Updated on 18 November 2026, by João Leite, Founder of murmur.

A glossary on your site helps you appear in AI when each term has its own page and the definition is in the first sentence. The question it answers is the definition question, the “what is” that a buyer asks when they come across a word from the sector and do not know what it means. Each term page carries the definition, an example, the neighboring term and the link to the guide that covers the subject in full.

murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, counted glossaries among the sources in its own weekly measurement. That is 90 questions about GEO, AI search and GEO agencies, asked of ChatGPT, Claude, Perplexity and Google AIO in 10 rounds, from 4 September to 2 November 2026, with 2,437 answers with a source. Glossary pages were a source in 55 of those answers, in 26 of the 90 questions, almost all of them definition questions. Of the 44 glossary URLs cited, 40 were the page for a single term.

  • The glossary on your site pays off on the “what is” question. In murmur’s measurement, the 26 questions in which a glossary was a source were almost all definition questions.
  • The glossaries appeared in all four AI models measured. There were 55 answers with a glossary as a source, out of 2,437 answers with a source, from September to November 2026.
  • One page per term. Of the 44 glossary URLs cited in that measurement, 40 covered a single term.
  • The term page opens with the definition and carries the example, the neighboring term and the link to the full guide.
  • The pages most used in answers carried more definition markers. In the study by Zhang, He and Yao, with 18,151 cited pages, the difference was 57.33% over the least used pages.

The glossary on your site helps when each term answers the “what is” question

The glossary on your site helps when each term answers, right in the first sentence, the “what is” question the buyer asks the AI. Someone who asks what a word from the sector means wants a sentence that defines it, and the page that starts with the definition delivers that sentence without a detour. The page that opens with the history of the term, with an invitation to talk to the team or with a list of links forces the person to look for the answer in the middle of the text.

The definition question usually comes before the purchase question. The person comes across the term in a proposal, in a contract or in a conversation, asks what it means and only then asks who sells it. The company that publishes the definition on its own site has a page ready for that first step, with its name in the URL.

A glossary that helps has three markers: one term per page, the definition as the first sentence and the link to the page that goes beyond the definition. Without those three, it becomes a long list of words, good to read from top to bottom and hard to use as the answer to a single question.

How often the glossary on your site appears as a source, in murmur’s measurement, AI model by AI model

In murmur’s measurement, glossary pages were a source in 55 answers, across the four AI models measured, between September and November 2026. The questions are the same 90, about GEO, AI search and GEO agencies, repeated in 10 rounds, from 4 September to 2 November. Each row shows how many answers from that AI model had a glossary page among the sources, and out of how many answers with a source it started from.

AI modelanswers with a glossary page as a sourceanswers with a source in the 10 rounds
Perplexity22787
Claude22442
Google AIO7745
ChatGPT4463
the four AI models552,437

The glossary appeared as a source in all four AI models, with more answers in Perplexity and Claude, 22 in each. In Google AIO there were 7 answers, and in ChatGPT, 4. The 55 answers were spread across 26 of the 90 questions, and almost all were definition questions.

One page per term or a single page: what the cited URLs show

In the URLs the AI models cited, the glossary that appeared was almost always the page for a single term. Of the 44 glossary URLs cited in murmur’s measurement, 40 covered one term, and only 4 did not. The 44 are URLs, not answers: the same URL may have been a source in more than one answer.

The page for a single term has a URL with the name of the term, a title with the name of the term and a first sentence that defines the term. All three speak to the same question. The URL “glossario/share-of-voice” says what the page is about before it opens, and one URL per term was the format of 40 of the 44 glossary URLs cited.

One term per page also works the other way: the different spelling, the acronym and the synonym stay on the same term page, and do not get a page of their own. “Share of voice” and “presença da marca nas respostas” (brand presence in answers) are the same term and fit on a single page, with both forms written right at the start.

The term page in the glossary: the definition in the first sentence, the example and the neighboring term

The term page in the glossary opens with the definition and ends with the path to the full guide. Each part has a function: the title says which term it is, the first sentence says what it is, the example shows the term in the sector and the neighboring term leads the person to the next question.

  1. The term in the page title, written the way the buyer writes it.
  2. The definition in the first sentence, with the term as the subject, in up to 25 words.
  3. An example from the sector, with a situation, a number or a case the company knows.
  4. The other forms of the term, such as the acronym, the English name and the synonym.
  5. The neighboring term, with the link to its page.
  6. The link to the guide that answers the question in full.

A one-sentence definition does not carry everything, and does not need to: it answers the “what is it”, and the rest of the page gives the example and the context. What matters is that the first sentence, read on its own, still says which term it is about. The passage that makes sense outside the page is the subject of the guide on what AI citability is.

The glossary entry and the guide: when the term calls for a full page

The term calls for a guide, and not just a glossary entry, when the question about it goes beyond the “what is”: how much it costs, how to do it, which one to choose, what the difference is between two. In that case, the glossary page keeps the definition and the link, and the guide answers the rest. Each genuinely different question deserves a page with the full answer, as the guide on how many pages to publish to appear in ChatGPT shows. The acronyms of the GEO market itself, such as AEO and LLMO, have a guide of their own, on which acronym to use and when, to appear in AI answers.

Which terms go in the glossary: the ones the buyer in the sector asks about

The terms that go in the glossary are the ones the buyer in the sector comes across and asks about, not the ones the company would like to teach. The source of those terms is conversations with people who have not bought yet: the questions that reach the sales team, the words that appear in proposals and contracts, the technical names the client repeats without knowing what they mean.

Internal names that only the company uses stay out, such as the nickname of a process or the name of a plan, because nobody asks the AI what they are. Terms the buyer already knows also stay out. The glossary that works is the one for the sector vocabulary that confuses buyers, term by term.

A term that changes with the market calls for review, and a stable one, not so much. In Martinez’s review, with 45 studies published from November 2023 to July 2026, the conclusion about date is direct: “A recent date helps with a time-sensitive query, but not necessarily with a stable definition” (arXiv 2607.14035). The definition of a term that does not change stays as it is; the definition of a term that follows price, rule or tool changes along with it, and the review date appears on the page.

The definition with a number and a source: what the studies by Zhang, He and Yao and by Aggarwal and colleagues measured

A definition that brings a number with a source combines two markers that the studies on AI answers measured. In a study by Zhang, He and Yao published on arXiv in April 2026, with 18,151 pages cited by ChatGPT, Google AIO and Perplexity, the pages most used in the answer carried definition markers 57.33% above the least used pages.

In the study by Aggarwal and colleagues, published on arXiv in 2023, three changes to the page delivered more than the others: citing sources, adding quotations and adding statistics. With them, the share of the AI answer taken from the page grew from 30% to 40%, in a word count that gives more weight to the passage that appears earlier. For the glossary, this means that the definition that cites the standard, the body or the study it came from, with the number written out, has what these tests measured in its favor.

murmur’s method starts with mapping the sector’s questions

The sector’s definition question gets a page that answers it, like any other buyer question. The method murmur publishes has four stages, in its own words: “mapeamento das perguntas do setor, diagnóstico, produção e estruturação do conteúdo, monitoramento contínuo” (“mapping the sector’s questions, diagnostic, production and structuring of the content, ongoing monitoring”). Production is the “escrita dos guias que respondem as perguntas do setor” (“writing of the guides that answer the sector’s questions”), and the mapping comes before it.

The FotoRegistro case study, a print and photo personalization e-commerce, shows the result of covering the market’s questions. On 2 September 2026, 84% of the new questions that came up in its market already had a FotoRegistro article answering them.

Working with each client, murmur studies the company and the sector, analyzes what AI models answer when the buyer asks and who appears in the brand’s place, and tracks the result week by week in ChatGPT, Claude, Perplexity and Google AIO. The content goes out on a subdomain of the client’s brand, without touching the main site, and remains with the company after the contract ends.

Frequently asked questions

How many terms should go in the glossary?

Put one term for each word of the sector that the buyer encounters and asks about, and none just to fill space. The number comes from sales conversations, proposals and contracts, and it changes from one sector to another. The acronym, the synonym and the different spelling of the same term stay on the same page and do not count as new terms.

Can the glossary definition cite another source?

It can, and it gains from it when the definition comes from a standard, a body or a study. Write who defined it and the date in the same sentence, with the link to the original page. In the study by Aggarwal and colleagues, citing sources was among the three changes that most increased the page’s space in the AI answer.

Glossary in alphabetical order or by subject?

The glossary index can be in alphabetical order, which is how a person looks up a word, and each term still has its own page. The subject appears in the links: each term page points to the neighboring terms and to the guide on the topic. That way the index order helps the reader, and the links connect terms on the same subject.

Do English terms go in the glossary?

They do when the sector’s buyer uses the English term, such as “share of voice” in marketing. The page carries the term as the buyer writes it, the definition in Portuguese in the first sentence and the Portuguese name alongside it, when one exists. The English term and the translation stay on the same page, not on two.

Where to start

Start with the sector’s words that your buyer already asks the AI models. The free diagnostic from murmur shows, without sign-up and in less than two minutes, what AI models answer when someone searches for what your company sells, and the full report arrives by email. The sector studies from murmur show which brands AI models recommended in the markets murmur measured, with the questions and the date of the round.

Read also

References

  • Zhang, He and Yao, “From Citation Selection to Citation Absorption”, arXiv 2604.25707v2, 29 April 2026. In 18,151 pages cited by ChatGPT, Google AIO and Perplexity, definition markers 57.33% higher on the most used pages, in a descriptive comparison. https://arxiv.org/abs/2604.25707 (accessed on 8 October 2026).
  • Aggarwal and colleagues, study on GEO and AI answers published on arXiv (2311.09735), 2023. Citing sources, adding quotations and adding statistics had a relative gain of 30% to 40% in the word count adjusted for position. https://arxiv.org/abs/2311.09735 (accessed on 3 October 2026).
  • Martinez, “Optimizing Visibility in Generative Engines: A Critical Survey of GEO (2023-2026)”, arXiv 2607.14035v1, 15 July 2026. Review of 45 studies; the recent date helps with the question that depends on the moment, and not necessarily with the stable definition. https://arxiv.org/abs/2607.14035 (accessed on 8 October 2026).