ColaDigital Study · Original Research · November 2026
We asked ChatGPT, Google AI Mode, Gemini, Perplexity, and Claude who has the cheapest weed and the best deals in 20 US cities. We then compared every price and deal they gave with the dispensary's own live menu from the same day.
This is our third study and the second on what AI tells cannabis shoppers. The first AI study looked at what AI gets wrong about dispensaries, and our cannabis SEO study looked at who wins Google in 39 cities. This one is about the number a shopper cares about most, the price. More of our work is on the cannabis marketing research page.
Vee runs ColaDigital, which helps cannabis retailers get found, and has worked in search marketing for over 20 years.
The Short Answer (November 2026): When we could check an AI's price or deal against the dispensary's own live menu, 64% matched, 22% were close and 14% were wrong. The bigger gap is what we couldn't check. 227 of the 560 prices and deals the AI gave us weren't on the menu we could read. When a price was wrong, the AI said the dispensary costs more than it does 66 times out of 69, and most wrong answers traced to an old page.
Study at a Glance
What we found
Shoppers now turn to AI for where to buy weed and what it costs. We wanted to know how close those answers come to the price on the dispensary's own menu. So we asked five AI assistants four price questions in each of 20 US cities, then checked what they said the same day.
The AI tools gave us 560 prices and deals. We could check 271 of them against a live menu or deals page. 174 matched, 60 were close and 37 were wrong. The other 289 didn't get a verdict. 227 weren't visible on the menu page we could read, and 62 weren't answers because the assistant declined or gave no pick.
The misses leaned one way. Where an assistant's price was off, it was above the live menu 66 times and below it 3 times. Shoppers were told a dispensary costs more than it does.
Most wrong answers traced to a page that was old or gone. That's good news for an owner, because an old page is something you can fix.
| AI assistant | Prices and deals | We could check | Matched | Close | Wrong | No answer |
|---|---|---|---|---|---|---|
| ChatGPT | 128 | 61 | 41 (67%) | 12 | 8 | 1 |
| Google AI Mode | 132 | 72 | 44 (61%) | 16 | 12 | 0 |
| Gemini | 105 | 59 | 36 (61%) | 18 | 5 | 1 |
| Perplexity | 90 | 30 | 23 (77%) | 4 | 3 | 44 |
| Claude | 105 | 49 | 30 (61%) | 10 | 9 | 16 |
Perplexity matched most often, but it declined to recommend or compare dispensaries 44 times out of 90. Claude declined 16 times, usually saying it couldn't verify a current price.
A shopper who asks an AI about your prices is reading an answer you didn't write. In our data that answer was more likely to be too high than too low, and it went wrong most often on the two questions that name your dispensary: your cheapest eighth and your first-time offer.
The cost is quiet. A shopper who hears you're expensive picks another dispensary and you never find out. A shopper who's quoted a first-time deal you stopped running shows up annoyed at the counter.
If an AI tool quotes an old page about your prices, that's the page it trusts. Fix that page, or put a current one on your own site, and the answers follow.
Only 25 of the 560 prices and deals said how old their source was, so a shopper had no way to tell a fresh price from a stale one.
Our first study, When AI Gets Your Dispensary Wrong, and How to Fix It, looked at what AI says about 40 dispensaries: where they are, when they're open and whether to go. Prices were the weakest part of those answers, and they're the part most likely to change by the week.
Price is also what shoppers weigh most. Price drives where people shop. In Sweed's 2025 shopper survey, 77% of 955 US adults said promotions decide where they buy. And 52% of consumers in SOCi's 2026 Local Discovery Index said they'd used AI to look for a local business during the previous 30 days.
An owner can fix a wrong address in a day. A wrong price, or a deal that ended last spring, costs a sale and nobody tells you. So this study asks one thing: when an AI tells a shopper what your dispensary charges, does your own menu say the same?
We asked four questions in every city. The first two are open questions that any shopper might type. The last two name a specific dispensary, which is how a shopper checks out a dispensary before visiting. In each city we named one dispensary: the top-rated one if its menu was readable, otherwise the lower-rated one.
| Question | Prices and deals | We could check | Matched | Close | Wrong |
|---|---|---|---|---|---|
| Cheapest weed in the city | 141 | 39 | 28 | 11 | 0 (0%) |
| Best deals in the city | 152 | 47 | 29 | 14 | 4 (9%) |
| Cheapest eighth at a named dispensary | 134 | 88 | 53 | 22 | 13 (15%) |
| First-time deal at a named dispensary | 133 | 97 | 64 | 13 | 20 (21%) |
This question gave us the fewest prices and deals we could check, 39 out of 141. Most answers came back as a list of dispensaries or a promise to check a menu, and what was inside couldn't be matched to a price on a live page. The 39 we could check had no wrong answers, but 11 were close.
Deals were easier to check than a ranking, and 4 of 47 were wrong. The 4 wrong answers named deals the dispensary's own page didn't show, or a dispensary in the wrong state.
Here the AI tools had a specific dispensary and a specific product, and they still missed. 13 of 88 prices and deals we could check were wrong. Every miss we could give a direction was above the live menu: the assistant named an eighth that cost more than one already on the dispensary's own page.
This was the hardest question for the assistants, with 20 wrong answers out of 97. A first-time offer is easy for a dispensary to change or drop, and old versions can live on in roundups and listings long after.
| Type | Prices and deals | Matched | Close | Wrong | Not checked | No answer |
|---|---|---|---|---|---|---|
| Deal | 279 | 105 | 34 | 24 | 102 | 14 |
| Price point | 173 | 52 | 23 | 12 | 77 | 9 |
| Cheapest or best | 83 | 2 | 1 | 0 | 41 | 39 |
| Fact | 25 | 15 | 2 | 1 | 7 | 0 |
We kept the same 20 cities as our first study, so the two line up. Every city got two dispensaries, 40 in all, and we put the same four questions to all five AI tools. Here's where they are:
If your dispensary is in one of these cities, the findings above are the closest match to what your shoppers see. If it isn't, the fixes still apply, because the same old-page problem shows up everywhere we looked.
For one dispensary, several answers offered 30% off a first purchase, or 99% off a first eighth with a $30 minimum. Its own FAQ said 10% off for first-time customers. Some answers marked the bigger numbers as coming from a third-party listing, and some noted that the sources disagreed. Others stated them as the offer.
For a second dispensary, four answers quoted 20% off a first purchase, one of them from a customer report dated October 2025. Its specials and location pages showed no first-time deal at all.
Three answers named $29.75 as a dispensary's cheapest eighth. Its own menu showed one at $22.50 on sale. In another city, an answer built its cheapest eighth from a producer's listing, not the dispensary's menu, and quoted $27. The dispensary's own menu showed one at $10.
The gap runs from a few dollars to more than half the price. A shopper who reads $27 for an eighth and never checks the menu is a sale you lost without knowing it.
Asked for the best deals in one Illinois city, an assistant listed a dispensary that belongs to a Massachusetts group and attached prices to it. Separately, Perplexity declined to recommend dispensaries but still showed place cards for dispensaries outside the shopper's city, including one in another state.
For one dispensary, an AI tool quoted a one-cent first-time gram, a half-ounce buy-one-get-one deal with a $1 eighth and a 20% weekend deal. Its online menu showed none of them, and its offers page returned a 404 error.
An AI tool told shoppers a dispensary gives $25 off $100 on a second visit with a rewards account. Its listing showed $10 off a second visit with a $60 minimum. The $25 off $100 offer is a real one, but it's that dispensary's medical first-order offer.
One assistant said it couldn't find a deal listing for a dispensary in a large California city and wasn't sure the dispensary had a location there. It does, and we read its page the day before. The same assistant had named it the day before.
Price answers often came with a tax or legal note. We found 24 such statements in 22 answers and checked each on the state regulator's own page on November 12. We checked 20. The other 4 were rough totals, like “about 30% in taxes,” that depend on local rates we didn't verify.
14 of the 20 were correct. 4 were incomplete and 2 were out of date. None were wrong.
Google AI Mode made 13 of the 24 statements, Claude 7, Gemini 3 and ChatGPT 1. Perplexity offered to explain taxes but gave no figures.
Everything below is something you control. We've put the fixes in the order of how many wrong answers they would have prevented in our data.
Old offers did the most damage, and first-time offers had the most errors of any question. An assistant will quote a deal from a roundup or a customer report long after the dispensary stopped running it.
An assistant can only quote the price it can read. When a menu sits inside an app, behind a gate or on page 7 of a list, the assistant falls back on whatever third-party page it finds. Our guide to making your dispensary website easy for AI to read shows how to set up a menu so it can be quoted. Our dispensary menu SEO guide covers how to build the menu page itself.
We found 27 answers that pointed to a dispensary in the wrong city, state or location. AI tools mix up chains, near-duplicate names and dispensaries with the same name in different states.
Two Google AI Mode answers told Baltimore shoppers Maryland's sales tax is 9%. The rate rose to 12% on July 1, 2025. If a state changes its rate, assistants keep the old one until a current page says otherwise.
Curious what AI is quoting for your prices today? I'll run these checks on your menu and tell you what I find.
You can run our test on your own dispensary in about 20 minutes. Use a private browser window and sign out where you can. Begin a new chat for each question, and use a temporary or incognito chat if the assistant offers one. Google AI Mode may still tailor results to you, even when you're signed out.
Put each of these to all five assistants, in a new chat every time:
We asked these four exactly as written above, so you can compare your answers with ours.
Then open your own live menu and deals page and compare. For every price or deal an assistant gives, write down whether it matches, how far off it is and which source it cited. That source is where your fix starts.
Rather not spend the time? We'll run the check for you.
Have Us Check Your Dispensary
We'll send you:
This study comes out of the daily work of getting dispensaries found. Here's what that work has produced:
The data file has every grade, with dispensaries shown as codes and not names.
Download the data we used: every price and deal we checked, the tax and legal statements, the rerun results and the README. The charts are here too: by question and by AI assistant. Dispensaries appear as codes, not names. If you use the data or charts, please link back to this page and name ColaDigital as the source.
Not reliably. Where we could check an answer against the dispensary's own live menu, 64% matched and 14% were wrong. But 227 of the 560 prices and deals couldn't be checked from the menu page we could read, so how accurate those were is still unknown.
In our data, most wrong answers came from an old or dead page. 23 of the 37 wrong answers traced to one, and the live menu was usually cheaper. When a price missed, the assistant was above the live price 66 times out of 69.
We don't rank them. Each AI tool gave us between 30 and 72 answers we could check, and the gaps are small. Perplexity matched most often, 23 of 30, but gave no answer to 44 of its 90 prices and deals. Claude declined 16 times, usually because it couldn't verify a current price.
Treat it as a lead, then check the dispensary's menu. Only 2 of the 83 answers that named a cheapest or best option matched a live menu.
Remove old offers and date every promotion. Old pages caused 23 of the 37 wrong answers, and first-time offers had more errors than any other question.
Vee Popat, ColaDigital. “What AI Tells Shoppers About Cannabis Prices, and How to Fix It.” Published November 2026 at https://coladigital.ca/ai-cannabis-price-answers-study/. If you quote our charts or numbers, please cite ColaDigital and add a link back here.
We checked the tax and legal sources below on November 12, 2026. The two surveys were last checked on November 9, 2026, and any date inside an entry is the source's own.
SOCi 2026 Local Discovery Index: soci.ai/insights/local-discovery-index
Sweed, The 2025 Cannabis Shopper Experience Report: online survey of 955 US adults 21+, October 6 to 10, 2025, conducted by Wired Research (info.sweedpos.com)
Maryland Comptroller, sales and use tax notice on the cannabis rate change, May 22, 2025
Illinois Department of Revenue, cannabis tax FAQ
Illinois Cannabis Regulation Oversight Office, FAQ on transporting cannabis across state lines
Ohio Revised Code 3796.40, version in force from March 20, 2026
Nevada Department of Taxation, cannabis tax page and Tax Fundamentals for Cannabis Establishments (January 2025)
Massachusetts Department of Revenue, tax rates for marijuana (mass.gov)
California Department of Tax and Fee Administration, cannabis retailers page
Missouri Department of Revenue, marijuana tax page; KBIA, April 4, 2023, on Columbia and Boone County 3% taxes
New Mexico Taxation and Revenue Department, 2025 tax changes
Arizona Department of Revenue, adult-use marijuana page
Prices are one piece of how a shopper picks a dispensary. These two studies cover the rest.
Browse all of it on our research page.
Want to know what AI is telling shoppers about your menu, prices and deals? Book a 30-minute call with Vee and he'll go through the answers with you, then list what to fix first. We handle fixes like these inside our dispensary SEO work. Our dispensary SEO case study shows what that did for one dispensary.