---
title: "The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress"
description: "An AI visibility report can show mentions up 40%, a score of 78, and 300 bot hits while telling you almost nothing about buyers."
url: "https://groas.com/post/the-ai-visibility-glossary-18-terms-vend"
image: "https://groas.com/media/blog/df8f7cf54608582d87caafc9a4560e258756e181379347ff15d89e6047e926dd.png"
published: "2026-10-05T05:48:28.275Z"
modified: "2026-10-05T05:48:28.866Z"
---

[AI For Google Ads](https://groas.com/category/ai-for-google-ads) · October 5, 2026 · 10 min read

# The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress

[DavidFounder & CEO @ groas](https://groas.com/author/david)

![Cover image for: The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress](https://groas.com/media/blog/df8f7cf54608582d87caafc9a4560e258756e181379347ff15d89e6047e926dd.png)

In this article

1. [Mention vs. citation](#mention-vs-citation)
2. [Prompt set](#prompt-set)
3. [AI share of voice](#ai-share-of-voice)
4. [AI Overview, AI Mode, answer engine](#ai-overview-ai-mode-answer-engine)
5. [Grounding and retrieval](#grounding-and-retrieval)
6. [Query fan-out](#query-fan-out)
7. [Training crawler vs. user-triggered fetcher](#training-crawler-vs-user-triggered-fetcher)
8. [Verified vs. unverified bot hits](#verified-vs-unverified-bot-hits)
9. [Live-answer fetch](#live-answer-fetch)
10. [robots.txt and AI bot rules](#robotstxt-and-ai-bot-rules)
11. [llms.txt](#llmstxt)
12. [Render-blocking JavaScript](#render-blocking-javascript)
13. [Entity](#entity)
14. [Answer-first passage](#answer-first-passage)
15. [Schema markup](#schema-markup)
16. [Off-site citation sources](#off-site-citation-sources)
17. [AEO vs. GEO vs. SEO](#aeo-vs-geo-vs-seo)
18. [Visibility score](#visibility-score)

An AI visibility report can show mentions up 40%, a score of 78, and 300 bot hits while telling you almost nothing about buyers. I watched the same trick in Google Ads: impressions and optimisation scores climbed while CPA went nowhere.

The labels have changed; the temptation has not. Each term below has a useful meaning and a way to misuse it. **A mention is not a citation. A visibility score is not revenue. A crawler hit is not a reader.**

## Mention vs. citation

A **mention** names your brand in an AI answer; a **citation** links to a page the reader can click.

I used to tell clients a mention was halfway to a click. I was wrong. In one study, [61.7% of citations were ghost citations: the source was linked, but the brand was never named](https://fitsmallbusiness.com/marketing/ai-visibility-metrics-small-business/). The distinction cuts both ways. A report that rolls mentions and linked citations into one win count hides which outcome you got. Ask for two columns: times named and times linked. If the vendor cannot split them, you are buying impressions again.

## Prompt set

A **prompt set** is the list of questions a tool repeatedly asks AI engines to check whether your business appears.

The list matters more than its size. Answers vary between runs, so 14 appearances in 20 runs tell you more than one flattering screenshot. But a set stuffed with branded questions like _is \[Brand] any good_ or obscure questions no buyer asks will inflate the average. Keep branded prompts in their own bucket. For category coverage, start with [25 to 40 prompts across category, problem, comparison, and brand types, then prune monthly](http://aiclicks.io/blog/how-to-choose-prompt-to-track). **Inspect the questions before you trust the percentage.**

## AI share of voice

**AI share of voice** is the percentage of tracked answers in which your business appears, either by name or by link.

The misuse is blending every appearance across every engine. ChatGPT, Perplexity, and AI Overviews retrieve differently, so [track mentions and citations by platform while holding wording, intent, and market constant](https://fitsmallbusiness.com/marketing/ai-visibility-metrics-small-business/). Say you spend $20k a month and a report gives you 32% share of voice. Is that 50% in Perplexity, 8% in AI Overviews, and zero in ChatGPT? Find out before you decide what to fix. **One blended number conceals the weak spot.**

## AI Overview, AI Mode, answer engine

**AI Overviews** are summaries within standard Google results, **AI Mode** is a separate conversational search experience, and an **answer engine** is a broader label for services that generate answers rather than just lists of links.

Those are different places for a buyer to encounter you. A page that appears in one will not necessarily appear in another, and an answer-engine total tells you even less about where it appeared. If a vendor gives you one AI visibility number for Google, ask which surface it measured. **Do not let a surface label stand in for a placement.**

## Grounding and retrieval

**Grounding** is an engine’s use of outside material to support an answer; **retrieval** is the process of finding that material.

I learned the access lesson managing Shopping feeds: if Google could not read the product information, polishing the campaign did not help. The same question comes before any AI content rewrite: can the relevant system reach and read the page? Engines can assemble answers from related searches and selected passages rather than the page you would have picked for the exact query. **Check access before you pay someone to rewrite a passage the engine never saw.**

## Query fan-out

**Query fan-out** is the process of turning one question into several related searches before assembling an answer.

A buyer may ask one broad question while the engine looks for comparisons, costs, safety details, or recent information. A page tuned only to the buyer’s exact wording can miss those searches; [content irrelevant to the expanded set may never be retrieved](https://searchengineland.com/how-ai-mode-ai-overviews-work-patents-456346). The misuse is single-keyword thinking with an AI label stuck on it. If you sell pest control, cover the comparison, price, and safety questions, not just the head term. **Track the question cluster, not one phrase.**

## Training crawler vs. user-triggered fetcher

A **training crawler** collects pages for model development, while a **user-triggered fetcher** requests a page in response to someone’s interaction with an AI service.

GPTBot and a live page fetch do different jobs, even when the names on the log lines look related. Vendors point to a spike in training-crawler hits as proof that buyers are seeing a brand. It proves no such thing. A user-triggered fetch is more relevant to an answer being assembled, but even that does not prove the page was cited. **Split the agents in your logs before you celebrate.**

## Verified vs. unverified bot hits

A **verified bot hit** is a request whose claimed crawler identity you have checked, rather than accepted from its user-agent name alone.

Any script can put `GPTBot` in a header. The name in the log is someone knocking and announcing who they are; verification is checking. Reverse DNS and a forward lookup can help establish whether an IP belongs to the claimed provider. Vendors misuse raw hit counts by calling them demand. First separate verified requests from impersonators, then separate training crawls from user-triggered fetches. **An unverified crawler hit is not a reader.**

## Live-answer fetch

A **live-answer fetch** is a page request made while an AI service is gathering material for a user’s answer.

It is more interesting than a training crawl because it happens in the answer pipeline. It is not a click, a citation, or revenue. Logs may show that a user-triggered fetcher requested a buying page; they will not, on their own, show what the final answer said. That distinction matters when a report promotes bot traffic into a business result. If relevant pages are inaccessible, start with [the technical issues that block AI visibility](https://groas.com/post/fix-technical-seo-issues-hurt-ai-visibility) before rewriting copy. **A fetch is a chance to be used, not proof that you were.**

## robots.txt and AI bot rules

**robots.txt** is a site file that tells named crawlers which paths they may or may not fetch.

The misuse is treating one bot rule as a master switch for every AI use. Training crawlers and bots that fetch material for answers can have different names and purposes. Blocking a training crawler does not necessarily block live access to a public page; [training restrictions and inference-time access are different questions](https://searchengineland.com/llms-txt-isnt-robots-txt-its-a-treasure-map-for-ai-456586). Check the rule against the specific agent and path you care about. **Write bot rules for the fetch you want to allow or prevent.**

## llms.txt

**llms.txt** is a Markdown file at `/llms.txt` that presents a curated list of pages for AI systems to read.

It is [not a replacement for robots.txt and does not block crawling or dictate indexing](https://searchengineland.com/llms-txt-isnt-robots-txt-its-a-treasure-map-for-ai-456586). Think of robots.txt as an access instruction and _llms.txt_ as a proposed menu, not a guarantee that anyone orders from it. Publishing one does not earn a citation; deleting one does not prevent an answer. I would put together a useful list if it took twenty minutes. **I would not pay a retainer for the file.**

## Render-blocking JavaScript

**Render-blocking JavaScript** is code that leaves important page text unavailable until a browser runs the scripts needed to display it.

A buying page can look complete in Chrome and give a simple text fetch almost nothing. If a fetcher cannot run the scripts or wait for them, the answer is missing before anyone can judge how well it is written. Vendors misuse this problem by selling more content. Fetch a buying page as a bot would first. **If the text is absent, you have a delivery problem, not a word-count problem.**

## Entity

An **entity** is the identifiable business behind the names, services, locations, and other facts an engine finds across sources.

Think category, service area, reviews, and consistent business details. If those details disagree from one listing to another, the engine has a harder job identifying and corroborating you. The misuse is selling _entity work_ as though cleaning up listings buys a linked citation tomorrow. It does not. Make the business legible across the places buyers and engines check; then measure whether it actually appears in relevant answers. **Consistency supports recognition. It is not a citation count.**

## Answer-first passage

An **answer-first passage** is a short section that states the answer before giving its supporting detail.

A direct, self-contained passage is easier to use when an engine selects excerpts to build an answer. Ten fluffy paragraphs give it more work and may leave your point unattributed. For the writing method, see [how to get cited with passage-first writing](https://groas.com/post/how-to-get-your-business-cited-in-ai-ove). The misuse is commissioning a rewrite of the whole blog before looking at the pages closest to buying intent. **Fix the answer on those pages first.**

## Schema markup

**Schema markup** is code that labels page details, such as prices or reviews, for machines.

It has uses in traditional search. That does not make it an AI citation strategy. An Ahrefs test of 1,885 pages that added schema against 4,000 controls found [small citation changes across AI Overviews, AI Mode, and ChatGPT that were statistically indistinguishable from zero](http://marcodiversi.com/blog/does-schema-help-ai-citations/). Google also says [new files or special schema.org markup are not required to appear in AI Overviews or AI Mode](https://searchengineland.com/how-ai-mode-ai-overviews-work-patents-456346). Keep schema for the job it does. **Do not buy it as proof of future citations.**

## Off-site citation sources

**Off-site citation sources** are third-party pages, such as reviews, directories, forums, and publisher lists, that an engine can use when answering questions about your business.

Your homepage says you are good at the job. Other sources can help establish whether that claim holds up. For a local business, consistent directory categories and reviews can matter alongside a clear site; for another category, the relevant discussion may happen on forums. The misuse is calling on-site content the whole job. Find where your category is discussed, then check whether those sources describe your business accurately. **Write the answer at home; build corroboration elsewhere.**

## AEO vs. GEO vs. SEO

**SEO** focuses on earning search visibility and clicks, **AEO** on appearing as a direct answer, and **GEO** on making information usable in generative answers.

That is a [working distinction vendors often blur](https://www.yext.com/blog/seo-vs-aeo-vs-geo). The misuse is selling three retainers for three labels when much of the work overlaps: readable pages, direct answers, accessible content, and consistent off-site information. I would skip a vendor who spends more time separating the acronyms than showing the passage, the fetch, and the link. **Labels do not get cited. Pages do.**

![Cartoon of a marketer celebrating a vanity score while buyer clicks go elsewhere](https://groas.com/media/blog/f4d82c06abebdb1dd31f487d955324bafe5ea87e8a8babf66f1fa05060de12b4.png)

## Visibility score

A **visibility score** is a vendor’s formula for combining AI appearances into one number, often on a 0-to-100 scale.

This is the term I would retire. The vendor chooses the prompt set, engines, and weights. Change that mix and the score can move without a buyer doing anything different. On a fixed methodology it can serve as a trend line; it cannot stand in for revenue. What the deck calls a health metric, I call optimisation score with a new haircut.

If a report lands on your desk, ask for three things before discussing its score:

1. **Linked citations on buying prompts, split by engine.** Keep mentions in a separate column.
2. **The prompt list itself.** Put branded questions in their own bucket so they cannot inflate category coverage.
3. **Verified user-triggered fetches to relevant pages.** Do not mistake these for citations or buyers.

Analytics needs the same care: [Google Analytics’ AI Assistant channel for recognised AI referrers excludes Google AI Overviews and AI Mode, which remain under Organic Search](https://fitsmallbusiness.com/marketing/ai-visibility-metrics-small-business/). GA alone will not show the whole picture. A fixed prompt panel run daily for 30 days, compared engine by engine, gives you a more useful view than one screenshot; I lay out the setup in [how to measure visibility](https://groas.com/post/how-to-measure-your-brand-s-visibility-i-2).

![Diagram distinguishing live fetches, training crawls, and off-site sources in an AI answer](https://groas.com/media/blog/38939d4089c85c1ffc01970969d13a8b650379ffe011071fc35c8ef72f45e640.png)

I watched optimisation score keep clients calm in Google Ads while CPA drifted. I do not need its AI-search cousin doing the same job. If the score rises but linked citations on buying prompts stay flat, ask what changed. If the answer is a new prompt mix or more crawler hits, you are buying impressions again, with better fonts.

## Frequently Asked Questions

### What is the difference between an AI mention and a citation?

A mention names your brand in an AI answer, while a citation links to a page the reader can click. In one study, 61.7% of citations were ghost citations, meaning the source was linked but the brand was never named. Ask vendors for two separate columns: times named and times linked.

### How many prompts should I track for AI visibility?

Start with 25 to 40 prompts covering category, problem, comparison, and brand types, then prune the list monthly. Keep branded questions like 'is \[Brand] any good' in their own bucket so they cannot inflate category coverage, and inspect the actual questions before trusting any percentage.

### Why should I track AI share of voice by platform instead of one blended number?

ChatGPT, Perplexity, and AI Overviews retrieve content differently, so one blended number conceals the weak spot. A 32% overall share could be 50% in Perplexity, 8% in AI Overviews, and zero in ChatGPT. Track mentions and citations by platform while keeping wording, intent, and market constant.

### Are AI Overviews, AI Mode, and answer engines the same thing?

No. AI Overviews are summaries within standard Google results, AI Mode is a separate conversational search experience, and answer engine is a broader label for services that generate answers rather than lists of links. A page that appears on one surface will not necessarily appear on another, so ask which surface a vendor's number measured.

### Do bot hits in my logs mean buyers are seeing my brand in AI answers?

No. A training crawler hit only means a page was collected for model development, and any script can put GPTBot in a header, so verify crawler identity through reverse DNS and a forward lookup. Even a user-triggered fetch during an answer does not prove the page was cited.

### Does adding schema markup get my pages cited in AI answers?

No. An Ahrefs test of 1,885 pages that added schema against 4,000 controls found small citation changes across AI Overviews, AI Mode, and ChatGPT that were statistically indistinguishable from zero. Google also says new files or special schema.org markup are not required to appear in AI Overviews or AI Mode.

### Should I trust a vendor's AI visibility score?

Treat it with caution. The vendor chooses the prompt set, engines, and weights, so the score can move without any buyer behaviour changing. On a fixed methodology it can serve as a trend line, but it cannot stand in for revenue or replace linked citations on buying prompts split by engine.

### Does llms.txt block AI crawlers from my site?

No. llms.txt is a Markdown file at /llms.txt presenting a curated list of pages for AI systems, and it does not block crawling or dictate indexing. Use robots.txt as the access instruction for named crawlers, and treat llms.txt as a proposed menu rather than a guarantee that anyone orders from it.

## Related Posts

- [![Clay figure pours a wheelbarrow of articles into a wall funnel whose pipe is disconnected, so the pages pile on the floor beside an unused wrench.](https://groas.com/media/blog/e1a99963f159a4756e614472ee73400643c5900fa7bde718abe8fa563e8cdc77.png)](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
  ### [Can ChatGPT Read Your Site? A 45-Minute Check Before You Write More AEO Content](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
  [October 6, 2026 • 10 min read](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)

  [Written by Alexander Perelman](https://groas.com/post/can-chatgpt-even-read-your-site-a-45-min)
- [![Cartoon of a marketer noting 'position two' after one pull of a slot machine whose reels are already spinning to a new result.](https://groas.com/media/blog/038bfefb4237637a1718ff66b379d5bdc42d57615f25797529aa5ef5682a4f7e.png)](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
  ### [Six PPC Habits That Misread AI Visibility](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
  [October 6, 2026 • 10 min read](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)

  [Written by David](https://groas.com/post/i-measured-ai-visibility-like-a-ppc-acco)
- [![A tiny worker on a ladder fixes up five model shopfronts while a giant dusty 'take a number' dispenser looms behind, its ticket queue ignored.](https://groas.com/media/blog/35c8ce6c06242ab08178711c5abe08dda0df51ec927a2dd6bc9770aa5f83bdbf.png)](https://groas.com/post/one-strategist-zero-developers-five-clie)
  ### [One Strategist, Zero Developers, Five Domains: A 30-Day AEO Launch Playbook](https://groas.com/post/one-strategist-zero-developers-five-clie)
  [October 6, 2026 • 10 min read](https://groas.com/post/one-strategist-zero-developers-five-clie)

  [Written by David](https://groas.com/post/one-strategist-zero-developers-five-clie)

## The Machines Already Run Search, You Should Probably Own One

[get started free](https://groas.typeform.com/to/xC1bQNUT)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#webpage","url":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend","name":"The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress","description":"An AI visibility report can show mentions up 40%, a score of 78, and 300 bot hits while telling you almost nothing about buyers.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://groas.com/media/blog/df8f7cf54608582d87caafc9a4560e258756e181379347ff15d89e6047e926dd.png"},"breadcrumb":{"@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#breadcrumb"},"dateModified":"2026-10-05T05:48:28.866Z","mainEntity":{"@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"AI For Google Ads","item":"https://groas.com/category/ai-for-google-ads"},{"@type":"ListItem","position":4,"name":"The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress","item":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#article","url":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend","mainEntityOfPage":{"@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"The AI Visibility Glossary: 18 Terms That Can Make Nothing Look Like Progress","description":"An AI visibility report can show mentions up 40%, a score of 78, and 300 bot hits while telling you almost nothing about buyers. I watched the same trick in Google Ads: impressions and optimisation scores climbed while CPA went nowhere.…","datePublished":"2026-10-05T05:48:28.275Z","dateModified":"2026-10-05T05:48:28.866Z","image":{"@type":"ImageObject","url":"https://groas.com/media/blog/df8f7cf54608582d87caafc9a4560e258756e181379347ff15d89e6047e926dd.png"},"author":{"@id":"https://groas.com/author/david#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"articleSection":"AI For Google Ads","keywords":"Google Ads, AI For Google Ads","wordCount":2020,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/david#person","name":"David","jobTitle":"Founder & CEO @ groas","image":"https://groas.com/media/blog/363be9ad5255816654dac065d01bf1c6063110b400bb38cc199570267232027c.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"FAQPage","@id":"https://groas.com/post/the-ai-visibility-glossary-18-terms-vend#faq","mainEntity":[{"@type":"Question","name":"What is the difference between an AI mention and a citation?","acceptedAnswer":{"@type":"Answer","text":"A mention names your brand in an AI answer, while a citation links to a page the reader can click. In one study, 61.7% of citations were ghost citations, meaning the source was linked but the brand was never named. Ask vendors for two separate columns: times named and times linked."}},{"@type":"Question","name":"How many prompts should I track for AI visibility?","acceptedAnswer":{"@type":"Answer","text":"Start with 25 to 40 prompts covering category, problem, comparison, and brand types, then prune the list monthly. Keep branded questions like 'is [Brand] any good' in their own bucket so they cannot inflate category coverage, and inspect the actual questions before trusting any percentage."}},{"@type":"Question","name":"Why should I track AI share of voice by platform instead of one blended number?","acceptedAnswer":{"@type":"Answer","text":"ChatGPT, Perplexity, and AI Overviews retrieve content differently, so one blended number conceals the weak spot. A 32% overall share could be 50% in Perplexity, 8% in AI Overviews, and zero in ChatGPT. Track mentions and citations by platform while keeping wording, intent, and market constant."}},{"@type":"Question","name":"Are AI Overviews, AI Mode, and answer engines the same thing?","acceptedAnswer":{"@type":"Answer","text":"No. AI Overviews are summaries within standard Google results, AI Mode is a separate conversational search experience, and answer engine is a broader label for services that generate answers rather than lists of links. A page that appears on one surface will not necessarily appear on another, so ask which surface a vendor's number measured."}},{"@type":"Question","name":"Do bot hits in my logs mean buyers are seeing my brand in AI answers?","acceptedAnswer":{"@type":"Answer","text":"No. A training crawler hit only means a page was collected for model development, and any script can put GPTBot in a header, so verify crawler identity through reverse DNS and a forward lookup. Even a user-triggered fetch during an answer does not prove the page was cited."}},{"@type":"Question","name":"Does adding schema markup get my pages cited in AI answers?","acceptedAnswer":{"@type":"Answer","text":"No. An Ahrefs test of 1,885 pages that added schema against 4,000 controls found small citation changes across AI Overviews, AI Mode, and ChatGPT that were statistically indistinguishable from zero. Google also says new files or special schema.org markup are not required to appear in AI Overviews or AI Mode."}},{"@type":"Question","name":"Should I trust a vendor's AI visibility score?","acceptedAnswer":{"@type":"Answer","text":"Treat it with caution. The vendor chooses the prompt set, engines, and weights, so the score can move without any buyer behaviour changing. On a fixed methodology it can serve as a trend line, but it cannot stand in for revenue or replace linked citations on buying prompts split by engine."}},{"@type":"Question","name":"Does llms.txt block AI crawlers from my site?","acceptedAnswer":{"@type":"Answer","text":"No. llms.txt is a Markdown file at /llms.txt presenting a curated list of pages for AI systems, and it does not block crawling or dictate indexing. Use robots.txt as the access instruction for named crawlers, and treat llms.txt as a proposed menu rather than a guarantee that anyone orders from it."}}]}]}
```