Franchise Chatbot Earnings Claims: What Your AI Is Allowed to Say About Income

$1.85 million is what a martial arts franchisor and its former franchise sales organization will pay to settle FTC charges announced in October 2026, which include making earnings claims that were not in the franchise disclosure document. Franchise chatbot earnings claims fall under the same federal rule: a chatbot may give a prospect a sales, income or profit figure only if that figure sits in Item 19 of your FDD and is presented with its basis. Any other number it volunteers is likely an unauthorized financial performance representation, which is the rule's term for an earnings claim.

This post covers the federal FTC Franchise Rule (16 CFR Part 436, as shown in the eCFR current through October 20, 2026). Some states add their own franchise laws, which are outside its scope.

What counts as a franchise chatbot earnings claim?

A chatbot answer is an earnings claim when it gives a prospect a specific figure or range of sales, income or profit, stated directly or by implication. The federal definition covers oral, written and visual statements, so a chat window, a voice agent and a payback calculator all qualify. How the answer is delivered does not change the analysis.

The definition of a financial performance representation reaches any statement to a prospective franchisee that states, expressly or by implication, a specific level or range of actual or potential sales, income, gross profits or net profits. It also names charts, tables and mathematical calculations that show possible results based on a combination of variables, which describes most ROI calculators.

The dissemination rule, 16 CFR 436.9(c), allows such a representation only when the franchisor has a reasonable basis and written substantiation for it at the time it is made, and the representation appears in Item 19. When Item 19 makes no earnings claim, the required statement in Item 19 says the franchisor does not authorize its employees or representatives to make any such representations, orally or in writing. A bot that answers prospects for the brand is arguably a representative in that sense.

What the October 2026 FTC case shows about off-document claims

The FTC alleges that the franchisor, Premier Franchising Group, and its former sales organization, Franchise Fastlane, told prospects without martial arts backgrounds they could profitably run one or more studios on a semi-absentee basis, working under 15 hours a week. The agency says more than 200 consumers paid an initial franchise fee of $49,500 or more, and that the companies made financial performance representations that were not in the FDDs.

The FTC also alleges that the earnings figures that were in the FDDs came from existing studios of 2,000 to 7,000 square feet, while new franchisees were advised to open 1,200 to 1,600 square feet, and that most existing franchisees had martial arts experience while most new ones did not. The orders are proposed and take effect only when a federal judge in the Eastern District of Tennessee signs them. These are allegations in a settlement, not findings.

That link to chatbots is an inference, not something the FTC said. A prospect who types "can I run this while keeping my job?" is asking for the semi-absentee answer, and a model trained on sales decks and webinars can produce it unprompted.

Who is responsible for what a franchise chatbot says?

The company that deploys a chatbot answers for it. In Moffatt v. Air Canada (2024 BCCRT 149), a British Columbia tribunal called the airline's claim that its chatbot was a separate legal entity responsible for its own actions a remarkable submission. The case is Canadian and involves an airline passenger, so it does not govern a US franchise sale.

It does show how a decision-maker treats the chatbot excuse, and the FTC action points the same direction on the franchising side. The agency charged both the franchisor and the sales organization, so outsourcing the conversation, whether to a salesperson or a software vendor, does not move the exposure off the brand's name.

A worked example: one acceptable answer and two risky ones

Assume a hypothetical Item 19 reports average gross sales of $1,200,000 for the 60 of 80 franchised units that were open all of last year. The 20 others are excluded. This is invented for illustration and is not any brand's disclosure.

An acceptable answer quotes the table as written: the 60 units, the $1,200,000 average, the definition of gross sales, the fact that 60 of 80 is 75 percent of units, and the statement that a new franchisee's results may differ.

The first risky answer drops the denominator: "Our owners average $1.2 million a year." If the 20 excluded units averaged $700,000, the real 80-unit average is (60 x $1,200,000 + 20 x $700,000) / 80 = $86,000,000 / 80 = $1,075,000. The bot overstated it by $125,000, or 11.6 percent, and the $700,000 figure appears nowhere in Item 19, so the franchisor could not substantiate the 80-unit number the bot implied.

The second risky answer does math the FDD never did: "At an 8 percent net margin, that is $96,000 a year." It is $1,200,000 x 0.08 = $96,000, and the margin is invented. That is a new profit figure with no reasonable basis, produced by a calculation, which the definition expressly covers.

A three-lane test for franchise chatbot earnings claims

Sort every question a prospect can ask into one of three lanes before the bot goes live. The lanes are a working framework, not an FTC requirement.

The first lane is quote. Questions about Item 19 itself ("what do your units earn?") get the current table verbatim with its basis, its date and the may-differ admonition the rule requires, plus a link to the FDD section. The bot adds no adjectives, no rounding and no "typical".

The second lane is hand off. Questions about a prospect's own outcome (hours, profit, payback, financing return) go to a human franchise development representative who has been trained to point back to Item 19. The bot offers a call and logs the question. The third lane is block. The bot never builds a projection, compares unit profit to a competitor, or uses the words "semi-absentee" or "passive" about ownership.

Run four checks on every answer in the log. Does it contain a figure or range of sales, income or profit? Is that figure in the current Item 19? Does it carry the basis and the admonition? Can you produce the written substantiation if asked? Keep the bot out of signing and payment talk as well, because the 14-day waiting period carries its own timing rules.

What to pull from your chatbot logs this week

Search 90 days of transcripts for dollar signs, "per year", "profit", "payback", "ROI", "hours a week" and "absentee", then have counsel review every hit. Check what the bot retrieves from: if its knowledge base includes sales decks, webinar transcripts or last year's FDD, it can repeat numbers that no longer match the current Item 19. Item 19 tables often stop at revenue, a gap covered in Item 19 tables that stop at revenue, so a bot asked about profit has nothing to quote and may invent it.

Voice agents add a second layer, because the same call carries consent and recording rules, covered in the recording-consent side of an AI voice agent. If you run a candidate-facing bot or are replacing one, Revscale builds AI chatbot agents for franchise lead conversations; whichever tool you pick, test it against the three lanes before launch.

Review franchise chatbot earnings claims the way you review any other earnings document: substantiation first, launch second, and a saved transcript of what the bot said. This is general information, not legal, tax or financial advice; confirm specifics with a qualified advisor.