AI chatbots for lead capture: worth it, once you disclose them

Short answer: Yes, a chat bot is worth it for a US small business that gets inquiries outside office hours, because it answers and takes details while a form just waits. The risk is not the bot. It is that most chat tools are run by an outside vendor that stores the transcript, and California’s wiretap law, which carries US$5,000 per violation, has been used to sue over exactly that when nobody told the visitor. Say it is automated, say it is recorded and who handles it, and offer a person.

Most advice about chat bots comes from the companies selling them, so it stops at conversion lifts. The honest version has two halves: a bot can catch leads your form loses, and a bot you did not disclose can cost more in one demand letter than a year of the software. The second half is cheap to fix.

Below: what a bot catches, the California wiretap claim that has followed chat windows, what a recent ruling and a new bill do and do not change, and the three lines that belong above the first message. For the wider system, start with our guide to AI lead generation in the US. This is general information, not legal advice, so confirm the details with a lawyer who knows your state.

What a chat bot catches that a contact form loses

A form collects details and then waits until a person reads it. That is a problem at 9 pm on a Saturday, when the visitor is comparing three businesses and will talk to whichever one answers first.

A bot answers the two or three questions people ask before they commit: do you serve my area, roughly when can you come, do you handle this kind of job. Then it takes a name and number and can offer a booking slot at 9 pm. Its value is highest for businesses whose inquiries arrive evenings and weekends.

The conversion figures chat vendors publish are marketing. Measure your own: chats started, contact details left, jobs booked.

Missed-call text-back and AI answering usually pay back before chat does. We cover them in the AI lead tools ranked by risk. If the phone is not fixed, fix it first.

Why most visitors would still rather reach a person

Gartner’s press release of July 9, 2024 reported a survey of 5,728 customers, conducted in December 2023. It found that 64% would prefer that companies did not use AI in customer service, and 53% would consider switching to a competitor if they found out a company was going to use it. The top concern was that it would become harder to reach a person.

Be fair about what that survey is. It asks about customer service, not sales inquiries, and it predates today’s tools. It does say what people fear: a wall. So the bot must never be one. A visible route to a human is the feature that makes the bot acceptable to the people who would rather not use one.

The 1967 wiretap law now aimed at chat windows

The California Invasion of Privacy Act (CIPA) dates from 1967 and was written for telephone wiretapping. Its section 631 is the wiretapping provision. Plaintiffs’ lawyers have used it against ordinary website chat windows.

The theory: most chat tools are supplied by an outside vendor whose servers carry and store the conversation. So the vendor is cast as a third party listening in without the visitor’s consent, and the website owner as having helped.

The money is in the remedy. Penal Code section 637.2 lets an injured person sue for US$5,000 per violation or three times actual damages, whichever is greater, and says the plaintiff does not need to have suffered actual damages. A flat sum with no harm to prove is why a chat window looks like a target.

These suits have been filed against businesses of many sizes and are often aimed at a quick settlement. Like the bot-disclosure law, the claim turns on the visitor being in California, not on where the business is.

What the Converse ruling changed, and what it left open

On July 9, 2025, the Ninth Circuit decided Gutierrez v. Converse Inc., affirming summary judgment for Converse in a case over its website chat feature. The plaintiff had no evidence that the chat vendor actually intercepted and read messages in transit. The court said being capable of intercepting a communication is not the same as intercepting it. With no violation by the vendor, there was nothing for Converse to have aided.

A concurring judge wrote that section 631(a), written in 1967, was never intended to apply to internet communications. A law firm summary walks through the ruling.

That is good news, but do not oversell it:

• It turned on that plaintiff’s lack of evidence, so it raised the bar rather than closing the door.
• The concurrence is one judge’s view, not the holding.
• State courts are not bound to follow it.
• Winning at summary judgment still means paying lawyers to get there.

The ruling makes these cases harder to win, not impossible to file. A disclosure shown before the first message avoids the argument more cheaply than a defense.

Did California’s new SB 690 fix this? Not for chat

You may have seen headlines saying California “reformed CIPA.” Governor Newsom signed Senate Bill 690 on September 30, 2026 (Chapter 976). Read the bill text before assuming it covers you.

It amends only Penal Code section 637.2. It adds that an action against a private actor for a violation of section 638.51, the pen register and trap and trace provision, arising from conduct on a website, online application or mobile app, may be brought only by the Attorney General. That provision is the theory used against tracking pixels and analytics. The bill applies to pending claims in actions commenced within two years before it takes effect, and it takes effect January 1, 2027.

It does not amend section 631. Earlier versions carried a broad “commercial business purpose” exemption that would have reached further, and that did not survive into the final text. So the headlines are about pixels, not chat, and the chat-window theory is untouched.

Two states that make the bot introduce itself

A separate family of laws is about honesty over who is talking.

California’s BOT Act, Business and Professions Code section 17941, took effect July 1, 2019. It makes it unlawful to use a bot to communicate online with a person in California with intent to mislead them about its artificial identity in order to incentivize a sale. A clear and conspicuous disclosure avoids liability. Our guide to the wider AI disclosure rules covers it in full.

Maine enacted Title 10, section 1500-DD in 2025. A person may not use an AI chatbot in trade and commerce in a manner that may mislead or deceive a reasonable consumer into believing they are engaging with a human being, unless the consumer is notified in a clear and conspicuous manner that they are not. A violation is a violation of the Maine Unfair Trade Practices Act.

Bot-disclosure laws are about who is talking. The wiretap claim is about who else is listening. One line fixes the first, and a different line fixes the second. You rarely know which state a visitor is in, so disclose to everyone.

Three lines to put above the first message

Treat this as a baseline to confirm with a lawyer, not a guaranteed defense. Show all three before the visitor types anything, because consent after the fact is the weak point.

You are chatting with an automated assistant.

This chat is recorded and handled by our chat provider. Privacy policy: (your link).

Want a person? Call us at (your number) or ask for a human at any time.

Then the housekeeping:

• Name the chat vendor category in your privacy policy.
• Ask the vendor in writing whether it uses transcripts for its own purposes, such as training its models or advertising. If it does, turn that off.
• Do not let the bot ask for health details, payment card numbers or Social Security numbers.

Texting is the last trap. A phone number typed into a chat to get a callback about this inquiry is not blanket permission for promotional texts. The consent rules are in our guide to marketing automation for small business in the US.

A hypothetical garage door company in Reno, worked through

This is a hypothetical, with invented round numbers for illustration only, not benchmarks. A garage door repair company in Reno, Nevada also takes jobs across the state line in Truckee, California, which is why California law reaches a Nevada business. The site gets 1,200 visitors a month, and 3% open the chat: 36 chats. A third leave a name and number: 12 inquiries. One in four books: 3 jobs. At an average job of US$450, that is US$1,350 a month. The chat tool costs US$100 a month, so the owner is US$1,250 ahead.

Now the other side. One demand letter citing US$5,000 per violation would wipe out several months of that gain. That is why the three lines go in on day one, before the first Truckee visitor types. Then the owner tracks the three real numbers for a month and decides from their own data whether the tool stays.

When a chat bot is the wrong purchase

Skip it, or wait, in these cases:

• Low traffic. A few hundred visitors a month means a handful of chats. Fix the phone and the form first.
• Nobody will follow up. If no one reads the transcripts and contacts the lead within the hour, the names go cold.
• The phone goes unanswered. Answering comes first.
• Regulated intake. Medical, dental or legal intake, where what people type is sensitive, needs advice first. Health details bring their own rules.
• No human route. If you cannot offer a way to reach a person, the bot is the wall visitors fear.

If a bot fits and you want it set up with the disclosure built in, see our AI automation services or start with a free audit.

Frequently asked questions

Is a chat bot better than a contact form for capturing leads?

For a business that gets inquiries outside office hours, usually yes. A bot answers the two or three questions people ask before they commit and takes a name and number, while a form just waits for someone to read it. Do not trust the conversion figures chat vendors publish. Measure your own: chats started, contact details left and jobs booked.

Can my business be sued over a website chat window if we are not in California?

Yes, it is possible. Like the bot-disclosure law, the wiretap claim turns on the visitor being in California, not on where your business is. California Penal Code section 637.2 lets an injured person sue for US$5,000 per violation or three times actual damages, whichever is greater. This is general information, not legal advice, so ask a lawyer about your own setup.

Does SB 690 stop website chat wiretap lawsuits?

No. SB 690, signed September 30, 2026 and effective January 1, 2027, amends only Penal Code section 637.2. It limits suits over section 638.51, the pen register provision used against tracking pixels, to the Attorney General. It does not amend section 631, the wiretapping provision used against chat windows, so the chat theory is untouched. Broader exemption language in earlier drafts did not survive.

What should a chat window tell visitors before the first message?

Three things: this is an automated assistant, this chat is recorded and handled by our chat provider with a link to the privacy policy, and how to reach a person, such as a phone number or an offer to ask for a human at any time. Show them before the visitor types anything. Treat it as a baseline to confirm with a lawyer, not a guaranteed defense.

Can the bot text a visitor who typed a phone number into the chat?

Not for promotions. A phone number typed into a chat to get a callback about that inquiry is not blanket permission for promotional texts. Use it for the callback the visitor asked for, and get clear, separate consent before sending marketing messages. The consent rules for texts are covered in our guide to marketing automation for small business in the US.

Should a small business use a chat bot or live chat with a person?

A person is better when someone is reliably there to answer. A bot covers the hours when nobody is. The workable setup is a bot that hands over to a person when one is available. Either way, the recording disclosure is the same, because the chat vendor still carries the conversation and stores it.