AI Verification Policy

Where we use AI

• Finding and summarising. AI tools help us monitor sources, surface candidate stories and draft first summaries of what a source says.
• Reader tools. Ask Franki and the AI tools on this site run on Anthropic’s Claude models. Ask Franki is instructed to answer only from the published FranchisePulse corpus, to cite the page a fact came from, and to say so when the corpus does not cover a question rather than invent an answer.

Where a person checks the work

• No figure is published because a model produced it. Every figure about a brand links to the page that states it, and that page is opened and compared with the figure before the figure goes live.
• The date that check was made is stored with the source and shown as the review date. When a source is re-checked, the date moves.
• Newsletter issues are approved by an editor before they are sent; an issue without an approval on record is not sent.
• Figures from vendors and brands are presented as their claims. We do not have access to their internal data and do not audit it.

How the AI Adoption Score is built

Each system in the Franchise AI Index is scored 0–100 on four pillars, then combined with fixed weights: operations AI (35%), data infrastructure (30%), marketing AI (20%) and training & support AI (15%). Evidence comes from public disclosures, product announcements, operator interviews and field reporting. The score is an editorial benchmark — our judgement of that evidence — not a number reported by the company.

• 70–100 · AI-led — centralised data platform and AI in production across operations and franchisee-facing tools.
• 50–69 · AI-active — AI live in marketing or forecasting; operations and training AI partial or piloting.
• 0–49 · AI laggard — point solutions or none; fragmented unit-level data.

The full method and the current industry rate are on the methodology page.

How to read the labels

• ✓ Verified — confirmed against a primary source: the company itself, its filing, or the association that compiles the figure.
• ⚠ Unverified — a claim in circulation that we could not confirm against a primary source. We publish it because it is being repeated, and label it so you do not repeat it as fact.
• Official / press — on a brand page, whether a source is the company’s own document or a media report that should be cross-checked.
• Source pending — we have not yet found a source for that figure. It is not an estimate and should not be cited.

When the AI is wrong

Models make mistakes, and so do editors. If you find an error — in a figure, a score or an answer from Ask Franki — tell us through the contact desk(“Correction or claim dispute”). Corrections follow our Editorial Standards. We do not use what readers type into our AI tools to train any model (see the Privacy Policy).