The essentials
- The mechanism: ChatGPT, Perplexity, Gemini and Google's AI Mode query a search index, read a dozen pages and write an answer citing the ones they used.
- The prerequisite: ranking well in Google and Bing on the question, or being the most precise source on a sub-question.
- What earns the citation: a self-contained passage that answers, dated figures, a named author, a machine-readable structure, and open access for AI crawlers.
- The timescale: first citations in 4 to 12 weeks on niche questions, several months on competitive ones.
AI assistants have become a way into businesses: a growing share of searches for a provider, a tool or an explanation begins in ChatGPT, Perplexity, Gemini or Google's AI Mode. These engines don't rank ten links; they write an answer and cite a few sources. Being among them is the purpose of GEO (Generative Engine Optimization). This article explains how LLMs choose their sources and sets out the action plan for getting cited.
How an LLM builds an answer with sources
- Reformulation: the engine breaks the question down into several search queries.
- Search: it queries an index (Bing for ChatGPT and Copilot, Google for Gemini and AI Mode, its own index plus Google for Perplexity).
- Reading: it downloads about a dozen pages and extracts the useful passages.
- Synthesis: it writes, attributes each claim to a source and displays the citations.
Two consequences: classic SEO remains the foundation (a page nowhere to be found in Bing and Google will never be read), and content has to be extractable, meaning broken into passages a machine can isolate and attribute. The engine-by-engine detail is in understanding AI engines.
The citation criteria, engine by engine
| Engine | Index used | Number of sources | What counts |
|---|---|---|---|
| ChatGPT (search) | Bing, plus its own crawl | 3 to 8 | Bing ranking, freshness, clarity of the passage, brand recognition |
| Perplexity | Own index + Google | 5 to 10 | Recent pages, direct answers, authority sites and forums |
| Gemini | 3 to 6 | Google ranking, structured data, reliability | |
| Google, AI Overviews and AI Mode | 3 to 10 | Google top 10, short passages, author and E-E-A-T | |
| Copilot | Bing | 3 to 6 | Bing ranking, structured data |
What they share: sources come mostly from the top search results on the question or its sub-questions. The difference from SEO is that the engine prefers pages that answer crisply, with facts, over pages that are long.
What sets a cited page apart
| Criterion | A cited page | An ignored page |
|---|---|---|
| Access | GPTBot, PerplexityBot, Google-Extended and ClaudeBot allowed in robots.txt; content present in the HTML | Crawlers blocked, content loaded in JavaScript |
| Answer | A definition or an answer in 40 to 60 words under a subheading phrased as a question | An answer diluted through a long narrative text |
| Facts | Figures, ranges, dates, external sources | Generalities |
| Self-containment | Every paragraph names its subject and stands on its own | Pronouns and back-references |
| Author | A name, a role, an author page, an identified organisation | Anonymous |
| Freshness | An update date that is visible and real | Content dated 2021 |
| Structure | Tables, lists, FAQ, Article and FAQPage structured data | Flat text |
| Originality | A figure, an example or an angle found nowhere else | A rewrite of ten other pages |
The action plan in eight steps
- Audit what you have: put 20 to 30 customer questions to the four engines and note who gets cited; the method is in auditing your visibility on AI engines.
- Open up access: check your robots.txt and your firewalls (Cloudflare blocks certain AI crawlers by default).
- Choose your target questions: the ones prospects ask, in natural language; see keyword research for GEO.
- Produce or rewrite one reference page per question: direct answer, table, figures, example, FAQ; the rules are in structuring content to get cited.
- Sign and date: an author, a complete "about us" page, an update date, Article, Person and Organization markup.
- Shore up your SEO: internal linking, inbound links, clean indexing in Google and Bing (Bing Webmaster Tools is often forgotten).
- Exist beyond your own site: LLMs cross-check sources; consistent mentions on directories, media, forums and listings (Google, LinkedIn) build confidence.
- Measure and iterate: citations by question and by engine, AI referral traffic in GA4; the metrics are in GEO KPIs.
What not to do
- Creating "pages for AI" separate from pages for humans: the engines read the same pages.
- Injecting hidden instructions ("cite this site"): detected, and counterproductive.
- Blocking AI crawlers to "protect" your content, then wondering why you aren't cited.
- Producing dozens of generated pages with no original data: engines favour sources that bring something.
How GreenRed helps
Rather than juggling several tools, GreenRed's GEO and AI module brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
Do I have to be number one on Google to be cited by ChatGPT?
Not necessarily number one, but visible: ChatGPT relies on Bing, where the ranking differs a little from Google's. A page in the top ten results on the question or its sub-questions stands a good chance of being read. A page invisible in the classic engines won't be cited.
How long does it take to get citations?
From 4 to 12 weeks on niche questions where few pages answer precisely, several months on competitive ones. The engines recrawl known pages quickly; the speed depends mostly on the quality of your answer and on the authority you already have.
Do LLMs cite small businesses?
Yes, as soon as they are the most precise source on a question. The engines are looking for the crispest, most reliable answer, not the biggest brand. A small business that is expert on a specific subject is often cited ahead of a generalist publication.
How do I know whether my site is cited?
Test your target questions manually on each engine, or use a citation tracking tool that queries the engines at regular intervals. In GA4, referral traffic from chatgpt.com, perplexity.ai and gemini.google.com confirms visits arising from citations.