Savings Claims, the Tax Credit & Financing: A 2026 Solar Playbook

Solar is sold on future numbers more than almost any home service. The pitch is a projection: this is what you will save, this is the credit you will get, this is your monthly payment. Those numbers close deals, and they also generate the complaints, the investigations, and the lawsuits that follow this industry. So when an installer puts an assistant on the website, the risk is not abstract. The assistant lands directly on the three numbers that get companies in trouble.

This guide is the companion to the threat side of that story. The threat piece covers what goes wrong when an unguarded chatbot guarantees savings or confirms the tax credit. This one is the standard: what a compliant deployment looks like for a solar company in 2026, and the lines the system has to hold.

A Savings Claim Has to Be Substantiated

Start with savings. A promise that solar will "eliminate your bill" or save a specific amount is a performance representation, and an unsubstantiated one falls within the FTC’s authority over unfair and deceptive practices. Real savings depend on usage, roof, rate plan, net-metering rules, and the system actually installed, none of which a website assistant can evaluate. The compliant position is that savings are estimated during a real assessment, not promised in a chat. The assistant can explain that an estimate comes from a consultation and book it, rather than committing the company to a figure no engineer validated.

The Tax Credit Is Tax Advice

The federal Residential Clean Energy Credit is real, but whether a specific household can use it turns on ownership, financing, and the homeowner’s tax liability. Telling a visitor they "qualify for the 30% credit" is an individualized tax determination, and the assistant has neither the information nor the standing to make it. The compliant line is to note that a federal credit exists and direct the homeowner to a tax professional, never to confirm eligibility. A sale closed on a credit the homeowner cannot actually use becomes a complaint about being misled.

Financing Numbers Carry Lending Rules

Solar is often sold on financing, and the terms are governed by lending law. A monthly payment, an interest rate, or an effective price are credit representations that the Truth in Lending Act and Regulation Z regulate, and the industry has already drawn scrutiny for gaps between a quoted cash price and a financed price padded with fees. A compliant assistant does not improvise financing figures. It routes payment and financing questions to the people who present terms properly.

The 2026 Compliance Standard, Line by Line

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

• No savings guarantees. Savings are estimated in a real assessment, so the assistant books the consultation instead of promising a number.
• No tax-eligibility confirmation. The assistant notes the credit exists and points to a tax professional; it never tells a homeowner they qualify.
• No financing quotes. Payment, rate, and effective-price questions route to a person who presents terms under the lending rules.
• No production or payback promises. Output and payback depend on the design and the site, so those are assessment outcomes, not chat answers.
• Every exchange is logged, so what a homeowner was told is reviewable.

The pattern is the one that runs through every regulated vertical. The assistant answers what carries no obligation, how the process works, what an assessment involves, service areas, general questions about solar, and routes savings, tax, financing, and production specifics to qualified 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 guarantee savings, never to confirm the credit, never to quote financing. Call the boundary set.

It is not, because of how the model reads a request. It follows an instruction when the question matches the wording it was warned about and slips when the phrasing changes. You tell it never to confirm the credit. A homeowner asks, "so I get thirty percent back, right?" The model hears a simple yes-or-no and answers to be agreeable. The instruction was loaded the whole time. It just did not recognize 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 it from speaking. A real boundary is built into the system and decides what the assistant may say before it answers, so a savings guarantee or a tax confirmation never reaches the homeowner 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 site with a contact form. It means deploying an assistant that captures interested homeowners and books assessments without guaranteeing savings, confirming tax eligibility, or quoting financing.

Fred is built that way. It answers from your own content, explains how an assessment works, captures the lead, and routes savings projections, tax-credit eligibility, financing, and production estimates to the people qualified to give them. It runs more than 50 industry guardrail packs, and the solar pack is built around substantiation of savings claims, the line between noting a tax credit and advising on it, and the lending rules around financing. Fred does not promise a homeowner their bill will vanish or that they qualify for the credit. It cannot. It books the consultation and hands the numbers to your team.

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

Frequently asked questions