The five engines, one at a time.
Lesson 3 covered the general mechanics. This lesson is the practical tour — how each major AI engine actually finds and cites sources, and why the same page can win in one and vanish in another.
The engines do not cite the same sources.
An analysis of 680 million citations found only about 11% of domains are cited by both ChatGPT and Perplexity. Winning one engine tells you almost nothing about the others.
People talk about "AI search" as if it were one thing. It is not. It is at least six different systems, each with its own index, its own retrieval method, and its own taste in sources. A page that ChatGPT loves can be invisible to Gemini, because they are not even reading from the same library.
That single fact reshapes strategy. You do not optimize "for AI." You build broad authority and clean structure that travels across engines, and then you measure each engine separately because visibility does not transfer automatically. Here is each one.
1. ChatGPT (and why Copilot rides along)
ChatGPT is the most-used AI assistant in the world, and when it answers a question that needs current information, it runs a web search. That web search runs against Bing's index. This is the most under-appreciated fact in GEO: if your site is not indexed in Bing, ChatGPT's web search largely cannot find you.
It also means you get two engines for the price of one. Microsoft Copilot runs on the same Bing index. Get indexed and ranking in Bing, and you have addressed the retrieval layer for both ChatGPT search and Copilot at once. (Lesson 6 and the paid Quickstart cover Bing Webmaster Tools, which is the closest thing to first-party AI citation telemetry that exists.)
One nuance: ChatGPT does not always search. For some questions it answers from its trained knowledge with no retrieval and no citations at all. Citations appear when it actively browses for a specific sub-question — which is exactly why clean, retrievable, claim-level content matters.
2. Perplexity — the retrieval-first engine
Perplexity is the easiest engine to reason about because it is built around sources. It is retrieval-first: nearly every claim in a Perplexity answer carries a numbered citation, because the system retrieves documents first and then writes the answer from them. There is very little "answering from memory."
For a GEO practitioner, this is good news. Perplexity rewards exactly the things this course teaches — clearly structured content, direct answers near the top of a section, and proof-grade sourcing (statistics with attribution, named experts). If your content is well-organized and genuinely useful, Perplexity is often the first engine where you will see yourself appear.
3. Google AI Overviews — grounded in Google Search
AI Overviews are the AI-generated summaries that now sit at the top of a large share of Google results. They are grounded in Google Search — the system synthesizes an answer from pages that already perform in Google's index. So for this surface, traditional SEO still does a lot of the heavy lifting: pages that rank well are the candidate pool the Overview draws from.
The catch is reach. AI Overviews now appear on a large and growing share of queries — especially informational ones — and on those queries they absorb the click that used to go to the top organic result. Being the source inside the Overview is the new version of ranking first.
4. Gemini — Google's assistant, different evidence chain
Gemini is Google's conversational assistant. It can ground answers in Google Search, but it builds its evidence chain differently from AI Overviews — which is why Gemini and AI Overviews, despite both being Google products reading similar conclusions, frequently cite different URLs for the same question. Gemini also has tools that can pull content from specific URLs it is given.
The practical takeaway: do not assume "I'm in Google, so I'm in all Google AI surfaces." AI Overviews, Gemini, and Google's AI Mode can each land on different sources. Treat them as related but distinct targets.
5. Claude — the careful citer
Claude (the engine you are likely reading this course's brand voice through) is widely used in professional and enterprise settings. When it retrieves, it tends to be conservative and deliberate about what it cites, favoring clearly authoritative, well-structured sources. For most businesses Claude is a smaller citation channel than ChatGPT or Google surfaces, but it over-indexes among exactly the high-value professional audiences many B2B brands care about.
What this means for your work
Six engines, three different "homes" for retrieval:
| Engine | Retrieves from |
|---|---|
| ChatGPT (web search) | Bing's index |
| Microsoft Copilot | Bing's index |
| Perplexity | Its own retrieval (always cites) |
| Google AI Overviews | Google Search |
| Gemini | Google Search (different evidence chain) |
| Claude | Its own retrieval (conservative) |
Two homes dominate the work: get indexed and visible in Bing (covers ChatGPT search + Copilot) and in Google (covers AI Overviews + Gemini), and you have addressed the retrieval layer for four of the six. Perplexity and Claude then reward the same content-quality signals on top.
Because the citation pools barely overlap, the winning move is not to chase one engine — it is to build authority and structure that travel across all of them, then measure each engine so you know where you actually stand.
That "measure each engine" point is exactly what a Reffed audit does — it checks how all six engines see your brand and scores each. Lesson 5 gives you a way to spot-check this yourself with eight specific prompts, and Lesson 6 walks through reading the full audit.