The AI Search Field Guide

AI SEO is the broadest term we use for search optimization shaped by AI discovery. It can include visibility in generative answers, citations, recommendations and AI-search interfaces alongside the organic foundations that make web information discoverable.

At DankRank, AI Search owns the commercial service. This guide owns the reference material.

GEO stands for Generative Engine Optimization. In practice, people use it for work intended to improve visibility inside generative search and answer systems: being surfaced as a source, cited, mentioned or recommended.

Different vendors draw the boundary in different places. If somebody sells you GEO, ask what they will actually change and what they will measure.

AEO stands for Answer Engine Optimization. The term has covered direct-answer work across featured snippets, voice assistants, question-answering systems and, more recently, AI-generated answers.

AEO and GEO overlap heavily in practice. Our full terminology argument is in AEO vs GEO vs AI SEO.

AIO might mean AI Optimization, AI Overviews optimization or broader AI-search work, depending on who is using it.

I don't assume the acronym tells me the scope. I ask which part of search and which outcome the person means.

LLMO usually means Large Language Model Optimization. The label is useful when the work specifically concerns model-driven retrieval or discovery, although the evidence environment is wider than the model itself.

That environment can include the web, search indexes, first-party content, third-party sources, entities and reputation.

AI visibility is the observed presence of a brand, page, product or source inside AI search. That presence can take several forms:

citationbrand mentionrecommendationsource appearancelinked referencereferral visit

Each one tells you something different.

A citation is a source reference associated with an AI-generated answer, with mechanics that vary by product.

We record citations separately from brand mentions because a system can talk about a brand while citing somebody else. See How AI Search Citations Work for the deeper version.

A brand mention records that the brand appeared in the response. The cited source may still belong to somebody else.

Mentions can matter for category visibility and recommendation tracking, but they should be labelled honestly.

Recommendation visibility asks whether a brand is presented as an option in a prompt where a recommendation makes sense. Factual sourcing is a separate measurement question, and a strong program keeps the two apart.

Source selection is the pattern of domains or pages an AI-search product uses or cites for a prompt. We can inspect that output, repeat the observation and compare what changed.

The proprietary retrieval and ranking process remains hidden. I would rather base optimization on repeatable observations and known discovery controls than pretend there is one universal citation formula.

An entity is a distinct thing a system can identify: a company, person, product, location, organization, place or other concept.

For a business, clear entity information comes from consistent facts and relationships across first-party pages and credible external sources. Structured data can help machines interpret those facts when the markup matches the visible page and the real-world entity. Fictional or inconsistent facts don't become authoritative because they were marked up.

Structured data is machine-readable markup that describes the content or entities on a page. For search work, common types include Organization, Person, Service, Product, LocalBusiness, Article and BreadcrumbList when the page genuinely qualifies.

There is no special "AI schema" to invent. Use the schema that accurately describes the thing on the page.

A prompt is the input given to an AI system. For visibility measurement, prompt design matters because the answer can change dramatically with intent, wording, geography, context and follow-up questions.

A useful tracking set usually includes a stable core of non-brand questions tied to real discovery behaviour.

A prompt set is the defined group of prompts used for repeated measurement. Document it well enough that later results can be compared with the same questions.

Changing the set after an optimization and blending new questions into the old score can make improvement look cleaner than it really is.

AI visibility tracking means recording citations, mentions, recommendations, source presence, competitors and related observations across a defined set of AI systems so we can compare what changes over time.

The practical guide lives here:

READ HOW TO MEASURE AI SEARCH VISIBILITY→

An AI visibility score summarizes a defined observation set. It can help with trend reporting when the method stays consistent, but the evidence underneath still needs to remain visible.

For more on scoring, see How to Measure AI Search Visibility. DankRank's own methodology and launch state are published in the Launch AI Visibility Baseline.

AI companies use automated user agents for different jobs, including search discovery, user-requested retrieval and model-related data collection.

Those purposes matter because a rule for one crawler may say nothing about another product or use case from the same company. The technical reference is AI Crawlers, llms.txt & Machine Discovery.

robots.txt is a crawler-access control file at the root of a site. It tells compliant crawlers which paths they are allowed to request.

Indexing removal and non-crawler uses of information involve other controls. For AI-search work, check the actual user agent and provider documentation before adding one giant "block AI" rule.

llms.txt is a proposed convention for publishing an LLM-friendly summary and links, usually at /llms.txt.

Major AI-search products don't universally require it for visibility. I treat it as something a suitable site can test, not proof of ranking influence or an eligibility requirement.

Retrieval is the process of fetching or selecting information that may be used to answer a query. Depending on the product, that can involve live web access, search indexes, product-specific stores, licensed datasets and other systems.

Optimization gets more reliable when we work on accessible source material and measured outputs instead of claiming access to a hidden retrieval formula.

Grounding connects a generated response to external information or supplied context so the answer can use evidence beyond the model's internal parameters.

Products describe the mechanism differently. For marketers, the useful questions are which external sources the system can access and how the resulting answer exposes them.

AI referral traffic is traffic arriving at a site from an AI product where the referral can be identified in analytics. It is one of the more concrete business signals available, but it captures only visits with an identifiable click.

AI search added new surfaces. Traditional search results, Maps and paid search still influence how customers find and choose businesses.

A useful search program watches the surfaces that matter to the business and moves evidence between them when it changes a decision. That is the argument behind Search Is One Ecosystem.

For the commercial AI-search service, go to AI Search.

For the terminology argument, read AEO vs GEO vs AI SEO.

For citation mechanics, read How AI Search Citations Work.

For measurement mechanics, use How to Measure AI Search Visibility.

For crawler controls, use AI Crawlers, llms.txt & Machine Discovery.

If you want us to look at how you currently show up across search first, start with the free Visibility Baseline.