Compliance and Privacy11 min read
Nevada Lawyer Advertising Rules for AI Outreach
Nevada lawyer advertising rules applied to AI outreach: a clearance grid for calls, mailers, SMS, chatbots and video ads, plus the 30 day injury window.
Nevada's lawyer advertising rules do not carve out AI. Under the State Bar of Nevada's April 2018 compilation of Rules 7.1 to 7.5, an AI drafted letter to a personal injury victim or a family member is prohibited within 30 days of the incident, every advertisement must name at least one responsible lawyer, and targeted mail must carry a red ink notice that it is an advertisement. The 30 day bar also binds any agent sending for a law firm, including an automation vendor.
This is for Las Vegas law firms, mostly personal injury practices, buying or building automated outreach. The compilation is dated April 2018, so the newest text quoted here is older than 2025; confirm the current wording on the Nevada Legislature's Rules of Professional Conduct page. This is general information, not legal advice.
Nevada lawyer advertising rules reach the AI and the vendor running it
Rule 7.2 opens permissively: a lawyer may advertise "through written, recorded or electronic communication, including public media," subject to Rules 7.1 and 7.3. So the question is never whether a firm may use AI for marketing. It is which outputs, to whom, and when.
Rule 7.3(a) bars soliciting employment from a stranger, meaning someone with no family or prior professional relationship, when pecuniary gain is a significant motive. It defines solicit to include telephone, letter, and any "other communication directed to a specific recipient." A personalized message generated from a row in a lead file is directed to a specific recipient by construction.
The Legislature's page records Rule 7.3 as added effective May 1, 2006 and amended effective December 13, 2012, well before language models. There is no AI exemption and no AI guidance. The rules attach to the communication, not to whoever or whatever wrote it.
For firms that outsource, the key clause sits in 7.3(d): the 30 day limit "applies as well to lawyers or law firms or any associate, agent, employee, or other representative of a lawyer or law firm." A vendor running a crash list campaign for the firm is a representative. Contracts can shift cost between you and the vendor; they do not change who answers to the bar.
Six outreach channels cleared against six duties
Here is how six common channels land against the six duties. Cells marked as a question are judgment calls, explained below the grid.
| Channel | Solicit under 7.3(a)? | 30 day bar, 7.3(d) | 7.2(b) disclaimers | Language parity, 7.2(c) | Filing, 7.2A | Retention |
|---|---|---|---|---|---|---|
| Automated outbound call to a stranger | Yes, telephone is named | Barred at any date: 7.3(d) opens only written contact | Not reached | Not reached | Not reached | Do not build it |
| AI drafted mailer to an accident list | Yes, a letter to a specific recipient | Yes, no send within 30 days | Responsible lawyer, plus fee and results rules if used | Yes, every language used | If disseminated for value | 3 years after last regular use, if filed |
| SMS follow up to a web form lead | Arguably no, the lead asked | Question: no stated carve out for an injured person who wrote in | Any fee, results or quality claim | Yes, if it texts in Spanish | Usually no | Not set by 7.2A; keep threads |
| Chatbot reply on the firm site | No, visitor initiated | Question, if it pitches an injured visitor | Same claim rules | Yes, every language it speaks | No, websites excluded | Not set by 7.2A; keep transcripts |
| AI generated video ad | No, public media | No | Actor or depiction disclosure, responsible lawyer, fee and results rules | Yes | Yes, if paid placement | 3 years after last regular use |
| Review response | No, unless it pitches | Only if it pitches an injured reviewer | Results rules if it cites an outcome | Yes, if not in English | Usually no | Not set by 7.2A |
The SMS and chatbot rows turn on who started the conversation. A person who filled out the form or opened the chat asked for contact, so those rows are arguably not solicitation. But 7.3(d), as quoted in the 2018 compilation, prohibits written communication seeking to represent an injured person within 30 days and states no exception for someone who wrote in first. If your sequence pitches representation to someone hurt last week, get an ethics opinion before automating it.
Review replies are low risk until the model gets creative. A reply that mentions a settlement is a past results reference, and 7.2(b)(7) then requires that the lawyer was lead counsel or primarily responsible, that the amount was actually received by the client, that fees and litigation expenses are stated, and a disclaimer that past results do not guarantee, warrant or predict future cases. Generators write the brag and skip the conditions.
Video carries the heaviest load. A synthetic spokesperson or a generated scene is a depiction under 7.2(b)(1), so the disclosure must identify which persons are actors and stay up for as long as they appear.
The red ink notice and the guardrail block your generator needs
Rule 7.3(c) requires this warning on the outside of the envelope and on the communication side of each page, in red ink:
NOTICE: THIS IS AN ADVERTISEMENT!
A model cannot satisfy that alone; it does not control ink, envelopes, or layout. The notice belongs in the print template, and the send step should reject any rendered piece missing it on any page. For a bilingual mailer, 7.2(c) carries required disclaimers into each language used, and we would treat the notice the same way.
Everything else belongs in the prompt and, more importantly, in checks on the output. This is the guardrail block we would put in front of any system drafting for a Nevada firm. Values in braces come from firm controlled data, never from the model.
``` NEVER PRODUCE (Nevada law firm outreach):
- Any message to an injured person, family member or legal representative unless incident_date is verified and days_elapsed > 30.
- Any voice or phone script for a person with no prior relationship to the firm.
- Any ad or letter that omits the responsible lawyer: {responsible_lawyer}.
- Any synthetic person, voice or dramatized scene without an on screen actor or depiction label for its full duration.
- Any contingency fee mention without: the client may be liable for the opposing parties' fees and costs.
- Any stated fee without its duration and limiting conditions.
- Any past result not in {approved_results}, or without the no guarantee disclaimer.
- Any quality claim not in {approved_claims}.
- Any disclaimer in fewer languages than the body copy.
- Any mail piece without the red NOTICE block on the envelope and each page.
```
Items 1 and 7 depend on facts the model does not have, so they must be checked in code before generation. A prompt that says do not contact anyone within 30 days fails silently on a record with no date. Item 8 exists because 7.2(b)(6) makes quality claims subject to proof of verification on request.
The 30 day clock has to run before the model sees the record
Take a solo personal injury lawyer buying a crash report feed. Assume 400 records a week, and assume 1 in 5 arrives without a usable incident date. That is 80 records a week the system cannot clear, and they should sit until a person confirms the date, not get a date guessed from a report timestamp.
For dated records, do the math in code. A crash on November 3 reaches day 30 on December 3. Because 7.3(d) bars the letter within 30 days and allows it only after, we would set the earliest send to December 4 and treat day 30 as blocked. A record arriving November 8 waits 26 more days before any letter is generated.
Generation should be triggered by the date gate, not by record arrival. If the vendor drafts on arrival and schedules the send, the letter already exists and one misconfigured scheduler mails it early. Build it so there is nothing to send.
The skeptical owner asks whether a model told the date could simply be trusted to wait. It cannot, for a structural reason: a crash record often carries a report date, a filing date, and an incident date in different fields. One mapping error and the clock starts on the wrong day. A code gate keyed to one verified field is auditable; a prompt instruction is not.
After day 30 the letter still has to meet 7.3(b) and (c): the red notice, the responsible lawyer under 7.2(b)(2), and, if it mentions a contingency fee, the 7.2(b)(4) disclaimer that the client may be liable for the opposing parties' fees and costs. If it quotes a fee, 7.2(b)(5) adds how long the fee is in effect and any limiting conditions.
One bilingual video campaign means 16 ads to file and keep
Now assume a 12 lawyer firm running AI video ads with a synthetic spokesperson in English and Spanish, testing 8 variants on paid social. 8 variants times 2 languages is 16 distinct ads. Each one disseminated for something of value is a 7.2A filing, and Rule 7.2A(c)(4) keeps each copy for 3 years after last regular use.
Retention is where generated ads get messy. A pipeline produces new variants weekly, and each paid variant is its own artifact. If the last one runs regularly through June 2027, the firm keeps it until June 2030. A system that regenerates per viewer may leave the firm unable to produce what ran, which is a design flaw, not a storage problem. If the firm runs one evergreen spot instead of rotating variants, the load collapses to 2 files, a fair reason to resist variant testing.
Disclosures multiply too. Each Spanish cut needs the actor disclosure, the responsible lawyer, and any fee or results disclaimer in Spanish, and 7.2(d) requires required information to be "reasonably prominent and clearly legible if written, or intelligible if spoken." A Spanish voiceover over an English only fine print card fails parity.
The same video on the firm's own site is excluded from filing, though every disclaimer still applies. We lay out how we scope these builds on our AI for law firms page.
An AI voice call clears two gates, and a stranger call clears neither
The call row is the one firms are pitched most. The FCC's Declaratory Ruling FCC 24-17 on AI generated voices, adopted February 2, 2024 and released February 8, 2024, confirmed that the TCPA's limits on artificial or prerecorded voice "encompass current AI technologies that resemble human voices," effective on release with no phase in.
An AI voice call needs the called party's prior express consent, rising to prior express written consent for advertising or telemarketing. The message must identify the caller and the responsible entity, and advertising calls must offer an opt out. Emergency purpose calls may be exempt. The ruling also cites an earlier robocall action with a $5,134,000 forfeiture.
Stacked on Nevada's rules, that is two independent gates. NRPC 7.3 asks whether the contact may happen at all; the FCC asks whether the voice had consent. A stranger call fails the first gate, since 7.3(a) names telephone contact and 7.3(d) opens only written communication. A call to a former client may pass 7.3 through the prior relationship and still fail the FCC gate without consent.
Inbound is different: an AI receptionist answering calls the public places to the firm is not an outbound call under this ruling, and we cover that build on our custom AI assistants page.
Where this grid does not hold
First, the text comes from an April 2018 compilation, older than 2025. If 7.2, 7.2A or 7.3 has been amended since, cells change. Second, the 30 day window covers personal injury and wrongful death only; a family law or criminal defense firm still faces 7.3(a) and 7.2, but not that clock. Third, existing clients and family fall outside 7.3(a), but not outside the FCC consent rule for AI voices.
Fourth, the grid covers only the provisions quoted here. It does not address Rule 7.1, the rest of 7.3(b), or federal rules on text messages, which the voice ruling does not settle. It is also Nevada only: a firm advertising into California or Arizona answers to those states too. And if your crash feed vendor cannot supply incident dates at all, the mailer row is not automatable, which is the honest answer. Not legal advice.
Questions owners ask
Can a Nevada law firm use an AI chatbot on its website?
Yes. Rule 7.2 permits advertising through electronic communication, and a visitor who opens the chat started the conversation. The bot still follows the disclaimer rules if it states fees, results or quality claims, in every language it speaks. The open question is a bot pitching representation to someone injured in the last 30 days, since 7.3(d) states no exception for people who wrote in first.
Do AI generated lawyer ads have to be filed with the Nevada bar?
If they are disseminated in exchange for something of value, Rule 7.2A requires filing a copy or recording with the state bar, and each variant that ran counts as its own ad. Filed copies are kept 3 years after last regular use. Ads that live only on the firm's website are excluded from filing, though every 7.2(b) disclaimer still applies to them.
Can a lawyer use AI voice calls to contact accident victims?
Not strangers. Rule 7.3(a) names telephone contact as solicitation, and the 30 day opening in 7.3(d) covers only written communication, so the call is barred no matter the timing. Separately, FCC 24-17 treats AI voices as artificial voices under the TCPA, which requires prior express consent, and written consent for telemarketing. A Nevada firm has to clear both gates.
What to do this week
- List every automated channel the firm or a vendor runs and place each one on the grid above. Any row marked solicit or blocked without a matching control is the first fix.
- Ask your outreach vendor, in writing, where the incident date check happens and what happens to a record with no date. If the answer is the prompt, require a hold queue enforced in code.
- Pull every paid ad variant still inside its 3 year retention window, check each for the responsible lawyer, the actor disclosure, and language parity, and confirm a filed copy exists.
If you want a second set of eyes on the gate design, the free 15 minute audit on our AI consulting page is a quiet place to start.
Drafted with AI assistance, researched, edited, and fact-checked by Elias Musleh on November 19, 2026.
Free 15-min automation audit
Want this handled for you?
We build the automation, run it, and hand you the numbers. Book a fifteen minute call and we will tell you straight whether it is worth doing for your business.
Or reach us directly: vendors@vegasbusinessai.com 702.826.9055