Perplexity, Gemini, Claude: comparing citation patterns by engine

The four engines differ in how they show sources and which pages they favor, so a single optimization rarely gets you cited on all four at once. Perplexity openly lays out numbered source cards above its answer, while ChatGPT, in answers that use web search, slips links into the sentences of the body. Gemini inherits Google's search index and AI Overviews almost as they are, and Claude adds sources only to answers that use web search.

You first need to know which engines are missing your citations before you can set the order of work.

Citation patterns across the four engines: a comparison table

Enter a single question such as "best place for whitening treatments in Gangnam" into all four engines at once, and each engine differs in where it shows sources and how long its answer is.

The signals that earn points on all four engines are direct-answer paragraphs, clear subheadings, structured data, and consistent brand naming. Put these in place first and you lose nothing on any engine.

In client work, PION first enters the category's core question set directly into ChatGPT, Gemini, Google AI Overviews, and AI Mode to measure citation status by engine, then starts with the engine missing the most citations. Which engine to tackle first is decided by the measured numbers instead of a textbook priority order.

When does ChatGPT cite sources?

ChatGPT puts real-time results into the body as links only in answers that use web search. It searches on its own when a question needs current information, and users can also turn search on themselves. Answers built from training data without search show no sources.

Links are woven into sentences rather than set apart in separate cards, so users come across the supporting links as they read the answer. OpenAI Help Center's ChatGPT search documentation also explains that it uses both links within the body text and a separate source panel.

To be cited by ChatGPT, your content needs the following.

• The first sentence or two of the body is the answer: it should be in a form ChatGPT can slot straight into its answer. Content that stalls with a long introduction gets pushed down the list of citation candidates.
• Structured data: schema.org Organization, FAQPage, and Article markup needs to be reliably in place for the engine to trust and pick up the page.

A common mistake is treating "ChatGPT with search on" and "ChatGPT with search off" as the same. Citations appear only in answers that use search, so measurement is meaningful only with search turned on.

Perplexity shows sources first

Of the four engines, Perplexity shows its sources most openly. It lays out numbered cards at the top of the answer and uses footnotes to link each sentence of the body 1:1 to the numbered source it came from. Users see the source cards before they read the answer.

The outside analysis introduced in Semrush's article on optimizing for Perplexity finds that newly published or updated content gets a boost in Perplexity's rankings.

To get into Perplexity's cards, content has to be compressed. Write subheadings in the same natural language as the user's question, and make the very next one or two sentences the answer. One short, tightly cut direct-answer block has a better chance of making it into a card than one long explanatory paragraph.

Gemini inherits your SEO foundation

Gemini is effectively an extension of Google Search, so if your SEO is in good shape, citations follow. Answers appear as cards beside the body and in the AI Overviews source panel, and the signals for choosing sources overlap almost entirely with the authority, relevance, and freshness signals of Google's index. The E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and structured data that Google has long emphasized apply directly to Gemini citations as well.

Conductor's AEO/GEO benchmarks report analyzed 21.9M Google searches in the US and reports that AI Overviews appeared in 5.5M (25.11%) of them. In its guide to AI features, Google states that there are no additional requirements to appear in AI Overviews or AI Mode and that SEO best practices still apply. A page that isn't even indexed won't appear in Gemini either. So you can earn Gemini citations without a separate project, by widening the SEO you already do by one layer.

The order of work is to fix the SEO fundamentals first (indexing, internal links, page speed, E-E-A-T), then layer direct-answer paragraphs on top. Keep structured data consistent with the visible text on the page.

In which answers does Claude add sources?

Claude cites external sources only in answers that use web search. Answers built from training data without search show no sources.

In its web search announcement, Anthropic said it had expanded web search to all plans and that when Claude uses information from the web in an answer, it cites the sources directly.

PION treats Claude citations as work on getting the whole domain in order rather than on a single piece of content. The items to take care of are below.

1. Consistent brand naming: if basic information such as company name, service name, and founding year varies across your own site, press coverage, and social media, the engine cannot tie it together as a single entity. Start by aligning the naming.
2. Steadily publish original material worth citing: material others can use as evidence, such as your own measurement data or research, needs to accumulate on your domain.

Rather than pushing for Claude citations to show up in this month's results, it is better to handle them separately as cumulative asset work over 6–12 months.