to bill review. Treat AI like an assistant: sometimes it drafts quickly, sometimes it needs heavy edits. Your job (and your billable work) is applying law to facts, verifying sources, and making calls. Add a few AI‑specific moves, confidentiality checks, clean pass‑throughs, and you’ll have steady invoices, fewer disputes, and a straight story for in‑house counsel about how you saved time without cutting quality.

Why AI billing is a 2025 priority for firms

Clients expect efficiency and transparency. Recent 2024 surveys from groups like ILTA and Clio show firms ramping up generative AI while corporate departments ask how that shows up on the bill. If you don’t set a policy, you invite “What is this AI charge?” emails and write‑downs.

And courts are paying attention. After the Mata v. Avianca sanctions (S.D.N.Y. 2023) over fake citations, many law departments updated outside counsel guidelines to require human verification. Some judges, like N.D. Tex. Judge Starr in 2023, require certifications that filings were human‑reviewed. If you can explain your review process, you keep trust.

Two easy wins: pick a few low‑risk uses (e.g., first‑draft research memos) and pilot alternative fee options that share time savings. Then decide what’s overhead vs. pass‑through now, not after billing. When passing AI costs to clients, ethics rules reward clarity; a simple, consistent policy speeds payment.

Ethics framework that governs AI billing

Four rules set the tone:

Model Rule 1.5 (reasonable fees): no “phantom time.” Bill actual time or price by value with an AFA. Model Rule 1.1 (competence): tech competence is part of the job, several states say lawyers must understand benefits and risks of using AI. Model Rule 1.6 (confidentiality): vet vendors, block training on your data, control retention. Model Rule 1.4 (communication): tell clients when AI meaningfully affects fees, process, or confidentiality.

Also relevant: Model Rules 5.1 and 5.3, supervise lawyers, staff, and nonlawyer assistants (that includes tech and vendors). ABA Formal Op. 93‑379 says general operating costs are overhead, not billable disbursements.

Put plainly: your value is judgment. If AI surfaces cases, your work is picking controlling authorities and applying them. That’s billable analysis. Time spent learning a tool or tweaking settings? That’s overhead. Bake “model rule 1.5 reasonable fees ai billing” and “model rule 1.1 tech competence using ai tools” into your policy so every timekeeper gets it.

What is billable when using AI, and what isn’t

Billable work:

• Drafting, research, analysis, and revision where you exercised legal judgment and reviewed AI output
• Customizing text to client facts, verifying citations, and quality control steps
• “Prompting” when it’s part of substantive legal analysis (framing issues, refining arguments), consistent with Florida Bar Op. 24‑1

Not billable:

• Learning the tool, testing features, vendor troubleshooting, overhead under ABA 93‑379 and echoed in state guidance
• Dropping in generic content with little or no customization; avoid double billing across matters

Example: AI trims a research task from 4.0 hours to 1.2. You bill 1.2, not the “old way.” If the client wants certainty, offer a flat fee that reflects value even as cycle time drops. That’s aligned with billing best practices for ai in legal services.

Tip: when you reuse an internal AI‑assisted template, only record time for matter‑specific reasoning and edits. Keep an internal note about the source (e.g., adapted from prior memo) so you can defend the entry without revealing other client details.

Pass‑through AI costs vs. firm overhead

ABA 93‑379 draws the line: subscriptions, licenses, and general platform costs are overhead, baked into rates, not passed through. Metered, matter‑specific charges (per document, query, or token) can be passed through at actual cost with clear disclosure. Florida Bar Op. 24‑1 tracks this and warns against markups without consent.

Real‑world example: if a drafting run costs $6.42 for a matter, list “AI drafting run (actual vendor charge), $6.42.” If your tool is $99/month across the firm, don’t chop it up across clients, treat it as overhead. Use “ai charges as disbursements vs overhead law firms” as your policy heading so everyone follows the same rule.

For smoother billing, set a per‑matter cap (e.g., “AI runs not to exceed $50 without approval”) and map the expense to a UTBMS code. If a client allows markups where you add extra value, get express consent and describe the value. Otherwise, pass at cost. No vague “technology fee” lines, keep it clean.

Client disclosure and consent best practices

Disclose AI use when it meaningfully affects the work, the fee, or involves sharing client data with a third‑party vendor. Florida Bar Op. 24‑1 and California’s 2023 guidance emphasize informed consent if confidential information leaves your systems.

What to cover:

• Vendor promises: no training on your data, encryption, region controls, retention and deletion
• Costs: whether you’ll pass through metered charges at cost and any caps
• Scope: where you do and don’t use AI (first drafts, research synthesis, not final sign‑offs)
• Restrictions: respect courts or clients that prohibit AI

Sample ai disclosure language for engagement letters: “We may use secure AI tools to help with drafting and research. A lawyer will supervise and verify all outputs. If these tools incur per‑use charges for your matter, we will pass them through at actual cost and list them on your invoice. We do not permit vendors to train on your data.”

At intake, ask whether the client’s outside counsel guidelines mention AI. Capturing preferences early makes client consent for third‑party ai vendors legal faster and avoids mid‑matter surprises.

Billing models that align with AI

Hourly billing still works if your time entries show judgment and review. That said, AI often makes alternative fee arrangements for ai‑enabled work attractive, clients like predictability, and your margin improves as you get faster.

Options to consider:

• Fixed fees for defined deliverables (e.g., demand letter plus revisions), with a note about exclusions for unusual complexity
• Phase pricing in litigation (pleadings, discovery motions), with AI producing first drafts
• Hybrid cap models (hourly to a cap plus a success kicker) where AI reduces variance
• Subscriptions for ongoing advisory, with metered AI runs passed at cost

Model Rule 1.5 reasonable fees ai billing applies to AFAs too, price for value, not minutes. A simple tactic: quote both hourly and flat‑fee options for the same task with a brief note on AI‑enabled efficiency. Many in‑house teams pick the AFA when you’re upfront. In RFPs and OCG responses, frame AI as how you deliver more within the same budget, not as an add‑on fee.

Timekeeping and invoice description guidelines

Write narratives that show lawyering, not tool‑talk. Skip “used AI to draft brief, 1.0h.” Use “Drafted and revised motion with AI assistant; verified citations and authorities; tailored arguments to client facts, 1.0h.” Bill reviewers see the value and are less likely to cut.

Tips:

• Avoid “tested tool,” “training,” or “prompt experiments”, those are overhead
• Separate disbursements: “AI research run (actual vendor charge), $4.18”
• Use matter verbs: verified, analyzed, applied, tailored, reconciled

Quiet upgrade: tag AI‑assisted entries internally (clients don’t see it) and A/B test narrative language. You’ll learn what passes e‑billing review. Map AI disbursements to an approved UTBMS code to prevent auto‑rejects.

Examples:

• “Research: Generated case list with AI; confirmed controlling authorities; synthesized into client memo; removed non‑binding results, 1.4h.”
• “Contract: Used AI to spot deviations from playbook; reconciled with governing law; drafted fallback language, 0.9h.”

These fit billing best practices for ai in legal services and keep invoices clear.

Supervision, quality control, and audit trails

Courts expect human eyes on anything AI touches. After Mata v. Avianca (S.D.N.Y. 2023), several judges issued standing orders requiring certifications that filings were reviewed by a human (see N.D. Tex. Judge Starr, 2023). Your best defense: proof of supervision.

Build a lightweight workflow:

• Save versioned prompts and outputs for key deliverables
• Record checks: Shepardize/KeyCite, verify quotes, confirm local rules
• Peer review for high‑stakes filings with initials and timestamps

Supervising ai outputs duty and audit trails lawyers isn’t busywork, it satisfies Model Rules 1.1 and 5.1/5.3 and lowers malpractice risk. Bonus: if a bill reviewer questions a charge, your verification notes turn a 1.2‑hour entry from “AI black box” into documented quality control.

For sensitive matters, store logs in your DMS, not in the AI tool, and set short retention. In investigations, be careful with privilege, label and handle AI outputs so they don’t slip into discoverable channels.

Engagement letter and policy language (plug‑and‑play)

Cover four points in your engagement:

• Disclosure and consent: “We may use secure AI tools to assist with drafting, analysis, and research. A lawyer will supervise and verify all outputs.”
• Confidentiality: “Vendors may not train on client data; data is encrypted; retention is limited; and storage regions are controlled.”
• Costs: “Matter‑specific, metered AI charges will be passed through at actual cost and itemized. Subscription or platform fees are overhead and not billed as disbursements.”
• Carve‑outs: “If a client or court prohibits AI, we will comply and notify you if timelines or fees are affected.”

Internally, set an AI policy listing approved tools, no‑go data types (e.g., export‑controlled info), and review steps. Include ai disclosure language for engagement letters plus a confidentiality and ai vendors no data training clause. Add a client‑rules playbook (BAAs/DPAs where needed).

One operational trick: create a “no‑AI” flag at intake that disables AI features and updates staffing assumptions. That small step prevents accidental use and the billing disputes that follow.

Jurisdictional and forum‑specific considerations

States are trending the same direction. Florida Bar Ethics Op. 24‑1 (2024) addresses fees, confidentiality, and pass‑throughs (subscriptions = overhead; metered = pass at cost). North Carolina 2023 FEO 3 stresses competence and supervision. California’s 2023 Practical Guidance flags confidentiality and vendor risk. The NYC Bar’s 2024 report covers benefits, risks, and disclosure. Check your state’s latest before rolling out firm‑wide.

Courts are active too. Some require AI certifications or human‑review declarations (e.g., N.D. Tex. 2023). Others warn about hallucinated cites and may sanction careless use. Expect more orders in 2025. Track outside counsel guidelines that go further than court rules.

Practice‑area notes:

• Litigation: Watch local orders, be ready to certify review, keep audit trails for expert work
• IP: Don’t feed unpublished inventions or trade secrets to third‑party systems without tight controls
• Investigations/Regulatory: Keep chain‑of‑custody for AI‑assisted review; privilege issues can be tricky

Bottom line: follow court rules and OCGs first, then layer your internal policy. Track “court rules and ai certifications in filings 2025” in your knowledge base so no one is surprised before filing.

Common pitfalls and how to avoid them

• Billing for learning or vendor troubleshooting. Treat as overhead and build a short internal FAQ.
• Passing subscription overhead as a disbursement. Follow ABA 93‑379; only pass metered, matter‑specific costs at cost.
• Weak disclosure when third‑party systems touch client data. Get informed consent and pick vendors with no‑training and clear retention terms.
• Overreliance on AI without documented review. Require verification steps and peer review for high‑stakes work.
• Reusing work across matters without real customization. Bill only for new analysis; keep provenance notes to defend entries.
• Marking up AI costs without consent. If you add extra value, explain and obtain consent; otherwise, no markup.

One sneaky risk: prompt contamination. Pasting confidential production into a shared sandbox can create conflicts or disclosure issues. Limit third‑party data to approved environments with logging. And to avoid reusing ai work product across matters double billing concerns, tag internal templates as “non‑billable reuse.”

Implementation playbook for your firm

90‑day rollout:

• Weeks 1‑2: Pick vendors with private models and “no training on your data.” Turn on region locks and retention controls. Choose secure vendors and configure privacy.
• Weeks 2‑4: Draft policy and engagement language. Define matter types: approved, restricted, prohibited. Add intake flags and billing rules.
• Weeks 4‑6: Pilot on 5 to 10 matters. Set metering, caps, and approval thresholds. Test narratives and disbursement lines in e‑billing portals.
• Weeks 6‑8: Train teams on supervision, verification, and documentation. Share model entries and expense codes.
• Weeks 8‑10: Review results, write‑offs, client notes, cycle time. Tune and roll out firm‑wide.

Two underrated moves: appoint a billing liaison to pre‑clear AI disbursement lines with top clients so invoices pass on the first try. And add a “red team” pass for complex filings to stress‑test AI output before partner review. Monitoring and small tweaks each quarter keep your program healthy and support alternative fee arrangements for ai‑enabled work without drama.

How LegalSoul supports ethical AI billing

LegalSoul fits how firms actually work:

• Per‑matter metering that tracks each run and passes it through at actual cost, with caps and client‑specific rules
• Engagement workflows that generate jurisdiction‑aware consent language and capture approvals
• Confidentiality guardrails: private models, no data training, encryption, region controls, and retention aligned with your DMS
• Supervision and audit logs: versioned prompts, outputs, redlines, and reviewer sign‑offs
• Billing‑ready narratives: clear descriptions and UTBMS‑mapped disbursements that e‑billing systems accept
• Policy enforcement: matter‑level “no‑AI” flags that disable features where clients or courts say no

Because usage data ties to outcomes, you can show how AI cut cycle time while keeping quality high, great material for RFPs and rate reviews. The conversation shifts from “Is AI billable?” to “Here’s the value you’re getting.”

Sample time entries and invoice language

Time entries:

• “Drafted research memo with AI assistant; identified controlling authorities; verified citations; applied to client facts, 1.2h.”
• “Prepared first draft asset purchase agreement using AI clause library; reconciled with client playbook; negotiated revisions, 2.3h.”
• “Reviewed opposing motion; used AI to surface analogous cases; developed rebuttal arguments; confirmed accuracy, 1.0h.”

Disbursements:

• “AI research run (actual vendor charge), $3.87.”
• “AI drafting run (actual vendor charge), $6.42.”

Flat‑fee note:

• “Scope includes AI‑assisted first draft plus attorney verification and customization; pricing reflects efficiency gains from supervised AI.”

Focus the language on supervision and accuracy, not the tool. Sample time entries for ai‑assisted drafting should read like legal work with tech assist, not tech pretending to be a lawyer. If the client tracks savings, a short note on cycle‑time improvements helps tell the value story and aligns with billing best practices for ai in legal services.

FAQs lawyers ask about billing for AI

• Can I mark up AI costs? Usually no, unless you add extra value and get informed consent. ABA 93‑379 discourages markups on out‑of‑pocket expenses.
• Do I need to disclose every use of AI? Disclose when AI materially affects the work, the fee, or involves third‑party processing of client data. Many firms also disclose generally in engagements.
• Can I bill for “prompt engineering”? If it’s part of substantive legal work (framing issues, refining analysis), yes. Learning the tool is overhead.
• What if a client forbids AI? Follow it. Flag the matter “no‑AI,” adjust staffing and timelines, and note the impact on fees.
• How do I handle reused AI templates across matters? Bill for new analysis and customization only. Avoid double billing and keep internal provenance notes.
• What about court rules? Track court rules and ai certifications in filings 2025; some judges want human‑review certifications.

One more tip: if clients ask for AI‑based discounts, offer AFAs that share savings while protecting margins instead of cutting rates everywhere.

Quick compliance checklist

• Check state guidance (e.g., Florida 24‑1; NC 2023 FEO 3; California 2023 guidance)
• Update engagement letters: AI disclosure, confidentiality, pass‑through language
• Classify costs: subscriptions = overhead; metered charges = pass at cost
• Approve vendors: no data training, encryption, retention controls, region locks
• Define matter categories; add a “no‑AI” intake flag
• Train teams on supervision, verification, and timekeeping narratives
• Map AI disbursements to UTBMS; test in e‑billing portals
• Keep audit trails: prompts, outputs, verification steps, sign‑offs
• Set caps/approvals for pass‑through AI charges
• Review write‑offs and client feedback quarterly and adjust

This checklist ties to billing best practices for ai in legal services and helps you live up to model rule 1.1 tech competence using ai tools.

Key Points

• Bill for legal work done with AI, drafting, research, analysis, review, when you apply judgment. No billing for “phantom time.” Training and testing are overhead.
• Pass‑throughs: subscriptions and platform licenses are overhead. Metered, matter‑specific AI runs can be passed at actual cost with clear labels; markups need consent.
• Disclosure and review: get informed consent when third‑party tools touch client data; verify citations and accuracy; keep audit trails; follow court orders and OCGs.
• Make billing fit AI: precise narratives, UTBMS‑coded disbursements, flat or hybrid AFAs, updated engagement language and policies. LegalSoul can handle metering, consent, audit logs, and clean descriptions.

Conclusion and next steps

You can bill ethically for AI‑enabled work, cut cycle time, and keep client trust. The recipe: bill for lawyer effort, treat subscriptions as overhead, pass metered charges at cost, disclose material AI use, and document your supervision.

Next steps:

• Confirm your state’s latest opinions and client‑specific rules
• Pilot AI‑assisted workflows on low‑risk matters with clear verification steps
• Offer AI‑aligned AFAs to share the efficiency gains
• Roll out engagement language, cost classifications, and timekeeping guidelines
• Run a 90‑day plan, then refine based on feedback

When clients ask if they’re paying for AI, you’ll have a clear, confident answer, and an invoice that tells a value story. Want a faster setup? LegalSoul can handle metering, consent, audit logs, and billing‑ready narratives so your team can focus on practicing law.