Can ChatGPT analyze franchise FDDs?

On its own, only if you paste in the PDF, and then it often misreads tables or invents numbers. Connected to FranDB's FDD database over MCP, ChatGPT can pull verified data from 2,488 FDD filings: fees, Item 19 earnings, outlet growth, litigation, and franchisee contacts. The difference is whether the numbers are extracted or guessed.

Does ChatGPT have current franchise data?

No. ChatGPT's training data has a cutoff and doesn't include the latest FDD filings, which franchisors update every year. If you ask it about a specific franchise's fees or earnings, it will often give you outdated or made-up figures. Connecting it to a live FDD database is the only way to get current numbers.

How do I connect ChatGPT to franchise data?

Open ChatGPT settings, go to Connectors, enable Developer Mode, and add FranDB as a custom connector using the MCP server URL. Sign in with your FranDB account and authorize. After that, FranDB's tools appear in your chats and ChatGPT can query franchise data directly. Setup takes about a minute.

Which ChatGPT plan do I need?

Custom MCP connectors are available on ChatGPT Plus, Pro, Team, and Enterprise. You'll also need a paid FranDB plan for the data access. The free ChatGPT tier doesn't support custom connectors.

How to Research Franchises with ChatGPT (Using Real FDD Data)

ChatGPT is great at explaining how franchises work and terrible at telling you what a specific franchise actually costs. Ask it for Subway's current franchise fee or a brand's Item 19 numbers and you'll get a confident answer that's often wrong, pulled from old training data or invented on the spot. The fix is to give ChatGPT a live data source. Connect it to FranDB over MCP and it answers with real figures from 2,488 FDD filings.

Why ChatGPT alone falls short

Two limits get in the way.

First, the training cutoff. ChatGPT learned from data up to a fixed date, and FDDs change every year. Franchisors are required to refile annually. So even when ChatGPT "knows" a brand, its numbers are usually a year or more stale.

Second, the hallucination problem. When you ask for a specific figure it doesn't have, ChatGPT tends to produce one anyway. It would rather give you a plausible number than admit it doesn't know. For casual questions that's harmless. For a decision involving a few hundred thousand dollars, it's dangerous.

Pasting the PDF in helps a little, but not enough. FDDs are 200–400 pages of dense tables, and models routinely grab the wrong row or misread Item 7. You end up double-checking everything, which defeats the point.

The better setup: ChatGPT plus a real database

MCP (Model Context Protocol) is an open standard that lets ChatGPT connect to outside data sources, like a plug-in. Once you connect FranDB, ChatGPT can query our FDD database on your behalf. You ask in plain English; it calls the right tool and returns extracted, checked numbers.

What that unlocks:

Current data. The figures come from the latest filings we've processed, not ChatGPT's memory.
Real numbers, not guesses. The Item 19 earnings and Item 7 ranges were already pulled and validated.
Cross-brand comparison. ChatGPT can line up several franchises at once because it's reading structured data, not juggling PDFs.

How to connect ChatGPT to FranDB

The full step-by-step lives on our AI integration page, but the short version:

In ChatGPT, open Settings → Connectors.
Turn on Developer Mode under advanced settings.
Create a new connector, name it FranDB, and paste in the MCP server URL.
Sign in with your FranDB account and click Authorize.

That's it. FranDB's tools now show up in your chats. It takes about a minute and you don't write any code.

What to ask once it's connected

This is where it gets fun. Some questions that return real data:

"Which pizza franchises have the highest average revenue per location?"
"Compare total investment for Chick-fil-A, McDonald's, and Wingstop."
"Show me franchises under $100K with an SBA default rate below 5%."
"How many units did this brand close last year versus the year before?"
"Pull the franchisee contact list for this franchise in Texas."

Notice the pattern: every question asks for a number, a count, or a comparison. That's how to use it. ChatGPT picks the right FranDB tool, runs the query, and hands you the figures.

A realistic workflow

Say you're weighing three home-service brands. Instead of downloading three FDDs and reading 900 pages, you ask ChatGPT to compare their investment ranges, royalty rates, and Item 19 earnings side by side. It comes back with a table in seconds.

Then you go deeper on the one that looks best: closure trends from Item 20, litigation from Item 3, SBA loan performance. Each is one more question. You spend your time judging the brands, not transcribing tables.

When you've narrowed it to a real contender, you pull the franchisee contacts and call current owners. Then, before signing anything, you hand the FDD to a franchise attorney for the legal items (8, 12, and 17 especially). ChatGPT got you to the shortlist. A lawyer gets you across the line.

The honest limits

A connected ChatGPT is far more reliable than a standalone one, but it's still a research assistant, not an oracle. The data is only as current as the most recent filings we've processed. Item 19 doesn't exist for every brand. About a third of franchises don't disclose it. And no AI should be making your final legal call.

Used for what it's good at (fast screening and comparison across a lot of brands), it changes how franchise research feels. Less data entry, more deciding.

Prefer Claude or Cursor? The same database connects to both. Here's how franchise research works in Claude.