When AI Gets Your Dispensary Wrong, and How to Fix It
A ColaDigital study, November 2026, by Vee Popat
Study page: https://coladigital.ca/ai-dispensary-answers-study/

WHAT'S IN THIS DOWNLOAD
- 01-key-numbers.csv: the headline numbers from the study
- 02-how-the-ai-assistants-compared.csv: the scorecard from the top of the report
- 03-should-i-go-answers.csv: "Should I go to [dispensary] or somewhere else?" answers, top-rated vs lower-rated dispensaries
- 04-where-is-it-answers.csv: "Where is it, what are its hours, and is it rec or medical?" answers, by AI assistant
- 05-complaint-answers.csv: complaint answers where the bad outweighed the good, by AI assistant
- 06-best-dispensary-consistency.csv: how often the "best dispensary" pick changed between two runs, by AI assistant
- 07-closed-dispensaries.csv: what each AI assistant said about three dispensaries that had closed for good
- Four chart images (PNG) used in the report

HOW WE COLLECTED THE ANSWERS
We asked ChatGPT, Google AI Mode, Gemini, Perplexity and Claude the same questions about 40 dispensaries in 20 US cities: the top-rated and a lower-rated dispensary on Google Maps in each city. Every question started a fresh chat with no follow-ups. Answers were collected November 5 to 10, 2026, through a US internet connection. ChatGPT, Google AI Mode and Gemini were used logged out; Perplexity and Claude on free accounts with no chat history.

We graded 800 answers (four questions about each dispensary, on all five assistants). We also collected 430 city-wide and closed-dispensary answers that sit outside the graded 800. A fact only counts as wrong when an official record contradicts it: the state license list, the dispensary's own website, or its Google listing. A second grader re-graded a random 80 answers without seeing our grades; we agreed on 61 of 80 (Cohen's kappa 0.61). Full method and limits are on the study page.

These files contain totals only. They don't name the dispensaries in negative findings.

USING THIS DATA
You're welcome to use these numbers and charts in your own reporting, as long as you credit ColaDigital and link to the study page above.
Suggested credit: ColaDigital, "When AI Gets Your Dispensary Wrong, and How to Fix It," November 2026, by Vee Popat. https://coladigital.ca/ai-dispensary-answers-study/
