Getting cited by ChatGPT means something specific: when a person asks an assistant a question inside your field, your name, your page or your phrasing appears in the answer. That is a different job from ranking on Google.
The short version is that language models cite material they can parse, corroborate and quote. Parse means your work exists as text a machine can read and attribute to you. Corroborate means the same claim about you appears in more than one place the model already trusts. Quote means at least one passage is specific enough to be worth lifting whole. Most experts satisfy the first two conditions and stall on the third, because what they publish restates what the field already agrees on.
Three Different Things People Call a Citation
Before optimizing for anything, separate the outcomes, because they have different causes.
The link in the sources tray. The assistant answers and your URL sits in the citation list underneath. This is the closest relative of a search result and the easiest to earn with clean, well structured pages.
The paraphrase with no attribution. The assistant explains your idea in its own words and names nobody. This happens constantly and it is invisible, which is why counting citations alone underestimates how far an idea has traveled.
The recommendation by name. Someone asks who can solve a problem, and the model answers with a person. This is the outcome worth building for, and it rests less on page structure than on whether the model has enough corroborated material to risk naming a human being.
Condition One: Material a Machine Can Parse
Models read text. Everything that carries your expertise as an image, a slide deck, a PDF locked inside a member area or a video with no transcript is expertise the machine cannot hold. An expert with a career of conference talks and no transcripts has published almost nothing, as far as retrieval is concerned.
The practical fix is unglamorous. Put the substance on a page you own, in text, under a heading that states the question the section answers. Give every recurring idea one stable home instead of seven partial mentions scattered across old posts. Add transcripts under videos. Where your work is described elsewhere, make sure the description carries your full name and your field in the same sentence, because that pairing is what lets a model connect the two later.
Condition Two: The Same Claim in More Than One Place
A single page announcing that you are the leading authority on something is a claim with one witness. Models weigh corroboration, and the pattern Storytellers keeps seeing is consistent: material that appears across community platforms, professional networks, editorial coverage and your own domain gets picked up more readily than material confined to a personal site.
Corroboration for an expert usually looks like this. Your own domain states the claim in full. A professional profile states it in compressed form, using the same wording. Third parties describe your work from the outside, in their language. And somewhere in a searchable community there is a real answer you wrote, signed, addressing a real question. Four surfaces agreeing on who you are and what you do is a stronger signal than any single page can be, however well optimized that page is.
Condition Three: Something Worth Quoting
This is where most experts lose. A model quotes a passage when the passage says something the rest of the corpus has left unsaid, in a form compact enough to lift. Generic competence produces generic text, and generic text gets absorbed into the average instead of being cited.
Quotable material tends to share four traits. It names the object precisely, including the industry, the role and the constraint. It carries a sequence, so a reader can follow what happens first and what happens after. It takes a position that a reasonable colleague could dispute. And it belongs to someone, which means the claim is anchored to a person with a track record rather than floating free of any author.
Storytellers works that problem through Narrative Intelligence, the Story side of the Palacios Method and the side that decides what: what is worth telling, at which moment, in what order, to which audience and for what effect. Strategic Entertainment is the Telling side, the side that decides how it lands, and the 8 Steps are the motor under both. A machine retelling your idea is the same problem as a colleague retelling it in a meeting you are not in. The version that survives is the version built to be repeated.
The Bayer Agroscience Scene
Bayer Agroscience was inside a cost cutting cycle, with an instruction from the German headquarters that blocked hiring external suppliers. The team took the challenge to ChatGPT. The model answered that the path was an external consultant, and it named Fernando Palacios. The board made an exception to the rule, legal fitted the contract outside the blocked category, and the contract was signed in June 2025, for the script of the annual convention.
At the in person delivery Fernando asked how they had arrived at him, given that they were forbidden from hiring a supplier. The answer was direct: ChatGPT. We talked to ChatGPT, it said it had to be an external consultant, and that it had to be you.
That is the recommendation by name, and it is worth reading for what produced it. Fernando Palacios founded Storytellers in 2006, the first storytelling company in Brazil, and 20 years of documented work sits behind that answer: a method with a name, cases described in public, and a body of material spread across enough surfaces that a model was willing to say one person out loud.
For a year that scene read as a single stroke of luck, and it stopped being one. Since February 2026, Storytellers has averaged about one inbound recommendation a week generated by ChatGPT or Google, with no prospecting behind it. Three things belong next to that average, because a number without them is decoration. The base is inbound recommendations, counted by Fernando himself, by hand, with no analytics panel behind the count. The period is February 2026 to August 2026, roughly twenty six weeks. The meaning is that the average measures the return on the AIEO work done across 2025, arriving a year later. One boundary matters as much as the number: a recommendation means a lead arriving, and whether it becomes a contract is a separate question, counted separately.
The shift is the part worth taking home. In 2025 this happened once, by luck. From February 2026 it happens by design, and it happens most weeks, which is the moment the machine stopped being an accident and pulled up a chair. That makes the Bayer scene the opening case of a pattern that now repeats, and the pattern is the real target of this work: a machine that sends someone your way, by name, most weeks of the year.
Five Moves to Run First
1. Write the definitive page for one question. Pick the question you answer better than anyone you know, and publish the full answer on a URL you control. One question, one page, no hedging.
2. Transcribe one talk. Take the best presentation you have given, put the transcript on your own domain, and give it a heading that matches how a person would ask about the subject.
3. Align the wording across surfaces. Your site, your professional profile and your speaker bio should describe your specialty with the same phrase. Three different phrasings dilute the same signal three ways.
4. Answer three real questions in public. Find live discussions in your field and leave answers with substance in them, signed, on platforms that stay indexed.
5. Date and source every claim. A result with a year, a client and a method attached survives scrutiny. A result floating without any of the three gets discarded by careful readers and careful systems alike.
Publish One Answer Worth Repeating
Getting cited by AI rewards the expert who publishes like an author instead of a brochure. The vocabulary around this work is covered in GEO and AIEO for experts, and the wider recognition problem in how to build authority as an expert. For the full track, see for experts, or look at how the work runs in the services.
Working Together
Mentoring and consulting for individual experts
The AI is not the author. The AI is the pen. Fernando Palacios is the author. If you want the Palacios Method applied to your own body of work, the services page is where that starts.