Do pages need to be shorter to be citable?

No. Bing’s Webmaster Guidelines name no page length but recommend one primary topic per page and key information near the top. A long page can be citable if each of its sections covers one aspect of that topic and makes sense on its own.

Should pronouns such as “it” or “they” be avoided entirely in citable content?

No. For citability, what matters is that each section names its subject explicitly, ideally in the first sentence, and that key statements stay unambiguous on their own. Pronouns remain normal and make the text easier to read.

Does writing for citability make content harder for people to read?

Usually not. The features of citable content—clear headings, explicitly stated facts, and unambiguous names—can also help people who skim a page or jump straight to a specific section.

Citability: when AI systems can cite content

Citability is a property of content, not of AI systems: it describes whether a single passage of a web page still works when it is lifted out of that page. This matters because AI systems such as ChatGPT, Gemini, Claude, and Perplexity retrieve web content for many answers and often show the pages they used as AI citations.

Why individual passages matter

AI systems that look up relevant content before answering (retrieval-augmented generation) often split documents into smaller passages that they retrieve individually. A passage can lose its context in the process. Anthropic, the maker of Claude, gives an example: on its own, the sentence “The company’s revenue grew by 3% over the previous quarter” reveals neither which company nor which period is meant.

In a study presented at the EMNLP 2024 conference, retrieval worked better when a language model first split the text into individual, self-contained statements instead of longer passages. In the Claude API’s web search tool, each citation also includes up to 150 characters of the cited text, so it points to a specific passage. Providers usually do not document in detail how their AI systems process web pages.

What makes a passage citable

In Bing’s Webmaster Guidelines, Microsoft states that services such as Copilot are more likely to select a page as a basis for AI answers (grounding) and for citations when its content stands on its own. Facts and definitions should be explicit rather than implied. From these and other points in the guidelines, a few rules of thumb can be derived:

The subject is named: A section names the company or product instead of referring back with “it” or “this.” Consistent names make companies and products clearly recognizable as entities.
Facts are explicit: Figures, prices, dates, and places are stated, not just implied.
One key statement, one clear sentence: The most important information in a section can be quoted as a single sentence. When that sentence opens the section, the section follows an answer-first structure.
One topic per section: A meaningful heading announces what the section is about. With semantic HTML, this structure is also recognizable to software.

Limits and distinctions

Citability improves the conditions for a citation but does not guarantee one. Microsoft, too, points out that GEO does not guarantee citations in AI answers. Whether an AI system retrieves web content at all depends on the system and the question. According to Anthropic’s API documentation, Claude does so mainly when a request depends on information that is current, changing, or outside its training data.

In the sense described here, citability is a term from GEO practice. In academic writing, by contrast, a citable source is one that is published, uniquely identifiable, and permanently accessible.

Citability

Definition

Why individual passages matter

What makes a passage citable

Limits and distinctions

Frequently asked questions