A growing share of buying decisions now start with a question to an AI assistant instead of a search box. "Best auto detailer near me." "Who can build an app for my restaurant?" If ChatGPT, Claude, Gemini, and Perplexity do not know your business exists, you are invisible to those customers. This is sometimes called AEO or GEO (answer engine or generative engine optimization). The label matters less than the mechanics.
How AI assistants decide what to recommend
Two paths lead into an AI answer. First, training data: what was written about you across the web before the model was trained. Second, live retrieval: most assistants now search the web and read pages before answering. You can influence both, but retrieval is where you can win this quarter.
1. Be the page that answers the question
When an assistant searches, it reads a handful of pages and synthesizes. Pages that state facts plainly get quoted; pages of marketing fog get skipped. Write like you are answering a customer directly: what you do, where you do it, what it costs, how long it takes. FAQ sections are extraordinarily effective here because they mirror the exact shape of the user's question.
2. Structured data is how machines read you
Schema.org markup (LocalBusiness, Service, areaServed, FAQPage) turns your site from prose into facts a machine can lift with confidence. An assistant deciding between citing you or a competitor will favour the site it can parse cleanly.
3. Add an llms.txt file
A plain-text file at /llms.txt that summarizes who you are, what you offer, and how to reach you is an emerging convention for exactly this purpose: giving language models a clean, canonical summary of your site. It takes an hour to write. This site has one, and yes, AI crawlers do fetch it.
4. Do not block the crawlers that feed the assistants
Check your robots.txt. Many sites accidentally block GPTBot, ClaudeBot, PerplexityBot, and friends because a plugin added a blanket rule. Blocking them means opting out of being recommended.
5. Third-party mentions still decide trust
Assistants weight independent sources: review platforms, local news, directories, industry lists. A Google Business Profile full of detailed reviews, a mention in a local roundup, a listing on the right directories. These are citations a model can trust more than your own homepage.
6. Classic SEO still carries it
Retrieval-based assistants mostly re-rank ordinary search results. If you are on page five of Google, you are not in the AI's reading list either. Everything in my local SEO guide applies here unchanged: speed, per-service pages, structured data, reviews.
I build sites with all of this baked in from day one, and I have watched client businesses start appearing in AI answers within weeks of shipping the fundamentals. If you want to know whether assistants can see your business today, email me and I will run the check for you.