Guaranteed Savings and the 30% Credit: Two Numbers Your Website Can’t Promise

A homeowner browsing your solar site at night asks the chat the two questions everyone asks: how much will this save me, and do I get the thirty percent tax credit? The assistant, built to move the conversation toward a quote, answers cleanly. It says the system will "eliminate your electric bill" and confirms the homeowner "qualifies for the 30% federal credit." Both answers feel like good salesmanship. Both are claims the company now has to stand behind, and one of them is tax advice the assistant has no business giving.

Solar is one of the most scrutinized sales environments in home services, precisely because the pitch lives on future numbers. Savings projections, payback periods, and tax incentives are the heart of the sale and the heart of the complaints. A general-purpose chatbot does not know which of those numbers it is allowed to promise, so it promises all of them.

A Savings Claim Is a Claim You Have to Substantiate

Start with savings. A blanket promise that solar will "eliminate your bill" or "save you X every month" is a performance representation, and an unsubstantiated one falls squarely within the FTC’s authority over unfair and deceptive practices. Real savings depend on usage, roof, rate plan, utility net-metering rules, and the system actually installed. None of that is known to a bot fielding a late-night question. When it guarantees an outcome, it has made the kind of claim that regulators have repeatedly pursued in this exact industry, and that a homeowner will quote back when the first true-up bill arrives.

The homeowner who signed because the website said "no more electric bill," and who still has one, is not just disappointed. They have a screenshot of a promise no engineer ever validated.

"You Qualify for the Credit" Is Tax Advice

The tax credit is the sharper trap. The federal Residential Clean Energy Credit is real, but whether a specific household can use it depends on facts only a tax professional should weigh: ownership of the system, tax liability, how the purchase is financed, and the homeowner’s own return. When an assistant tells a visitor they "qualify for the 30% credit," it has given an individualized tax determination. If that homeowner has too little tax liability to use it, or leases rather than buys, the number that closed the sale turns into a complaint about being misled.

Confirming eligibility is not a friendly recap of a public incentive. It is advice about that person’s taxes, made by software with no view of their finances and no license to give it.

The Financing Numbers Carry Their Own Rules

Then there is the monthly payment. Solar is often sold on financing, and the terms are governed by lending law. A bot that quotes a monthly payment, an interest rate, or an effective price is making credit representations that the Truth in Lending Act and Regulation Z regulate. Solar financing has already drawn attention for the gap between a quoted cash price and a financed price padded with fees. An assistant improvising payment figures is wandering into that exact disclosure regime with no idea it exists.

"Will Not" Is a Suggestion. "Cannot" Is an Architecture.

The standard reaction is to write rules into the bot. Tell it never to guarantee savings, never to confirm the credit, never to quote financing. That reads like a boundary. It is not one, because of how a model treats a question it was not specifically warned about.

You tell it never to promise the credit. A homeowner asks, "so I get thirty percent back, right?" The model hears a yes-or-no question with an obvious friendly answer and gives it. The instruction was loaded the entire time. The phrasing just did not match what it was told to refuse. A system built to produce confident, closing answers will keep producing them, and a disclaimer in the footer does not unsay the sentence the homeowner already screenshotted.

Who Owns the Number

Strip away the technology and the exposure is ordinary. A new sales rep who told every prospect their bill would vanish and that they definitely qualified for the tax credit would be a liability the company would coach hard, fast. When a website assistant says the same things at scale, the responsibility does not disappear. It lands on the company, with a written record, no engineer or tax professional in the loop, and consumer-protection and lending regimes that govern exactly those claims.

The installers who get hurt are not the ones who added an assistant to the site. They are the ones who dropped in a generic chatbot, assumed enthusiastic and compliant were the same thing, and learned otherwise when a homeowner’s lawyer pulled the transcript. The fix is not to take AI off the site. It is to run an assistant that captures the lead and books the consultation without guaranteeing savings, confirming tax eligibility, or quoting financing, and that hands those numbers to the people qualified to give them.

Frequently asked questions