Does an AI chatbot on a franchise website make an earnings claim?

It can, and the analysis does not turn on whether a human typed the sentence. A financial performance representation is information given to a prospective franchisee, orally, visually or in writing, from which a specific level or range of actual or potential sales, income, gross profits or net profits may be inferred. An assistant that answers 'what do your locations make?' with a number, a range, a payback period or even a strongly implied ordering of outcomes has supplied exactly that. If the same figure does not appear in Item 19 of the disclosure document, the franchisor has made a representation outside the document it is required to make it in.

Is an ROI or payback calculator a financial performance representation?

A calculator that takes a prospect's inputs and returns projected revenue, profit, break-even month or payback period is one of the clearest examples. The output is a specific level of potential income for that prospect. Two features make calculators worse than a paragraph of prose: the numbers are personalised, which strengthens the inference that they describe this prospect's outcome, and the underlying assumptions are usually invisible, which makes the required disclosure of material bases and assumptions difficult to satisfy inside the interface itself.

What does the substantiation requirement mean for a generated number?

A franchisor must have a reasonable basis and written substantiation for a financial performance representation at the time it is made, and must furnish that substantiation to a prospective franchisee on reasonable request. A language model does not create a reasonable basis; it produces a plausible sentence. If the number came out of a model's synthesis of marketing copy, a blended figure it inferred from context, or a hallucinated benchmark, there is no substantiation file to produce — and the request for substantiation is precisely how the absence gets discovered.

We only publish a system-wide average. Can the AI restate it?

Restating an Item 19 figure is permissible; reshaping it is where the exposure begins. Common failure modes are the assistant applying the average to a prospect's specific market, extrapolating a monthly figure to an annual one, dropping the cohort description that made the average meaningful, converting revenue into an implied profit, or omitting the accompanying disclosures the document attaches to the number. The output is then a new representation that no longer matches what was disclosed, which is the practical definition of an unlawful earnings claim.

Are franchise brokers and lead-generation vendors using AI our problem?

Broadly yes. Franchise sellers acting on a franchisor's behalf are within the disclosure regime, and many state franchise statutes reach both the seller and the franchisor for representations made in the sales process. Where a broker network runs AI qualification chats, auto-generated 'candidate fit' summaries or automated follow-up sequences that quote performance, the franchisor's practical position is that it will be asked to produce substantiation for numbers it never drafted. Contractual prohibitions on out-of-document earnings claims are necessary but are not a defence to a prospect who received one.

What is the cheapest control that actually reduces this risk?

A hard refusal boundary plus a logged transcript. The assistant should be constrained to decline any performance question and route it to the disclosure document rather than attempting a compliant answer, because a model asked to summarise a financial disclosure will reliably drop qualifiers. Retain the full conversation log with timestamps, because when a franchisee later alleges an oral misrepresentation, the transcript is either the franchisor's best evidence or the plaintiff's best exhibit, and which one it is was decided when the system was configured.

Your Chatbot Made a Financial Performance Representation

Six Surfaces, One Question

The useful exercise is not "do we make earnings claims" — franchisors know the answer to that. It is an inventory of every surface where a number can reach a prospect, and a verdict on each. Automation changes the inventory rather than the rule: the surfaces multiply, and most of them were built by a growth team that has never seen an FDD.

Surface

What it emits

Verdict

Why

Website sales assistant

"Most of our owners are profitable inside the first year."

FPR

A level of potential profit and a timeframe can be inferred. Outside Item 19 unless the document says exactly this.

ROI / payback calculator

Prospect enters a market and headcount; tool returns a break-even month.

FPR

Personalised projection of potential income. Bases and assumptions must be disclosed, and rarely are inside the widget.

Generated territory pro forma

A one-page model built from population, competitor density and a revenue-per-capita assumption.

FPR

Specific potential sales for a named territory. The revenue assumption is the substantiation question in one number.

AI-summarised validation calls

"Existing owners describe strong unit economics and quick ramp."

Risky

Aggregating franchisee statements into a franchisor summary can convert their anecdotes into your representation.

Automated follow-up email

"Based on your market we'd expect you in the upper half of the system."

FPR

A range inference tied to this prospect. Sequence copy is generated once and sent thousands of times before anyone reads it.

Item 19 restated verbatim

Quotes the disclosed figure, cohort, period and accompanying language in full, then points to the document.

The safe pattern. The failure mode is paraphrase, not quotation — models paraphrase by default.

Where in the Sale It Goes Wrong

Earnings-claim disputes are rarely about the disclosure document. They are about the eleven weeks before and after it, where automated systems talk to a prospect far more often than any human does — and where the record of what was said is either kept or quietly discarded.

First contactHighest

Lead form, chat widget or broker portal. No disclosure document has been delivered, and this is where the majority of AI-generated performance language is emitted.

QualificationHigh

Net-worth and liquidity screening, often with an automated 'you'd qualify for a two-unit deal' framing that implies unit-level economics.

Disclosure deliveryContradiction risk

The FDD is issued with whatever Item 19 contains — or with Item 19 stating that no representation is made, which contradicts every number already sent.

Discovery dayHigh

Presentations, slide decks and Q&A. Generated decks reintroduce numbers that legal removed from the document.

Waiting periodSilent drift

Follow-up sequences run automatically. Nobody re-reads campaign copy that was approved months earlier against the current disclosure document.

Signature and afterEvidence

A questionnaire asks the franchisee to confirm no representations outside the document were made. Transcripts saying otherwise still exist.

"No Financial Performance Representations" Is the Hardest Position to Hold

Many franchisors choose not to make a representation at all, and Item 19 says so. That choice is coherent on paper and extremely fragile in an automated sales channel, because it converts every stray number anywhere in the funnel into a direct contradiction of the disclosure document. A single calculator left live from a previous campaign, or one assistant reply estimating a payback period, is worse for a no-representation franchisor than for one whose Item 19 contains real data.

The corollary is that the decision about Item 19 is also a decision about what your website, chat tooling and broker network are permitted to generate. Those are usually made by different people, months apart, with no shared record.

The Substantiation File Is the Whole Game

Generated numbers fail a specific test rather than a vague one. Asked for the basis of a figure, a franchisor must produce it — a defined cohort of outlets, a stated period, the accounting treatment, the exclusions, and the reason the cohort is a fair comparison for this prospect. Model output has none of these properties; it has fluency. The gap between the two is invisible in the sales conversation and total in a written response to a request for substantiation, which is why the request is such an effective discovery tool for the other side.

Six Controls Worth Having

Refuse, don't summarise

Configure every prospect-facing assistant to decline performance questions outright and link to the disclosure document. A model asked to be helpful about earnings will be helpful about earnings.

One source of truth for numbers

If any figure is permitted, it should be injected from a maintained Item 19 record with its cohort and period attached, not retrieved from marketing pages or the model's own recall.

Substantiation file per figure

Every number the sales channel can emit needs a written basis on file before it is emitted, held to the standard of being produced on request rather than reconstructed after a demand letter.

Retain and review transcripts

Log conversations with timestamps and sample them against a keyword list — profit, revenue, payback, ROI, break-even, 'make', 'earn'. Sampling is what turns a control into evidence that it worked.

Extend the rules to brokers

Contract terms prohibiting out-of-document representations should name AI tools explicitly, require log retention, and give you audit rights over the prompts and sequences used on your brand.

Version-lock every asset

Decks, calculators, sequences and site copy should carry the FDD version they were cleared against, so an Item 19 amendment forces a review of everything downstream of it.

Find the Numbers Still Live on Your Site

Payback calculators, "average unit volume" pages, testimonial quotes with figures in them and old campaign landing pages outlive the disclosure documents they were cleared against.

See every claim your site is making in one pass. Run a free scan and check each figure against your current Item 19.

This article is general information and not legal advice. Franchise sales are governed by federal disclosure requirements and by separate registration and anti-fraud statutes in a number of states, whose requirements differ. Consult qualified franchise counsel before relying on any conclusion here.