Why a glossary page helps AI find you

One glossary page took my site from about 30 to 120 search impressions a day in 14 days. Why that matters when assistants search before they answer.

By Dasha. Updated November 1 2026. https://heardofyou.io/guides/glossary-page-ai-visibility

A glossary page is the cheapest page on your site that an AI assistant can quote. Buyers ask "what is X" long before "best X for Y". Assistants look things up before they answer.

I built one page defining 20 terms in my field. In two weeks:
- search impressions went from about 30 to 120 a day,
- clicks went from about 1 to 5 a day,
- distinct queries I appeared for went from 11 to 47.

Here is what I did, what I would change, and why it matters if you want ChatGPT, Claude, Gemini or Perplexity to name you.

Why a definitions page matters for AI

For most product questions, assistants search the web first (/guides/how-ai-decides-what-to-recommend). They split one question into several smaller searches.

A buyer types "best invoicing app for freelancers". The assistant runs the searches you would expect. It also runs "what is recurring invoicing" or "what does net 30 mean". Those are glossary questions.

If your site answers them clearly, the assistant may read you on its way to the recommendation. If not, it reads a competitor's page. That competitor becomes the brand it already knows.

A glossary catches a wide spread of that vocabulary on one page. No single blog post can. That was the whole trick, and it is not clever. It is a page that answers the small questions plainly.

What I built

I opened my search console, sorted by impressions, and filtered for queries with zero clicks. These were questions I already appeared for but was not winning: "what is an MCP server", "what is pgvector", that sort of thing. No page of mine defined any of them. So I made one.

One HTML page, 20 terms, one afternoon. Each entry runs about 120 words, in four parts:

1. The term and a one-line definition. This is what a search engine or an assistant quotes. Write a complete sentence that stands alone.
2. "What it actually means." About 80 words of plain English, as you would explain it to a smart friend.
3. "You would care about this if..." One line on who needs this and why. It keeps the reader on the page. It also tells an assistant which buyer the term belongs to.
4. A link to the page on your site that uses the idea: a guide, a comparison, your product page.

Every entry has its own anchor, so other pages can link straight to a definition. I added DefinedTermSet and DefinedTerm markup. I added FAQPage markup too, since half the entries were questions anyway.

What happened

I wrote these down as ratios, not screenshots. The baseline is small, so read them as a direction, not a law.

| Metric | Before | After 14 days |
|---|---|---|
| Impressions a day, whole site | about 30 | about 120 |
| Clicks a day | about 1 | about 5 |
| Indexed pages | 18 | 23 |
| Queries with at least one impression | 11 | 47 |

The five new indexed pages were the glossary plus four single entries. Google chose to show those as their own results. Each entry started acting like a small page of its own. I did not expect that, and it explains the jump in queries.

This was a search experiment, not an AI-mention experiment. I did not run buyer questions through four assistants before and after. I regret that, because it is exactly what I would do today. Our how we measure (/how-we-measure) page is the method I would have used. The honest claim: the glossary got me into more of the searches assistants run. That is a necessary step, not the finish line.

What I would change
- Start from queries you already get impressions for. They show the words buyers actually use. Your marketing words are not on that list. Keep them off the glossary too.
- Add 5 to 10 entries at a time. Search engines index anchors slowly. Batches show you which entries get picked up, so you can write more like those.
- Link first mentions. Whenever a guide uses a term, link it to the glossary anchor. That gives the assistant a path: question, definition, your product.
- Write the "you would care if" line first. If you cannot write it, nobody is asking about the term. Cut it.
- Do not chase uniqueness. Your "net 30" will read like everyone else's. Clear beats clever. The link to what you built is what makes it yours.

What does not work

A keyword-stuffed glossary is a 2014 trick, and nothing rewards it now. Each definition has to be a real, useful one, or the page is just a list of words.

The markup is housekeeping, not strategy. No evidence shows schema alone makes an assistant cite or recommend a page. We say so in what works and what doesn't (/guides/what-works-and-what-doesnt). Add the markup, match it to the visible text, and spend the real effort on the sentences.

Ours

We practise this. Our own glossary (/glossary) defines AI visibility, GEO, mention rate, fan-out, grounding, earned media, llms.txt and the rest. Each has a one-line definition, a plain paragraph and a line on who should care.

We log whether assistants ever cite it in our visibility log (/research/heard-of-you-visibility-log). It starts near zero and is published unedited.

A first version of this appeared on my studio blog, dashbuilds.dev.

What to do next

Before you write a glossary, find out which questions assistants already answer about your category, and who they name. The free check (/check) runs 5 buyer questions, 3 times each, on ChatGPT, Gemini, Claude, Perplexity and Muse. It shows which sources they cited. The words on those pages are your first 10 terms.

Questions

Did the glossary change what AI assistants said about my site?

I did not measure that at the time, and I will not pretend I did. I measured Google: 4x impressions, 5x clicks, 4x distinct queries in 14 days. Assistants search before they answer, so showing up for more questions is the first step, not the result.

How many terms do I need?

I shipped 20 in one afternoon. Next time I would start with 5 to 10 and add batches, since search engines index entries slowly anyway. Twenty good entries beat a hundred thin ones.

Does the schema markup matter?

I added DefinedTerm and FAQPage markup, and four entries got their own Google listings. That is a search result, not an AI recommendation. No evidence shows markup alone makes an assistant cite you. Treat it as housekeeping and put the effort into the definitions.

Will this work for a product that is not technical?

Yes. Every category has words buyers type before they know what to buy: candle wax types, tax terms, dollhouse scales. If you can write the "you would care about this if" line for a term, the term belongs on your page.