Thought Leaders

What AI Assistants Are Saying About Your Dispensary, and What You Can Fix

Vee Popat
Add MyCannabis.com to your preferred sources on Google
Key PointsAI recommended the top-rated dispensary in town less than half the time when shoppers…
View more
  • AI recommended the top-rated dispensary in town less than half the time when shoppers asked where to go.
  • Strong star ratings failed to protect dispensaries because a handful of reviews on other platforms outweighed thousands of Google reviews.
  • Most traceable errors began with information the dispensary itself could fix on its own listings and profiles.

A shopper in Cincinnati asks an AI assistant about a local dispensary. The answer says the doors lock at 9:50 PM. The dispensary’s own website says it’s open from 9 to 10 every day.

So where did 9:50 come from? One customer review asked, “Since when does the door lock at 9:50?” ChatGPT and Google AI Mode both read that complaint as house policy and passed it along as fact.

Nobody did anything wrong there. A frustrated customer left a review, and two of the biggest AI tools turned it into a closing time.

That’s the kind of mistake a dispensary operator won’t see unless they look for it. So we went looking.

What We Tested

In November, my team asked five AI assistants (ChatGPT, Google AI Mode, Gemini, Perplexity, and Claude) about 40 dispensaries across 20 US cities. In each city, we picked the highest-rated dispensary on Google and a lower-rated competitor. Then we asked what a shopper would actually ask: the address and hours, if it’s any good, what people complain about, and if they should go there or somewhere else in the city.

We graded 800 answers against state license records, each dispensary’s own website, and its Google listing. The full method, the data, and the limits are in the study, When AI Gets Your Dispensary Wrong.

The Question That Matters Most: “Should I Go There?”

Out of 100 responses to the prompt “Should I go to [dispensary] or somewhere else in [city]?”, AI recommended the top-rated dispensary only 45 times. In the other 55, the assistants hedged, wouldn’t pick a shop, or in eight cases sent buyers to a competitor.

That’s the best-rated dispensary in town, and AI told shoppers to go there less than half the time.

The lower-rated dispensaries were recommended 13 times out of 100 and passed over 29 times.

The gap shows up when you ask a simpler question. When the question was “Is it a good dispensary?”, the assistants said yes 72 times in 100 answers about the top-rated dispensaries. They thought well of these shops. They were much less willing to tell anyone to actually go.

Strong Reviews Didn’t Protect Anyone

You’d expect a high rating to work like insurance. It didn’t.

Several of the harshest answers in the study were about dispensaries averaging 4.9 stars across thousands of Google reviews. In one case, 27 Yelp reviews carried more weight in an AI’s warnings than roughly 12,000 Google reviews for the same 4.9-star dispensary. In Cleveland, a dispensary with about 900 Google reviews at 4.8 stars had only six Leafly reviews. One said “avoid,” and several assistants passed that along.

Across the study, 82 of the 800 answers, roughly one in 10, repeated claims about the dispensary’s honesty, safety, or legality. Some were old news. A 2023 license suspension that was settled within weeks showed up with no date and no outcome.

The star ratings were off too. ChatGPT showed one Las Vegas dispensary at 2.9 stars. On Google, it sits at 4.9 from about 29,000 reviews.

The Basic Facts Still Slip

Location, hours, and rec or medical status should be the easy part. Across 200 answers to those questions, 13 had a fact that didn’t match an official record. That’s about one in 15.

Another 47 of the 800 graded answers mixed the dispensary up with a different business.

We also tested three dispensaries that had closed. ChatGPT and Claude flagged all three as closed. Gemini called two of the three open. The errors ran the other way too. In Albuquerque, two assistants suggested an open dispensary might be permanently closed, and the warning traced back to a Leafly profile nobody had claimed.

Where the Wrong Answers Come From

If you only read one section, make it this one.

In the study’s words, “most of the errors we could trace started with information a dispensary can fix.” That includes its own website, its Google listing, unclaimed Leafly, Weedmaps, or Yelp profiles, and outdated reviews on third-party directories. Similar business names, old news stories, and deal or price sites played a part too.

That doesn’t mean you control what AI says. You don’t, and anyone who promises otherwise is overselling. You do control a good share of what it reads, though.

What I’d Check This Week

If it were my dispensary, this is where I would start.

  1. Ask the questions yourself. Use a private window, try all five assistants, and ask more than once. In our “best dispensary” test, ChatGPT’s first pick changed between two runs in 7 of 20 cities, so one answer isn’t enough to go on
  2. Take over your listings. Claim your dispensary’s pages on Yelp, Leafly, Weedmaps, and Birdeye, then confirm the hours and open status on each. An unclaimed profile is how an open dispensary ended up described as possibly closed
  3. Give every location its own page. Put the address, phone number, hours, and rec or medical status on it as plain text, not inside an image or a pop-up
  4. Clean up old location pages. Delete or redirect them, so an assistant doesn’t find two sets of details and repeat the wrong one
  5. Keep reviews away from your hours and policies. If a review widget sits next to your hours, keep it clearly separate. It’s a small change, and it removes one more chance for a complaint to be read as a rule
  6. Make sure price answers stay in your state. Some price answers in the study pointed to dispensaries in the wrong state. If your city shares its name with a city in another state, test this one first
  7. Report fake listings near you. Perplexity showed spam-like “dispensary” listings in 17 of its 40 answers. On Google Maps, use “Suggest an edit” to report them

What This Study Can’t Tell You

We asked each question once per assistant over six days, November 5 to 10, 2026. AI answers change, and a shopper who’s signed in may see something different from what we saw logged out or on a free account. Forty dispensaries across 20 cities is a sample, not a ranking of any city. And we measured what the assistants said, not what shoppers did next.

None of these fixes guarantees a better answer, and if you check again in a month, some answers will have moved. But the information AI reads about your dispensary is the part of this you can actually change, so that’s where I’d spend the time.

Disclosure: Vee Popat runs ColaDigital, which ran the study cited in this article.

Vee Popat is the founder of ColaDigital and has spent more than two decades in search marketing, including the last several years helping dispensaries in the US and Canada with dispensary SEO. Connect with him on LinkedIn.