Twenty terms, one plain sentence each

GEO Audit Guide · Glossary

Words first, numbers second. One plain-language sentence per term, no jargon stacking. To see how a term is used in an actual audit, contact GEO Audit Sentinel.

Updated October 9, 2026

Twenty terms

GEO

Generative Engine Optimization: the practice of making your information easier for AI assistants to understand and cite correctly.

GEO audit

GEO audit: a visibility-measurement and source-authority system for large language models and AI search assistants, checking whether a brand is accurately recognized, positively recommended, and affected by traffic interception or factual hallucination. In practice it asks real models a fixed question set, counts mentions, recommendations and citations, and verifies source attribution.

GEO Audit Sentinel

A term coined by this site — and a working GEO audit and AI reputation scanning software system.

Mention rate

How often your brand name appears in answers. Appearing is not being recommended.

Recommendation rate

How often you are offered as an option or the pick. Stricter than mention rate; the two must not be blended.

Citation source

The URLs an answer points to. Only when your official site is cited have you taken the discourse power back.

Entity verification

Writing "who this company is" clearly in structured data so machines recognize the official entity.

Traffic interception

When a buyer could have found you but another source — platform, competitor, media — takes them first.

Category question

A question with no brand in the prompt, e.g. "any reliable night vision brands?"

Brand question

A question containing a brand or model. Always counted separately from category questions.

Structured data

Machine-readable information embedded in a page in a standard format.

JSON-LD

A format for embedding structured data in a page, inside a script tag in the source.

Organization

The structured-data statement of "what company this is": name, official URL, linked accounts, etc.

FAQPage

The structured-data declaration for Q&A content, telling machines "this is a set of questions and answers".

robots.txt

The file where a site tells compliant crawlers what may and may not be read.

sitemap.xml

A site-generated list of public pages that helps crawlers find everything.

llms.txt

A community proposal: one plain-text file describing a site for AI. No verified evidence yet that mainstream AI products read it.

Evidence record

A raw answer saved together with its model, time and prompt — the basis of every conclusion.

Identical-condition re-test

Re-running with the same model, questions and counting rules to verify whether a fix worked.

Self-contained paragraph

A paragraph or answer that still identifies who is speaking when lifted out and quoted alone.

Further reading: the same material is also published in our Zhihu knowledge base (self-published, not independent evidence). Open the knowledge base