How to get ChatGPT and Google AI to cite your brand: the 6 steps that actually work

There's no single trick for showing up in AI. But there is an order that works, and most people run it backwards. The short answer: start with schema, not with llms.txt. Here are the six steps in the order that actually moves the needle, with the number behind each one, so you don't burn time or money on the easy part someone's trying to sell you as the hard one.

What makes AI cite a page?

Six signals do most of the work: a clear entity (the model knows who you are), authority, structure, freshness, depth, and being readable by crawlers in the first place. The steps below are those signals in priority order.

Step 1. Add schema markup

Schema is the API between your site and the model. It labels your content so there's no guessing: this is a question, this is the answer, this is the company. What the evidence actually supports is narrower than the usual sales pitch, and more useful to you. A 2026 preprint that analysed 730 AI citations from ChatGPT and Gemini found no independent effect from the mere presence of schema once Google ranking was controlled. What separated the pages was the content inside the markup: pages whose Product or Review schema carried concrete fields, price, aggregateRating, specifications, were cited at 61.7% against 41.6% for generic types like Article, Organization or BreadcrumbList (Fischman, 2026). So the generic schema your CMS inserts on its own is not what earns the citation. The schema that publishes your real data is, because it hands the engine a fact it can extract and attribute to you. Start with FAQ schema, which puts your answers in the question-and-answer shape the model produces, and fill it with specifications and answered questions rather than slogans.

Step 2. Quote a named expert

Attribute your claims to a real person with a name and a title. No published study puts a number on the name itself, so treat it as a mechanism rather than a multiplier: a line from "our head of data, Jane Smith" can be checked, and an anonymous assertion cannot. What has been measured is the shape around the claim. In the Princeton GEO study, the three best performing tactics were citing sources, adding quotations and adding statistics, worth a relative improvement of 30 to 40% in visibility inside generative answers (Aggarwal et al., 2024). A named expert is how a quotable, attributable line gets onto the page in the first place.

Step 3. Publish your own data

Publish numbers only you can publish. That 30 to 40% lift came from adding statistics, quotations and citations to a page, and your own data is the one kind of statistic a model cannot generate for itself. Run a small survey, share your benchmarks, publish what you learned from your own clients, then mark it up so it reads as data rather than as prose. If a page contains a number that exists nowhere else, the model has a reason to point at you specifically.

Step 4. Keep it fresh

Keep your cornerstone pages current, and say so in a machine readable way. There is no credible public source for a fixed freshness multiplier, so discount anyone who quotes you one. The mechanism is real all the same: live-web engines like Perplexity and Gemini fetch at query time and read your dateModified to judge whether a page still answers a time-sensitive question. Real dates on real updates, not a monthly rewrite of everything.

Step 5. Build a consistent entity

Say the same thing about yourself everywhere: your site, your Google Business Profile, Bing, LinkedIn, industry directories. Every contradiction is a leak. Consistency isn't housekeeping here. It's what lets the model be confident enough to name you.

Step 6. And llms.txt?

It sits at the very bottom of the list. In its May 2026 AI guidance, Google said llms.txt isn't needed for AI Overviews, AI Mode, or any of its generative features. Add it if you like, but don't let anyone sell it to you as step one.

The ordering mistake most people make

The common error is starting with llms.txt or cranking out more blog posts on your own site, while the schema, the data, and the off-site mentions get ignored. Work top down. Schema, then expert quotes, then your own data, then freshness, then a consistent identity, and only then the file at the bottom of the list. If you want to know which of these is your weakest link, the AI Visibility Scan will point right at it.