The words behind the work.

Answer engine optimization — work aimed at being the source of direct answers.

The term overlaps heavily with GEO and with parts of traditional SEO such as featured snippets. It is not a separate technical switch.

See also: GEO

Google's AI-generated summaries shown on some search results pages.

AI Overviews draw on Google Search systems. Controls for Search eligibility and snippets apply; no special markup is required. Citeroot does not collect AI Overviews today; it collects from ChatGPT, Claude, Gemini and Perplexity.

See also: Answer engine

The observable presence of a brand or site across a defined set of AI answers.

AI visibility is measured, not assumed: it depends on which questions you ask, which engines you collect from, where and when. A claim of visibility without those details is incomplete.

See also: Mention, Citation, Share of voice

A short, self-contained paragraph that answers a question directly.

Placed near the top of a page, an answer capsule states the answer, any key condition and the scope in a couple of sentences, so it can be quoted without surrounding context.

See also: GEO

A system that responds to questions with a generated answer rather than a list of links.

Examples include chat assistants and AI search experiences. Many combine a language model with retrieval from the web.

See also: AEO, GEO

A link or reference to a source within an AI answer.

A citation can exist without a brand mention and vice versa. Citeroot reports them separately, with their own rates.

See also: Mention, Source map

Independent sources repeating or confirming a claim about you.

A claim that appears only on your own website is weaker evidence than one repeated by independent sources.

See also: Citation

AI-related bots differ by purpose: building a search index, collecting training data, or fetching a page for a user.

A robots.txt policy can treat each purpose differently. Blocking a training crawler does not necessarily block a search crawler, and vice versa.

See also: robots.txt

Generative engine optimization — improving how a business and its content appear in AI-generated answers.

The term was popularised by a 2023 research paper presented at KDD 2024. It overlaps with SEO and AEO and is not a guarantee of any placement.

See also: AEO, AI visibility

Supplying a model with retrieved sources so its answer reflects them.

Grounded answers can cite their sources. Ungrounded answers rely on what the model learned in training.

See also: Citation

A proposed convention for a plain-text site summary aimed at language models.

It is not a standard that all engines honour, and no major search provider requires it. Treat it as optional documentation, not a ranking lever.

See also: robots.txt

Your brand named in the text of an answer.

A mention doesn't require a link. Mention rate is the share of valid answers that name the brand.

See also: Citation, Recommendation rate

The fixed set of questions you track over time.

A good panel covers the buying journey, mixes branded and non-branded questions and changes rarely, so results stay comparable.

See also: AI visibility

The share of answers that explicitly recommend a brand for the use case in the prompt.

Recommendations are a stronger signal than mentions and are tracked with their own denominator.

See also: Mention

The step where a system selects documents to use when composing an answer.

Being crawlable does not guarantee being retrieved, and being retrieved does not guarantee being cited.

See also: Grounding

A file that tells crawlers which paths they may fetch.

Each user agent can be allowed or blocked separately. It is advisory for well-behaved bots, not an access-control system.

See also: Crawler (search, training, user)

A brand's mentions as a share of all tracked brands' mentions in a set of answers.

Always report the denominator: which prompts, which engines, which period. Citeroot computes it at the level of answers: answers naming you ÷ the sum, across all tracked brands, of answers naming each brand. Failed runs are excluded.

See also: Mention

A view of which domains and pages are cited across a set of answers.

Source maps reveal which publishers influence a category and where competitors are cited but you are not.

See also: Citation, Corroboration

Machine-readable markup (often JSON-LD) that describes a page's content.

It can help systems understand entities and facts on a page. It is not a guarantee of selection for any feature.

See also: Retrieval

Run-to-run difference in answers to the same prompt.

AI answers are non-deterministic. Repeated runs let you tell a real change from noise.

See also: Prompt panel