How to Get Your Business Cited by ChatGPT and Google AI Overviews
Getting cited by an AI assistant is a retrieval problem before it's a content problem. Most answer engines fetch live pages at question time and ground their answer in what they find — which means a page published last week can be cited even though it post-dates the model's training.
That's good news. It also means the mechanics of being retrievable matter as much as the writing.
Step one: can the AI crawlers actually reach you?
Check your robots.txt for blocks on GPTBot, ClaudeBot, PerplexityBot and Google-Extended. Plenty of sites block these by accident, inherited from a template or a nervous developer, and then wonder why they never appear.
Decide deliberately. Blocking them is a legitimate choice — it just isn't compatible with wanting AI citations.
Step two: make passages liftable
Answer engines extract passages, not pages. A self-contained paragraph under a clear heading can be quoted safely; a paragraph that depends on the three above it cannot.
The test is simple: copy any paragraph out of your page, paste it somewhere with no context, and read it. If it still makes complete sense and still states who it's about, it's liftable. If it starts with "This means that…", it isn't.
Step three: publish the specifics
Models prefer sources that commit to facts. Service areas by name. Starting prices. Response times. Hours. Licence and insurance details. What you don't do.
A competitor who published "typically $400–$700 for a standard install in Nassau County" will be quoted ahead of your "competitive pricing, call for a quote" every single time.
Step four: fix what the internet already says about you
This is the half most people skip, and it's often the half that decides the outcome. Answer engines synthesise from many sources — directories, review sites, local press, social profiles, old listings you forgot existed.
If three directories carry a disconnected phone number and a former address, that is part of the model's picture of your business. Audit your NAP consistency across the major directories, claim and correct what's wrong, and give the accurate version enough corroboration to win.
Step five: structured data
Schema markup gives a machine unambiguous facts instead of inferences. For a local business the ones that earn their keep are LocalBusiness (with real geo and opening hours), Service for each offering, and FAQPage where you genuinely answer questions.
Don't mark up things that aren't on the page. It's a fast way to lose the rich result entirely.
What about llms.txt?
llms.txt is a proposed file at your site root that gives AI crawlers a curated map of your content. Adoption is voluntary and still uneven, so treat it as cheap insurance rather than a ranking factor.
The real return today is the clarity exercise it forces: deciding which twenty pages actually represent your business is useful regardless of who reads the file.
Set expectations properly
Structural fixes often show up within weeks, because retrieval is live. Reputation-driven changes — where a model has absorbed an outdated picture of you from third-party sources — take months, because you're waiting on other sites to update and be re-crawled.
Anyone promising you a fixed position in AI answers by a fixed date doesn't understand the mechanism.
