Reg Z, the Buyers Guide, and Your Dealership Website: A 2026 Playbook

For most of the web era a dealership website was a brochure with an inventory feed. Nobody read a stock photo as a binding promise. An assistant that talks back changes that, because the moment your site answers a question about price, payment, or coverage, it is making the kind of statement the law has policed on the lot for decades. The regulators who watch auto retail did not write a carve-out for software, and in 2026 they are reading chat transcripts the same way they read a newspaper ad.

This is the companion to the threat piece. That one walks through the answer that becomes a complaint. This guide covers the standard: what a compliant assistant on a dealership site is allowed to say, what it has to route to your desk, and how to deploy one without signing the store up for every number it produces.

The Website Speaks as the Dealer

Start with the principle that decides everything else. A representation made on your site, by a human or by a widget, is the dealership’s representation. There is no separate, lower standard for things a bot says. That is why the friendliest answers carry the most exposure: a payment figure, a warranty reassurance, an out-the-door estimate. Each one looks like customer service and lands like a regulated statement.

Truth in lending is the clearest example. Once a specific monthly payment or a rate appears, the site has advertised a credit term, and Regulation Z’s advertising rules require a defined set of disclosures to ride along with that figure. A bare "about $399 a month" is not a friendly estimate in the eyes of the rule. It is a triggering term missing its disclosures, and the dealership owns the gap.

Warranty Language Is a Promise the Store Honors

Federal law splits the used-car world cleanly down the middle: a vehicle is sold "as is" or it carries a warranty, and the FTC’s Used Car Rule requires the Buyers Guide on the glass to say which. An assistant that tells a shopper a unit is "still covered" can contradict that disclosure outright. Worse, a written promise of coverage can pull the deal under the Magnuson-Moss Warranty Act, turning an offhand reassurance into an enforceable obligation nobody at the store agreed to.

A compliant assistant treats coverage as a document, not a vibe. It points the shopper to the Buyers Guide and the actual warranty terms, and it hands any "is this one covered?" question to a person who can read the file. The same caution applies to anything the site cannot verify. Volunteering mileage or condition details it has not confirmed walks a dealership toward the federal odometer disclosure rules and toward a misrepresentation claim if the number is wrong.

What AI Compliance for Car Dealerships Actually Requires

Set against that backdrop, the 2026 standard is mostly a list of refusals. A compliant dealership assistant does not quote a payment or a rate, because that figure needs Reg Z disclosures and a look at the buyer’s credit it does not have. It does not confirm warranty coverage on a specific unit. It does not produce an out-the-door price that folds in tax, title, and doc fees it cannot calculate for a real deal, since a number that lands lower than the truth reads as exactly the sort of bait the FTC’s authority over deceptive practices exists to catch. Trade-in "values" get the same treatment, because a figure no appraiser validated still sets the expectation the negotiation starts from.

What it does is the wide, useful surface that needs no F&I review. Hours and directions. Whether a model is in stock. What a test drive involves. The general shape of the financing process, with the actual numbers booked for the desk. A clean capture of the lead so a salesperson picks up a real conversation. That is a genuinely helpful job. It simply stops at the line where a sentence becomes an advertised term.

One accuracy note worth stating plainly: the standard here does not lean on the FTC’s CARS Rule. That rule was vacated by the Fifth Circuit in early 2025 and is not in effect, so a 2026 program builds on the durable framework instead, the advertising rules, the Used Car Rule, Magnuson-Moss, and the FTC Act’s general prohibition on deception.

Why an Instruction Will Not Hold the Line

The shortcut every vendor reaches for is the prompt. Write "never quote financing, never confirm coverage" into the assistant’s instructions and treat the boundary as set.

It is not set, and the reason is mechanical. A language model obeys an instruction when the request looks like the wording it was warned about, and shoppers rarely use that wording. You tell it never to quote a payment. The shopper does not ask for a quote. They lean in: "ballpark, what would something like this run me a month?" The model hears a reasonable person asking for a rough number and gives one, because being helpful is its default and a prompt is only a request to suppress that default. The rule was loaded the whole time. It just never recognized the sentence that crossed the line.

That is the difference between an instruction and a standard. An instruction asks the model to behave. It does not stop the model from speaking. A real boundary is built into the system and decides what the assistant is allowed to say before it answers, so a payment quote or a coverage promise never reaches the shopper no matter how the question is phrased. "Will not" is a suggestion. "Cannot" is an architecture.

What a Compliant Deployment Looks Like

Meeting the standard does not mean pulling the assistant off the site. It means running one that was built to know which statements are regulated and to keep those with the people who answer for them.

Fred is built that way. It answers from your own inventory and store content, captures and scores the lead, and routes anything touching price, payment, rate, trade value, or warranty to your sales and F&I team. It runs more than 50 industry guardrail packs, and the automotive pack is built around advertised credit terms, the Buyers Guide line, and the claims a dealership cannot make in writing. Fred does not quote a payment or promise coverage. It cannot. It handles the everyday questions, logs every exchange, and books the regulated work for the desk.

The aim is not a chattier website. It is one you can show a regulator was incapable of advertising a credit term or a warranty it was never authorized to offer.

Frequently asked questions