There is a lot being sold into cannabis right now under the labels AEO and GEO. Google has read the same marketing you have, and it has written a page about it.
The five things Google says you can ignore
This is Google's guide to optimizing for generative AI features, quoted rather than paraphrased, because the paraphrases circulating are the problem.
While terms like Answer Engine Optimization (AEO) or Generative Engine Optimization (GEO) are common online, many suggested "hacks" aren't effective or supported by how Google Search actually works.
Google Search Central, Optimizing for generative AI features, read September 8, 2026
1. Special files, including llms.txt. "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them." Google adds that keeping one "will neither harm nor help your site's visibility or rankings in Google Search, as Google Search ignores them."
We publish an llms.txt and an llms-full.txt on this site. They do nothing for Google Search, and it would be dishonest to run this page without saying so. The reason they stay is the carve-out Google itself names: other services and systems that do read those files.
2. Chunking. "There's no requirement to break your content into tiny pieces for AI to better understand it. Google systems are able to understand the nuance of multiple topics on a page and show the relevant piece to users." And plainly: "There's no ideal page length."
3. Rewriting for machines. "You don't need to write in a specific way just for generative AI search. AI systems can understand synonyms and general meanings." Google adds that you do not have to worry about capturing every long-tail variation, which is the anxiety keyword tools are sold on and which the keyword research spoke covers in more detail.
4. Chasing mentions. "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem. Our core ranking systems focus on high-quality content while other systems block spam; our generative AI features depend on both." In a category where link and mention packages are sold hard, that sentence is worth keeping to hand.
5. Structured data. "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." Keep it for rich-result eligibility, which is what the schema spoke documents it doing.
Read the list together and most of what is marketed as AI optimization in this category is on it.
The one lever Google does name
The same guidance is not purely negative. It names a positive route for local businesses specifically:
Where appropriate, generative AI responses can include product listings, product information, and information about local businesses. Using products like Merchant Center and Google Business Profiles can help your products and services to be visible in both AI responses and other Google Search results.
Google Search Central, Optimizing for generative AI features
Half of that transfers to a dispensary. Merchant Center does not, because profile and Shopping product listings are governed by a policy that prohibits recreational drugs, which the Business Profile guide covers. What is left is the Business Profile, and it is the same asset the Map Pack draws on.
That collapses a lot of the mystique. For a dispensary, the documented AI-visibility work and the documented local-search work are largely the same work: a complete, accurate, well-reviewed profile, and content on your own site that answers real questions.
The layer nobody checks: can the crawlers read you at all
Before any of that matters, a more basic question. Assistants reach the web through named crawlers, and a line in a robots.txt file decides which of them may read your pages. Almost nobody in cannabis has looked at their own.
The distinction that matters, and it is the one people get wrong, is that search crawlers and training crawlers are different bots with different consequences. From OpenAI's own crawler documentation:
OAI-SearchBot is for search. OAI-SearchBot is used to surface websites in search results in ChatGPT's search features. Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links.
OpenAI, Overview of OpenAI crawlers, read September 8, 2026
OpenAI states that the settings are independent: a site can allow OAI-SearchBot to appear in search results while disallowing GPTBot so its content is not used for training. Google draws the same line for its own token. Google's crawler documentation says Google-Extended governs whether crawled content may be used for training and grounding Gemini, and that it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search."
So blocking a training crawler costs training data. Blocking a search crawler costs visibility. A blanket block of everything with "bot" in the name, which is what a nervous operator or an aggressive security setting produces, does both.
What eight cannabis-adjacent domains actually publish
We fetched and parsed the robots.txt of eight domains on 8 September 2026. This is a snapshot of public files that can change any day, which is why the date matters more than the table.
- Leafly disallows GPTBot across the site, with a short list of allow exceptions for individual articles, and disallows meta-externalagent. OAI-SearchBot is not restricted. Read against OpenAI's own documentation that is a deliberate and fairly sophisticated posture: opt out of training, stay visible in ChatGPT search.
- Weedmaps carries no AI-specific groups, and its wildcard group does not block them.
- Dutchie returned HTTP 403 for robots.txt itself to an ordinary request, so its crawler policy cannot be read from outside at all.
- The four cannabis SEO agencies we publish comparisons against: one names the AI crawlers in its file. Three have no AI-specific groups whatsoever.
- This site allows every one of them explicitly, in named groups, with a note in the file warning maintainers that Cloudflare's managed robots.txt feature would silently override that policy if it were switched on. That warning is there because it is a real failure mode, and the citations spoke is a live example of a robots policy being overridden by infrastructure.
The practical instruction is one command. Fetch your own robots.txt, look for the AI user agents by name, and decide each one deliberately rather than inheriting whatever your host or CDN chose for you.
How to measure it
Two routes, and only one is first-party.
Inside Google, the guidance points at a report: "To measure how your content is performing in generative AI features on Google Search and Discover, use the Generative AI performance report in Search Console." That is free, it is yours, and we have not pulled it for this site yet, which is now on our monthly export routine.
Outside Google, there is no report, and the same guidance is blunt about the tools being sold to fill that gap: "Be wary of third-party tools that promise ranking success or claim to use 'internal' Google metrics. No third-party tool has access to our internal ranking or AI systems."
Which leaves the unglamorous method: ask the assistants the same questions on a schedule and write down what they say, including the times they say nothing about you.
A twelve-query protocol you can run on your own store
Since no first-party report covers the assistants Google does not operate, the only honest method is to ask them the same questions on a schedule and write down what comes back. Here is the shape we use.
Twelve fixed queries, run monthly across five surfaces: ChatGPT with search, Perplexity, Google AI Mode or the AI Overview, Gemini, and Claude with web access. For each run, log the date, the assistant, the query, whether your store was named, the URL cited if there was one, and which competitors appeared.
- Entity, four queries: who is your store name; is your store open now; does your store deliver; your store hours and address.
- Commercial, four queries: best dispensary in your city; dispensary near a landmark you are close to; where to buy a category you are known for in your city; dispensary open late in your city.
- Informational, four queries: four questions a customer asks before choosing a store in your market, ideally ones you have answered on your own site.
The last four matter most. Entity queries mostly test whether your Business Profile is complete and consistent, which is the lever Google names. Informational queries test whether anything you published is being used as a source, which is the only part of this that content work can actually move.
Two rules that keep the exercise honest. Run it in a clean session with no prior context, because an assistant that has just been told about your store will happily repeat it back. And log the runs where nothing of yours appears, in the same file, with the same care. A visibility log that only records the hits is a marketing asset rather than a measurement, and it will not tell you when something breaks.
One thing that is genuinely new
Worth a short mention because it is in the guidance and it is not yet marketing. Google describes agents that "may access your website to gather the data they need to complete these tasks, such as analyzing visual renderings (like screenshots), inspecting the DOM structure, and interpreting the accessibility tree", and points to best practices for making a site agent-friendly.
For a dispensary the useful reading is that a machine trying to answer "what is in stock near me" has to get through the same obstacles a person does. An age gate that blocks crawling, a menu that only exists inside a third-party frame, and a store locator that needs a click to reveal an address are all things the website spoke and the multi-location spoke cover, and each of them is a wall for an agent as well as for a customer.
The working checklist
- Read your own robots.txt and find every AI user agent by name.
- Decide training and search separately. OAI-SearchBot and Google-Extended are different decisions from GPTBot.
- Check whether your host or CDN is overriding your policy, by fetching the live file rather than reading the one in your repository.
- Check your menu provider and directories, since your listings live on their domains and their robots policy is not yours.
- Finish the Business Profile, because it is the only positive lever Google names.
- Open the generative AI performance report in Search Console and record a baseline.
- Do not buy llms.txt, chunking, AI rewriting or mention packages as AI tactics. Google lists all four among the things you can ignore.
- Keep structured data for rich-result eligibility, not as an AI play.
- Run a fixed query set monthly and log the misses as carefully as the hits.
- Ask any vendor to cite Google's guidance, which is the test Google itself hands buyers.
Frequently asked questions
Does a dispensary need an llms.txt file to appear in AI answers?
Not for Google. Its guidance says you do not need to create new machine-readable files, AI text files, markup or Markdown to appear in Google Search including its generative AI capabilities, because Google Search itself does not use them, and that maintaining such files will neither harm nor help your visibility or rankings there. We publish one on this site and it does nothing for Google Search. The carve-out Google names is other services and systems that do read those files, which is the only honest reason to keep one.
What does Google say actually helps AI visibility?
One thing, by name. Its guidance says that generative AI responses can include information about local businesses, and that using products like Merchant Center and Google Business Profile can help your products and services be visible in both AI responses and other Google Search results. For a dispensary only half of that transfers, since Shopping policy prohibits recreational drugs, which leaves the Business Profile. That is the same asset the Map Pack draws on, so the AI work and the local work are largely the same work.
Is schema markup required for AI search?
No. Google lists overfocusing on structured data among the things you can ignore: structured data is not required for generative AI search and there is no special schema.org markup to add, though Google recommends keeping it because it helps with rich-result eligibility on Google Search. Anyone selling you markup as an AI tactic is selling against Google's published position.
Should I break my pages into small chunks for AI?
Google says there is no such requirement. Its wording is that there is no requirement to break content into tiny pieces for AI to understand it, that its systems can understand multiple topics on a page and show the relevant piece, and that there is no ideal page length. It adds that you should make pages for your audience rather than for generative AI search.
Which AI crawlers should a dispensary allow?
Distinguish the search crawlers from the training crawlers before you block anything. OpenAI documents that OAI-SearchBot is for search and that sites opted out of it will not be shown in ChatGPT search answers, while GPTBot governs training, and that the two settings are independent. Google documents that Google-Extended controls whether content is used for training and grounding Gemini and that it does not affect inclusion or ranking in Google Search. So blocking a training crawler costs training data, and blocking a search crawler costs visibility. Those are different decisions.
Do the cannabis marketplaces let AI crawlers read their listings?
It varies, and it is worth checking rather than assuming, because a dispensary's listings often live there. Reading the published robots.txt files on 8 September 2026: Leafly disallows GPTBot across the site with a short list of exceptions for individual articles, and disallows meta-externalagent, while leaving OAI-SearchBot unrestricted, which by OpenAI's documentation is an opt-out of training rather than of ChatGPT search visibility. Weedmaps carries no AI-specific groups. Dutchie's robots.txt returned a 403 to an ordinary request, so its policy cannot be read from outside. Those files can change any day; the date is the point.
How do I measure whether my dispensary appears in AI answers?
Two ways, and only one of them is first-party. Google documents a generative AI performance report in Search Console for measuring how content performs in generative AI features on Google Search and Discover. For the assistants Google does not operate, the only honest method is to ask them the same questions on a schedule and write down what happens, including the runs where you are not mentioned. Google separately warns to be wary of third-party tools that promise ranking success or claim to use internal Google metrics, and states that no third-party tool has access to its internal ranking or AI systems.
Is AEO or GEO a real service?
The terms are real and Google addresses them directly. It writes that while terms like Answer Engine Optimization or Generative Engine Optimization are common online, many suggested hacks are not effective or supported by how Google Search actually works, and it then lists five specific things you can ignore. Google also tells buyers to ask whether an agency's AI advice aligns with that guidance and whether it cites official Google documentation. That is a test with a right answer, and a large share of what is sold as AEO or GEO fails it.
