AEO / GEO
Work with Lifewood

Reference

Glossary of AEO and GEO terms

Terms as used across this site, defined tightly enough to quote without distortion. Where the wider market uses a term loosely or sells a contested idea as settled, that is noted rather than smoothed over.

Start reading The 2026 numbers

Updated 28 September 2026 · 5 min read · Sources linked throughout

Disciplines

The three practices

AEO, Answer Engine Optimization
Earning citations inside AI-generated answers. The unit of optimisation is an extractable passage rather than a page, and the metric is citation share across a fixed question set. See the explainer.
GEO, Generative Engine Optimization
Shaping how a generative system describes an entity, whether or not it links to it. Given its precise sense by Aggarwal et al., ACM SIGKDD 2024. Measured by mention rate and by whether the description is accurate.
SEO, Search Engine Optimization
Earning position in a ranked list of links, where the outcome is a click. Still the source of the large majority of referral traffic in 2026.
Answer engine
Any system returning a synthesized response rather than a list of links, ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews, Copilot. The defining property is that the question is answered in place, usually with no click.

Mechanics

How answers get assembled

Retrieval (mode)
An answer produced with live web search enabled. Reflects what the system can find at the time of asking; responds to published content within days to weeks.
Model memory (mode)
An answer produced from training data with no search. Changes only when a model is retrained, months to years, and is moved by third-party coverage rather than your own site.
RAG, Retrieval-Augmented Generation
The architecture behind retrieval answers: fetch relevant documents, then generate conditioned on them. Explains why extractable passages matter: the system quotes what it retrieved.
Chunking
Splitting a retrieved document into passages that are ranked individually. The reason the unit of selection is a span of text, not a URL.
Extractable passage
A self-contained span that answers a question without surrounding context. The practical unit of AEO; a page without one is hard to quote regardless of quality.
Grounding
Attaching a generated answer to retrieved sources so claims can be traced. An ungrounded answer is produced from parameters alone and will invent a description if the web supplies none.
Hallucination
A confidently stated claim with no basis in source material. Relevant to GEO because a system with no good source for your company will fabricate one.

Measurement

Metrics and their traps

Citation
A link to a source inside a generated answer. Worth tracking separately from mention. Note that AI Overview citations are clicked roughly 1% of the time, citation is an influence metric, not an acquisition channel.
Mention
An entity named in an answer without a link. Commercially meaningful, since it still shapes a shortlist, and invisible to any measurement that counts only links.
Citation share
The proportion of answers to a fixed question set in which an entity is cited. Only interpretable when reported per surface and per language.
Share of answer
Broader than citation share: how much of the answer space for a topic an entity occupies, counting citations and mentions together.
Zero-click
A query resolved without the user visiting any site. Reported at roughly 68% of Google searches in early 2026.
Control question
A tracked question in a category you do not sell into, included so a change in your target questions can be read against something. Without controls, everything moving at once is indistinguishable from the models changing.
'AI visibility score'
A single blended metric sold by several tools. Treat with caution: retrieval and model memory move on different timescales for different reasons, so one number averaging them cannot be acted on.

Entity

Identity and structure

Entity resolution
Whether a system can determine that a brand name, a legal name and a domain refer to one organisation. Failure here makes all downstream content unattributable, which is why it comes first.
Category association
Whether anything on the open web plainly states which category an entity belongs to. A common failure mode is a well-known company that no source explicitly places in the category it sells into.
alternateName
A schema.org property listing other names for the same entity. The standard way to state that a short brand name and a full legal name are one organisation.
sameAs
A schema.org property pointing at other authoritative profiles for the same entity. An unverified sameAs is worse than none, because it asserts an identity claim the crawler will follow and fail to confirm.
Structured data / JSON-LD
Machine-readable statements describing what a page is about. Helps parsing and entity resolution. The narrower claim that it directly raises citation rates is not well evidenced, treat as hygiene, not a lever.
Provenance
The traceability of a claim to a dated, attributable, identifiable source. The property that makes a statement safe for a model to quote.
llms.txt
A proposed convention for publishing a machine-readable site summary at a well-known path. Contested: Google states it has no effect on Search or AI Overviews, no major provider has committed to reading it, and 97% of such files received zero traffic in May 2026. Cheap housekeeping, not a ranking factor.

Content

Writing and freshness

Answer-first structure
Stating the question as a heading and answering it in the opening lines before elaborating. Supported by the finding that over half of sampled citations came from the first 30% of the cited page.
Freshness
How recently a page was substantively updated. A strong filter on commercial queries: 83% of citations came from pages updated within twelve months.
Authority signals
Statistics, citations and attributable quotations. The interventions that measurably outperformed cosmetic edits in the KDD 2024 research.
Keyword stuffing
Repeating target terms to signal relevance. Measured as counterproductive for citation in generated answers: one of the clearest cases where classical SEO instinct works against you.
Stale contradiction
An old page asserting a superseded fact that keeps resurfacing because nothing formally retired it. A common and fixable cause of engines describing a company wrongly.

Sources

Where these numbers come from

Every figure on this page is external and linked. The studies measure different populations over different windows, so read them as converging evidence rather than one authoritative number.

  1. Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan & Deshpande, ACM SIGKDD 2024Targeted content changes raised visibility in generative engine responses by up to 40%. arxiv.org ↗
  2. Omnibound, AEO statistics compilation 202655% of sampled AI Overview citations came from the first 30% of the cited page. www.omnibound.ai ↗
  3. Omnibound, AEO statistics compilation 2026For commercial and evaluation-stage queries, 83% of AI citations came from pages updated within the previous 12 months, and over 60% from pages refreshed within six months. www.omnibound.ai ↗
  4. Google Search RelationsGoogle documentation updated in June 2026 states llms.txt has no effect on Search rankings or AI Overviews. John Mueller noted that no AI crawler has claimed it extracts information from the file. baselinelabs.ai ↗
  5. SE Ranking and Ahrefs studies, reported by Digital AppliedA study of 300,000 domains found llms.txt on 10.13% of them; a separate study of 137,000 sites found 97% of llms.txt files received zero traffic in May 2026. www.digitalapplied.com ↗
  6. Search Engine LandGoogle zero-click searches reached 68% in early 2026. searchengineland.com ↗
  7. Pew Research CenterWhen an AI Overview is shown, users click one of its cited source links about 1% of the time. searchengineland.com ↗

Keep reading

Explainer

What is AEO

 Comparison

GEO vs AEO vs SEO

 Data

AI search statistics 2026

 Playbook

What actually works

FAQ

Common questions

Can I reuse these definitions?

Yes, with attribution and a link. They are written to be quoted, self-contained, without needing surrounding context to make sense.

Why are some entries marked contested?

Because the category is young and a good deal of its vocabulary is sold with more confidence than the evidence supports. A glossary presenting contested claims as settled facts is a marketing asset pretending to be a reference.

Want help getting into AI answers? <span class='tone'>Talk to Lifewood Data Technology.</span>

Lifewood Data Technology runs AEO and GEO programmes for enterprise brands. We measure where you stand first, then fix what is blocking you.

  • 100+ languages
  • 50+ delivery centres
  • Measured monthly
Visit Lifewood.com See GEO services

Lifewood Data Technology publishes this reference.