Key takeaways
- Utah requires disclosure when a consumer clearly asks, and prominent disclosure up front in high-risk interactions for licensed occupations; disclosing at the outset and throughout is a safe harbor[2].
- California's bot law covers websites, apps and social media, not phone calls[3][4]; its dialing-device law makes automated calls say when a voice is artificial[5].
- Colorado's 2026 law requires notice at the point of interaction for covered automated decision tools used in consequential decisions, which include housing[6].
- Federal rules treat AI voices as artificial for outbound calls[7]; a per-call AI disclosure rule has been proposed, not adopted[8].
Where the law stands in late 2026
Brokers adding automated phone answering or website chat in 2026 keep running into the same compliance item: whether the law makes them tell people. Only a handful of states have written rules for it, each covering specific situations, and the federal rules bite hardest on outbound calls. The useful work is knowing which situations those are and writing one disclosure that covers all of them.
Consumers have already made up their minds about wanting to know. In Pew Research Center's 2025 survey, 76% of Americans said it is extremely or very important to be able to tell whether content was made by AI or by people[1]. Real estate clients are already watching for it: among sellers who hired an agent, 21% knew and another 36% suspected their agent was using AI tools[9].
This guide tracks the state rules that matter to a brokerage's calls and chat. For the wording side in depth, by channel, see disclosing AI to callers and leads. For the broader risk picture, see using AI in real estate: compliance risks.
Utah: the most specific rules
Utah's Artificial Intelligence Consumer Protection Amendments, in force since May 7, 2025, replaced the state's 2024 provision with two tiers[2].
- Consumer transactions. A business that uses generative AI to interact with someone in a consumer transaction must disclose that the person is interacting with generative AI and not a human, if the person asks. The question has to be a clear and unambiguous request to find out whether they are talking to a human or to AI[2].
- Regulated occupations. Someone providing services in an occupation regulated by Utah's Department of Commerce that requires a license must prominently disclose AI use when the interaction is high-risk: verbally at the start of a spoken interaction, and in writing before a written one[2].
- High-risk interactions include collecting sensitive personal information such as financial data, and giving personalized recommendations or advice that could be relied on for significant personal decisions, including financial advice[2].
Real estate licensing in Utah runs through a Department of Commerce division, so a licensee's assistant that asks about a buyer's pre-approval amount or offers pricing guidance on a home sale is the kind of conversation that could be treated as high-risk. A sign-call assistant that only answers listing questions and takes a message is less likely to be, but the line is not drawn in the statute, so build to the stricter tier.
Utah also gives a safe harbor. A business is not subject to enforcement under the disclosure section if its generative AI clearly and conspicuously discloses, at the outset of the interaction and throughout it, that it is generative AI, that it is not human, or that it is an AI assistant[2]. Fines can reach $2,500 per violation, and violating an order carries up to $5,000 per violation[2]. One more line matters for brokers: it is not a defense to a consumer protection violation that generative AI made the statement[2]. If the assistant misstates an HOA fee, the business owns the misstatement.
California: online bots and automated calls
California's bot law makes it unlawful to use a bot to communicate with a person in California online, with the intent to mislead them about its artificial identity, to encourage a purchase or sale. A business that discloses it is a bot is not liable, and the disclosure must be clear, conspicuous and reasonably designed to inform the person[3]. Online means a public-facing website, web application or digital application, including social networks[4]. Website chat and social media messaging are in scope; ordinary phone calls are not.
A separate California statute governs automatic dialing-announcing devices, the equipment that plays recorded messages on outbound calls. Before the recorded message, a natural voice announcement must state the nature of the call and the business, ask whether the person consents to hear the message, and tell them if the message uses an artificial voice, defined as one generated or significantly altered using AI[5]. A brokerage that sends prerecorded or AI-voiced outbound calls into California needs that announcement as well as federal consent.
Colorado: notice when automated tools affect housing decisions
Colorado's SB26-189, signed May 14, 2026, repealed and reenacted the state's 2024 AI Act as a law on automated decision-making technology[6]. It requires deployers to give clear and conspicuous notice to consumers at the point of interaction with a covered tool, and it applies to consequential decisions, a list that includes housing as well as financial and lending services[6]. Developer documentation duties begin January 1, 2027[6].
For most sales brokerages, a phone assistant that answers listing questions and passes messages to agents is not making a consequential decision. Property managers are different. A tool that screens rental applicants, ranks them or decides who gets a showing slot is much closer to the housing decisions the law names. Inventory where any automated tool decides something about a person, and treat those uses separately from simple answering. Bias risks in that kind of screening are covered in AI and fair housing.
The federal layer
On February 8, 2024 the FCC ruled unanimously that calls made with AI-generated voices are artificial under the TCPA[7], which puts AI-voiced outbound calls under the same consent rules as prerecorded ones: telemarketers need prior express written consent before robocalling consumers[7]. The details for agents are in AI voice calls and the TCPA.
The FCC has also proposed, but not adopted, a rule requiring callers using AI-generated voice to disclose at the beginning of each call that it uses AI-generated technology[8]. Until it is final, it is a sign of direction rather than a rule. Building the disclosure into every outbound AI call now costs one sentence.
Keep inbound and outbound apart in your compliance file, because the rules are different. When a buyer calls a sign number and an automated assistant answers, the buyer started the call; the TCPA's consent rules for artificial voices are aimed at calls the business places. The state disclosure laws above can still apply to that inbound conversation, which is why the greeting matters. Outbound is where the heavier obligations sit: written consent for telemarketing with an artificial voice[7], California's announcement for dialing devices[5], and whatever per-call disclosure the FCC finally adopts[8].
Timeline: what is in force and what is coming
Feb 2024
FCC rules AI-generated voices are artificial under the TCPA[7]
Sep 2024
FCC proposes a per-call AI disclosure for AI-voice callers[8]
May 7, 2025
Utah's two-tier disclosure rules and safe harbor take effect[2]
May 14, 2026
Colorado signs its automated decision-making technology law[6]
Jan 1, 2027
Colorado developer duties begin[6]
Jul 1, 2027
Utah's AI Policy Act reaches its current repeal date[2]
Dates from the statutes and FCC documents cited. Proposed rules can change before adoption, and repeal dates are often extended.
Example (fictional): a brokerage in Boise runs one website chat assistant and forwards its main line to an automated assistant after 6 p.m. In a single evening it hears from a Utah buyer relocating for work, a California seller who owns a rental in Idaho and an Idaho renter asking about a lease. Idaho is not one of the states in this guide, but two of the three people are covered by other states' rules the moment they start the conversation.
The Utah buyer asks about loan amounts, which edges toward a high-risk interaction for a licensed occupation[2]. The California seller is using the website chat, which is an online bot interaction under California's definition[4]. The brokerage does not try to detect where each visitor lives. It uses one greeting and one chat header that satisfy Utah's safe harbor, and both conversations are covered without anyone checking an address.
The rules side by side
| Law | Channel | When disclosure is required | What satisfies it | Status |
|---|---|---|---|---|
| Utah: consumer transactions | Any generative AI conversation: text, voice or visual | When the person clearly asks if it is AI | Say it is generative AI and not a human | In force since May 2025 |
| Utah: regulated occupations | Licensed services, high-risk interactions | Up front: spoken at the start, written before | Prominent disclosure; safe harbor if at the outset and throughout | In force since May 2025 |
| California bot law | Websites, apps, social media | When a bot is used to encourage a sale | Clear, conspicuous statement that it is a bot | In force |
| California dialing devices | Automated outbound calls with recorded messages | Before any recorded message plays | Natural voice announcement, including if the voice is artificial | In force |
| Colorado ADMT law | Covered tools in consequential decisions, including housing | At the point of interaction | Clear and conspicuous notice | Signed May 2026; duties start in 2027 |
| FCC (federal) | Outbound AI-voice calls | Consent before telemarketing calls; per-call disclosure proposed | Prior express written consent | Ruling in force; disclosure rule proposed |
Recording sits next to all of this. Many assistants record and transcribe, and some states require every party's consent to record; Florida, for example, permits recording only when all parties consent[10]. One opening sentence can carry both notices. State-by-state recording rules are in call recording consent rules by state.
Wording that satisfies the strictest rule
Writing to Utah's safe harbor covers the other states too: say what it is at the start, keep saying it where the conversation could be mistaken for a person, and answer truthfully when asked. The examples below are written for a fictional brokerage; adjust the names and keep the structure.
| Moment | Example wording | Rule it serves |
|---|---|---|
| Phone greeting | "Thanks for calling Example Realty. You've reached our automated assistant, not a person. Calls are recorded. I can answer questions about our listings or take a message for an agent." | Utah safe harbor; recording notice |
| Caller asks "Is this a real person?" | "No, I'm an automated assistant for Example Realty. I can get your question to an agent right now." | Utah consumer-transaction rule |
| Before financial questions | "Before I ask about your budget or pre-approval: I'm an automated assistant, and an agent will review everything you share." | Utah high-risk interaction |
| Website chat header | "Example Realty assistant (automated, not a person)" shown for the whole conversation | California bot law; Utah "throughout" |
| Outbound AI-voice call | "This is an automated call from Example Realty using a computer-generated voice, about the home you asked us to value." | FCC consent context; California dialing devices |
If an SI Employee answers your line, the greeting can carry the disclosure and recording notice you approve, word for word, and you sign off on every word before calls are forwarded. Hear how a call sounds before writing your own. The solo listing agent example shows where a greeting like this fits on a sign line.
A tracking routine for brokers
List every automated touchpoint
Phone answering, website chat, social messaging, outbound AI calls, automated texts and any tenant or buyer screening tool.
Tag each one by state
Where the consumers are: a Utah buyer chatting with a Texas brokerage's website still raises the Utah question.
Write one disclosure per channel
Use the strictest wording above for every state, so nobody has to remember which caller lives where.
Check the transcripts monthly
Pull five calls and five chats. Was the disclosure given, and was every "are you a person?" answered truthfully?
What the assistant is allowed to say about listings, prices and representation is a license-law question rather than an AI one; see what an AI phone assistant may say under license law. This guide is general information, not legal advice.
Frequently asked questions
What states require AI disclosure?
Utah requires it when a consumer clearly asks, and up front in high-risk interactions for licensed occupations[2]. California requires bots used online to sell to disclose, or face liability[3], and automated dialing calls to say when a voice is artificial[5]. Colorado requires notice for covered automated decision tools in consequential decisions such as housing[6].
Does Utah's AI law apply to real estate agents?
The regulated-occupation tier covers licensed occupations regulated by Utah's Department of Commerce, and applies when an interaction is high-risk, such as collecting financial data[2]. Utah licenses real estate agents through that department, so plan for it. Disclosing at the outset and throughout the conversation is a safe harbor[2].
Is there a federal law requiring AI disclosure on calls?
Not yet. The FCC has proposed a per-call AI disclosure rule but has not adopted it[8]. AI-voiced outbound calls already count as artificial voice calls under the TCPA[7].
Does a chat assistant on my website need a label in California?
California's bot law applies to websites, apps and social networks[4]. A clear, conspicuous statement that the assistant is a bot avoids liability under it[3].
Do I need to repeat the disclosure during a conversation?
Utah's safe harbor covers disclosure at the outset and throughout the interaction[2]. For chat, a persistent label does that. For calls, repeat it when asked and before any financial questions.
References
Pew Research Center (2025). How Americans View AI and Its Impact on Human Abilities, Society (opens in a new tab) ↑
Utah State Legislature (2025). S.B. 226 Artificial Intelligence Consumer Protection Amendments (2025 General Session, enrolled) (opens in a new tab) ↑
California.Public.Law (text of the California code) (2019). California Business and Professions Code § 17941 (bot disclosure) (opens in a new tab) ↑
California.Public.Law (text of the California code) (2019). California Business and Professions Code § 17940 (definitions) (opens in a new tab) ↑
California.Public.Law (text of the California code) (2025). California Public Utilities Code § 2874 (automatic dialing-announcing devices) (opens in a new tab) ↑
Colorado General Assembly (2026). SB26-189 Automated Decision-Making Technology (opens in a new tab) ↑
Federal Communications Commission (2024). FCC Makes AI-Generated Voices in Robocalls Illegal (opens in a new tab) ↑
Federal Communications Commission (Federal Register) (2024). Implications of Artificial Intelligence Technologies on Protecting Consumers From Unwanted Robocalls and Robotexts (proposed rule) (opens in a new tab) ↑
Zillow Research (2026). Sellers: Results from the Zillow Consumer Housing Trends Report 2026 (opens in a new tab) ↑
The Florida Senate (2025). Florida Statutes § 934.03: Interception and disclosure of wire, oral, or electronic communications (opens in a new tab) ↑