State of Dispensary Search 2026: Client Data and a Menu Crawl Sample

In Bud Authority's State of Dispensary Search 2026, AI engines named the client in 1,843 of 3,275 runs (56.3%) across 30 clients, and menu pages drew 12.1% of 751,982 Google clicks across 31 clients from 2026-07-01 to 2026-09-30. In a separate crawl sample, 9 Dutchie storefronts fetched as GPTBot returned a median 2 readable words. The client figures are directional rates from one agency's book, not a measure of all dispensaries.

By Mark Wallace, Co-Founder & Chief Growth Architect, Bud Authority. Figures compiled read-only on 2026-10-13. No client is named.

• The sample is one agency's active clients. We hold no outside benchmark, so nothing here is compared with dispensaries in general, and nothing predicts the outcome any one store should expect.
• The 11 clients with a 2026-08-05 pass have 42 to 44 runs each, against 54 to 320 for the other 19, so their per-client rates are coarse.
• We hold a citation reading for 7 of Gemini's 824 runs (0.8%), so Gemini has no usable cite rate and the all-engine cite rate is in practice a ChatGPT, Claude and Perplexity figure.
• Mention and cite rates are presence rates across repeated runs. They are directional, and they do not test why any client's rate is high or low.
• The crawl sample was measured apart from the client data, so it cannot show whether menu rendering explains the menu click share. Our crawl requests sent the GPTBot user-agent string from a non-OpenAI IP address, so they may not match what OpenAI's crawler receives.
• Every figure comes from a single window. No period-over-period comparison is made.

Both ends of the range rest on thin samples. The client named in no run and one of the two named in every run had 2026-08-05 passes, with 42 and 44 runs.

Table C pools every Google click the 31 clients received from 2026-07-01 to 2026-09-30.

Home was the largest click class for 27 of 31 clients, location for 2 and menu for 2.

Location pages drew 16.3% of all clicks in Table C, yet the median client took 0.1% of its clicks from them. Only 24 of 31 clients have location pages tracked, and one client took 82.2% of its clicks from that class.

Menu-class pages make up 70,737 of 85,229 tracked pages (83.0%), and 30 of 31 clients drew at least one menu click. The 103 tracked home pages drew 485,144 clicks, against 90,656 for the menu pages.

That works out to about 1.3 clicks per tracked menu page across the quarter, the lowest of the six page classes. Menu pages took 20.9% of all impressions at a 1.8% click-through rate, against 7.1% for home pages.

Two clients break the pattern. One took 86.1% of its clicks from menu pages and another took 77.2%.

The dataset is published as a CSV at /data/state-of-dispensary-search-2026.csv. It holds the tables on this page, one row per table row, with a table column naming the source table. The counting rules are set out in the Method section.

Variables measured: AI runs, mentioned runs, verified citations and measured runs per engine, the spread of client mention rates, plus Google clicks, impressions, click-through rate, tracked pages and the median client share per page class. Measurement technique: read-only aggregation of Search Console data and each client's latest stored AI visibility pass, plus a separate GPTBot fetch of Dutchie storefronts.

Cite this as: Mark Wallace, Bud Authority, State of Dispensary Search 2026, https://budauthority.com/research/state-of-dispensary-search-2026. When quoting a figure, name the table letter (A to D) so the source row can be found.

Journalists who want the method detail behind any table can email service@budauthority.com. Named results for individual clients are in the case files on our portfolio page, and this dataset names none.

Pooled shares add up every client's clicks, so the biggest accounts set the result. The 3 largest clients held 310,183 of the 751,982 clicks (41.2%). The median takes each client's own split and counts every client once, which is why the home share reads 64.5% pooled and 77.5% at the median.

Absence on ChatGPT is common in this dataset: the median client was named in 43.0% of its ChatGPT runs, against 61.8% to 66.1% on the other engines. Our data does not test causes. One you can rule out: OpenAI's crawler documentation says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers".