Allergens, Access & Accuracy: A 2026 Restaurant Playbook

A restaurant website looks like the friendliest place to drop a helpful assistant. Hours, menu, reservations, easy wins. The trouble is that the same chat window also catches the questions that are not casual at all: is this safe for my allergy, can I bring my service animal, do you have a table at eight. Those touch food safety, accessibility law, and basic accuracy, and a general-purpose assistant answers them with the same cheerful certainty it uses for the parking situation.

This guide is the companion to the threat side of that story. The threat piece covers what goes wrong when an unguarded chatbot clears an allergy or invents a reservation. This one is the standard: what a compliant deployment looks like for a restaurant in 2026, and the lines the system has to hold.

Allergens Are a Safety Question, Not a Menu Question

Start with the answer that can actually hurt someone. Telling a guest a dish is "safe" for a serious allergy is a representation about cross-contact in a kitchen the assistant cannot see. Food-safety regulation treats allergen handling seriously, and most state food codes, built on the FDA model, expect staff to manage allergen information with care rather than guess. A false "it’s safe" is also a misrepresentation, and the FTC’s authority over deceptive practices reaches misleading claims to consumers. The compliant position is that the assistant never clears a dish for an allergy. It takes the allergy seriously, declines to guarantee safety, and routes the guest to staff who can check with the kitchen.

Accessibility and Service Animals Are Legal Answers

Restaurants are public accommodations under Title III of the Americans with Disabilities Act, and the rules on service animals and accessibility are specific. A confident but wrong answer, like turning away a service animal as a pet, conflicts with the law and creates a written record of it. A compliant assistant gives the correct general rule on service animals and accessible seating and routes specifics to staff, rather than improvising a policy that contradicts the law.

Reservations, Prices, and Alcohol Are Commitments

The everyday answers carry their own weight. Confirming a reservation the assistant cannot actually place, promising a table on a fully committed night, quoting a private-event price no manager approved, or telling someone underage the bar will serve them, each is a small promise or a regulated statement. Alcohol service and age rules are governed by state law and carry real penalties for the license. A compliant assistant does not confirm bookings it cannot verify, quote event pricing on its own, or make representations about alcohol service.

The 2026 Compliance Standard, Line by Line

A compliant restaurant assistant is defined by what it is built to refuse. Treat the list below as the floor.

• No allergy clearances. The assistant never says a dish is safe for an allergy; it routes the guest to staff who check with the kitchen.
• Correct accessibility answers. Service-animal and accessibility questions get the right general rule, with specifics routed to staff.
• No phantom reservations. The assistant captures a request and lets your system or staff confirm what is actually available.
• No unapproved pricing. Private-event and catering pricing route to a manager.
• No alcohol-service promises. Age and service questions are answered as state-governed, not guaranteed.
• Every exchange is logged, so what a guest was told is reviewable.

The pattern is the one that runs through every regulated vertical. The assistant answers what carries no risk, hours, location, general menu information, how reservations work, and routes allergy, accessibility specifics, bookings, pricing, and alcohol to people.

Why an Instruction Cannot Meet the Standard

The usual shortcut is to write these rules into the assistant’s prompt. Tell it never to clear an allergy and never to confirm a table. Call the boundary set.

It is not, because of how the model handles a question phrased a little differently than expected. You tell it never to say a dish is allergy-safe. A parent asks, "there’s no nuts in this, right?" The model hears a simple yes-or-no and reassures them. The instruction was loaded the whole time. The wording just did not match what it was told to refuse.

That is the difference between an instruction and a standard. An instruction asks the model to behave; it does not stop it from speaking. A real boundary is built into the system and decides what the assistant may say before it answers, so an allergy clearance or a false reservation never reaches a guest no matter how the question is framed. "Will not" is a suggestion. "Cannot" is an architecture.

What a Compliant Deployment Looks Like

Meeting the 2026 standard does not mean a static menu page. It means deploying an assistant that handles the genuine convenience questions while sending the ones with real stakes to your team.

Fred is built that way. It answers from your own content, helps guests with hours, menu, and how booking works, and routes allergy questions to the kitchen, accessibility specifics and reservations and event pricing to staff. It runs more than 50 industry guardrail packs, and the restaurant pack is built around allergen safety, ADA service-animal and accessibility rules, and not making bookings or pricing promises it cannot keep. Fred does not tell a guest a dish is safe for their allergy. It cannot. It answers what it should and hands the rest to people who can stand behind it.

That is the difference between hoping the assistant does not clear an allergy and being able to show why it cannot.

Frequently asked questions