Updated October 2026
Your CEO asks ChatGPT for the best flower brands in New York on the way to a store visit, and your name isn't in the answer. By lunch there's a pitch in your inbox for a "GEO package," short for generative engine optimization, with an llms.txt file and AI-ready rewrites of every page, along with a one-line note from the CEO: "Should we?"
Before you answer, it helps to know that Google has already said what it thinks of most of that package, and that the part it doesn't cover is something your team may be better at than any agency.
As of October 2026, Google's guide to generative AI features in Search puts it plainly: optimizing for AI search "is optimizing for the search experience, and thus still SEO." AI Overviews and AI Mode pull pages from the same index and ranking systems as regular results, and the guide even lists things you can skip, which reads a lot like the pitch in your inbox. Google Search doesn't use llms.txt, so the file "will neither harm nor help," and nobody needs to chop your pages into chunks or rewrite them just for AI. If the heart of the package is that file, you'd be paying for something Google says it ignores.
That still leaves the CEO's real question. When an AI answer names another brand, it's often because that brand's pages say something specific and yours don't. Google calls this "non-commodity content" and warns against pages that "could easily be produced by a generative AI model." It gives you a fair test for your own site: if a model could write the paragraph without knowing anything about your company, it has no reason to quote you for it. Anyone could write "quality you can see." Only your team can say how long your flower dries, and who decided that, and why.
You probably hear those specifics every week without writing them down. At Budtender XP, our evenings for dispensary staff, they come out when budtenders start asking brand teams the questions customers put to them at the counter, and the brand team answers out loud. A model can't invent answers like that, but unless somebody writes them down, they stay in the room. A simple question log fixes that:
Then make sure the engines can actually reach those pages. Your developer lets the search crawlers from ChatGPT, Claude and Perplexity in through your site's robots.txt file, and their names are in the answers below. The age gate shouldn't send every URL to a single verification page; we wrote up how to set up an age gate crawlers can see past.
So the reply to your CEO is no to the package and yes to a different plan. Skip the llms.txt file, since Google says it changes nothing. Let the search crawlers in, keep the age gate out of their way, and start the log so your pages fill up with answers only your team can give. Then, once those pages are up, ask ChatGPT the CEO's question again, the best flower brands in New York, and if a competitor still comes first, open their page and read what they wrote that you didn't.
What is GEO for cannabis brands? GEO, or generative engine optimization, is getting your brand quoted in AI answers from Google, ChatGPT, Claude and Perplexity, and Google says it's still SEO.
Do cannabis brands need an llms.txt file? Not for Google, since its AI optimization guide says Google Search doesn't use llms.txt and the file neither helps nor hurts.
Which crawlers should a cannabis brand site allow for AI search? Googlebot, OAI-SearchBot, Claude-SearchBot and PerplexityBot, while training crawlers like GPTBot and ClaudeBot are separate settings you can block.
How do you know if GEO is working? Google's Search Console has a Generative AI performance report, and Google says no third-party tool has access to its ranking or AI systems.
Source: Google Search Central. For persons 21 and older. General marketing information, not legal advice.
