AEO glossary: GEO, AI SEO and AI search terms explained

Definitions

AEO (answer engine optimization)

The work of making a brand easy for AI assistants to find, understand and name in their answers: ChatGPT, Gemini, Perplexity, Copilot, Claude and Google's AI Overviews and AI Mode. Also called GEO or AI SEO. It builds on SEO, since Google writes that its AI features are rooted in its core Search ranking systems (documented, Google AI optimization guide, 15 May 2026).

GEO (generative engine optimization)

A near synonym of AEO, named after generative engines: assistants that write an answer instead of listing links. GEO is the usual term in French and Dutch. On 10 November 2026, Google Ads showed 1,600 monthly searches for "agence geo" in France and 10 or fewer for "agence aeo". On its own, GEO can also mean geography.

LLMO (large language model optimization)

Another name for the same practice, aimed at large language models rather than at search engines. LLMO and "LLM SEO" are mostly used by English-speaking tech audiences (720 monthly US searches for "llm seo" on 10 November 2026). The work does not change: be reachable, be clear, be cited by others, and measure.

AI SEO

A term with two meanings: SEO done with AI tools (writing, clustering, audits) and SEO for AI answers, which is AEO. Google results for "ai seo agency" mix both. "ai seo" had 6,600 monthly US searches on 10 November 2026. When you hire an agency, ask which of the two it sells.

AI visibility

How present a brand is in AI answers: how often it is named, in which engines, languages and countries, next to which competitors and on the strength of which sources. It is measured from a fixed set of questions asked again and again, because the same question can get a different answer from one day to the next.

Answer engine

A system that answers a question in its own words and may cite sources, instead of returning a list of links. ChatGPT, Gemini, Perplexity, Microsoft Copilot, Claude, Google AI Overviews and Google AI Mode are answer engines. GeckoBoost collects answers from these seven, depending on the plan.

AI Overviews

The AI-written summary Google shows above the classic results for some searches, with links to supporting pages. Google states that a page must be indexed and eligible to show with a snippet, with "no additional technical requirements" (documented, Google Search Central, read 31 October 2026). On 10 November 2026, an AI Overview appeared on 37 of the 40 AEO and GEO result pages we checked.

AI Mode

Google's conversational search mode, where the whole result is an AI answer with links and room for follow-up questions. Google says AI Mode and AI Overviews "may use different models and techniques", so the links they show vary (documented, Google Search Central, read 31 October 2026). Track them as two separate engines.

Query fan-out (fan-out queries)

The searches an AI system runs on its own to build one answer. Google defines fan-out as "a set of concurrent, related queries generated by the model" (documented, Google AI optimization guide, 15 May 2026). On 10 November 2026, ChatGPT's internal searches for French questions such as "meilleur outil AEO" were written in English, and it added a city to questions that named no place.

Grounding

Feeding a model with current documents at the moment it answers, so the answer rests on real pages and not on memory alone. Google calls it retrieval-augmented generation, built on its Search index (documented, Google, 15 May 2026). Microsoft says Copilot and Bing answers rely on the same crawling and index as Bing search (documented, Bing Webmaster Guidelines, read 31 October 2026). A page that is not indexed cannot ground an answer.

Citation

A link or source an AI answer shows next to its text. A citation is not a mention: an engine can name a brand without linking it, or link a page that never names the brand. Some answers have no citation at all, and Google's Gemini help says not every response includes sources (documented, Gemini Apps Help, read 31 October 2026).

Mention

A brand named in the text of an AI answer, with or without a link. Counting mentions takes a dictionary of brand names and their variants, because engines spell names in different ways. A mention can be favourable, neutral or wrong, so read the answers as well as the counts.

Mention rate

The share of collected answers that name a brand, for a given set of questions, engines and period. GeckoBoost shows it with a margin of error, because a rate drawn from 40 answers is less certain than one drawn from 400. A change is reported as beyond the margin of error only when the two ranges do not overlap.

Share of voice

A brand's share of all brand mentions in a set of answers, against the competitors you track. It answers the question "of the brands named, how much is us?". Microsoft's Citation Share in Bing Webmaster Tools and the share of voice in Google Merchant Center are defined differently, so their numbers do not compare directly.

Prompt

The question or instruction a person types into an assistant. In AI visibility tracking, a prompt is one of the fixed questions a tool asks in each engine, language and country. No engine publishes prompt volumes the way Google publishes search volumes, so the question set is a deliberate choice, not a measured demand.

Accuracy check

A review of what AI answers say about a brand: prices, features, founders, locations. An answer can name a brand and still get it wrong. On plans that include accuracy checks, GeckoBoost compares collected answers with the brand's own fact sheet and lists the claims that contradict it, with the answers that made them, so a team can fix the pages they come from.

SERP (search engine results page)

The page a search engine returns for a query: organic results, ads, AI Overview, People Also Ask, videos, forums, local results. On the 40 result pages we checked on 10 November 2026, the top 10 often held only 5 to 10 organic links, because these blocks take places in it.

Zero-click search

A search that ends on the results page without a click to any site, because the answer already shows in an AI Overview, a featured snippet or a knowledge panel. It is one reason AI visibility is measured by mentions and citations, not only by visits.

People Also Ask

The Google block of related questions with short answers that unfold. It is a cheap source of real questions for FAQ sections. On 10 November 2026, its questions for a Dutch search combining AEO and "bureau" in Belgium were mostly about customs licences and certificates, which shows how ambiguous the acronym is in that market.

Training data

The text a model learned from before its release. What a model knows from training stays fixed until its next version, while live search lets it read current pages. Blocking a training crawler (GPTBot, ClaudeBot, or Google-Extended for training) is a separate decision from blocking a search crawler, and the vendors document them as independent settings.

OAI-SearchBot

OpenAI's crawler for ChatGPT search. OpenAI writes that sites opted out of it "will not be shown in ChatGPT search answers" and recommends allowing it in robots.txt (documented, OpenAI bots page, read 31 October 2026). It is independent of GPTBot, so a site can allow search and refuse training.

GPTBot

OpenAI's crawler for content that "may be used in training" its foundation models. Disallowing it tells OpenAI not to train on the site. It does not remove the site from ChatGPT search, which depends on OAI-SearchBot (documented, OpenAI bots page, read 31 October 2026).

ChatGPT-User

The agent ChatGPT uses when a person asks it to open a page during a conversation. OpenAI says robots.txt rules "may not apply" because a user started the fetch, and that this agent does not decide what appears in search (documented, OpenAI bots page, read 31 October 2026). In server logs, its visits show which pages people bring into ChatGPT.

Googlebot

Google's main crawler, which also feeds AI Overviews and AI Mode. Google says robots.txt rules for Googlebot are the control for Search, AI features included, and documents no token that blocks AI Overviews alone (documented, Google crawler documentation, updated 30 October 2026). Blocking Googlebot removes a page from classic results and from AI features at once.

Google-Extended

A robots.txt token, not a separate crawler, that controls whether Google may use a site to train Gemini models and to ground answers in Gemini Apps. It "does not impact a site's inclusion in Google Search" (documented, Google crawler documentation, updated 30 October 2026). It leaves no trace in server logs, so check it by reading robots.txt.

Search generative AI control

A Search Console setting, rolled out to all sites by 31 August 2026, that includes a site in AI Overviews, AI Mode and generative features in Discover, or excludes it. Inclusion is the default. Excluding a site removes its links and grounding from those features, without changing classic ranking (documented, Search Console Help, read 31 October 2026).

Bingbot

Microsoft's standard crawler. Bing writes that Bing and Copilot "rely on the same core crawling, indexing, and ranking foundation", so a page missing from Bing's index is unlikely to be cited by Copilot (documented, Bing Webmaster Guidelines, read 31 October 2026). Check Bing indexing as early as Google indexing.

PerplexityBot

Perplexity's crawler, "designed to surface and link websites in search results" and, per Perplexity, not used to train foundation models. Perplexity recommends allowing it, publishes its IP ranges and says it also relies on third-party crawlers (documented, Perplexity crawler docs, undated, read 31 October 2026). Perplexity-User fetches pages on a person's request and generally ignores robots.txt.

ClaudeBot and Claude-SearchBot

Anthropic's crawlers for Claude. ClaudeBot collects content that may be used for training, Claude-SearchBot indexes pages for search answers, and Claude-User fetches a page when a user asks. Blocking Claude-SearchBot "may reduce your site's visibility" in search answers (documented, Anthropic help article dated 7 April 2026). Each is a separate robots.txt decision.

robots.txt

The text file at the root of a site that tells crawlers which paths they may visit. It controls crawling, not indexing: Bing writes that it "controls crawl access, not indexing" and that noindex keeps a URL out of search and Copilot (documented, Bing Webmaster Guidelines, read 31 October 2026). For AI visibility, allow the search crawlers first: Googlebot, Bingbot, OAI-SearchBot, PerplexityBot, Claude-SearchBot (documented).

llms.txt

A proposed Markdown file at the root of a site that points language models to its key pages. Google states that Search does not use such files and that they have no positive or negative effect there (documented, Google Search Central, 15 June 2026). No other major vendor we checked documents reading it. Keep one if it costs little, without expecting a lift (heuristic).

IndexNow

An open protocol that lets a site tell participating search engines, Bing among them, that a URL was added, changed or deleted. One submission "will be shared across all IndexNow-enabled search engines" (IndexNow FAQ, read 31 October 2026). Bing lists IndexNow among the ways to make URLs discoverable for indexing and grounding (documented, Bing Webmaster Guidelines, read 31 October 2026).

Entity

A thing a search engine or a model can identify without doubt: a company, a person, a product, a place. A brand becomes a clear entity when its name, website, profiles and descriptions agree across the web. A consistent entity is easier to name correctly (heuristic: no vendor documents an entity signal for AI answers).

sameAs

A schema.org property that lists the other official pages of the same entity: LinkedIn, Wikidata, a YouTube channel, review profiles. It sits in a site's Organization markup and helps machines tie those profiles to one brand. No AI vendor documents it as a citation factor, so treat it as tidy housekeeping (heuristic).

Structured data (schema.org)

Markup in a page's code that describes its content to machines: Organization, Article, FAQPage, Product, DefinedTermSet. Google says it "isn't required for generative AI search", while Bing says it "may support clearer grounding" (documented, both read 31 October 2026). Either way, the markup must match the visible text of the page.

Server-side rendering

Sending the full text of a page in its HTML, so a crawler reads it without running JavaScript. Google can process JavaScript, but Bing warns against hiding critical content behind client-side rendering, and Anthropic's web fetch tool does not run JavaScript (documented, both read 31 October 2026). Server-rendered text is the safe choice for every engine (heuristic).

Bing AI Performance

A report in Bing Webmaster Tools, in public preview since 10 February 2026, that shows how often a site is cited in Microsoft Copilot, Bing's AI summaries and some partner integrations. Microsoft says the data is a sample and that a citation does not represent traffic (documented, Bing Webmaster Blog, 10 February 2026). ChatGPT and Gemini are not among the surfaces it lists.

AEO (customs)

In customs, AEO means Authorized Economic Operator, an EU status for trusted importers and exporters. It dominates Belgian searches and weighs on Dutch ones: on 10 November 2026, for a Dutch search combining AEO and "bureau", all 8 organic results in Belgium and 4 of 9 in the Netherlands were about customs. Marketing pages in these markets should spell out answer engine optimization.

Questions