Why This Glossary Exists
The AI visibility category generates new jargon faster than any single vendor can keep up with. Profound, AthenaHQ, Otterly, Peec, Goodie, Writesonic, and dozens of smaller tools all use overlapping but slightly different terms. This page is DataDab's working definition — the one we use internally, the one we use in client briefs, the one we ship to your team when you engage us.
If you need to win AI citations in 2026, the vocabulary you actually need is small: answer engine (the system you're optimizing for), citation (the unit of outcome), prompt set (the input you measure against), extractability (the property of your pages that determines whether the engine can use them), entity (the reconciled concept of your brand), and decision-stage content (the content format that earns citations). The rest of this page defines each, plus the terms that show up in vendor pitches so you do not get quoted a feature you do not need.
The Engines
The systems you are optimizing for. The vocabulary differs across engines because the retrieval architecture differs.
Answer engine
An answer engine is an AI system that responds to natural-language queries with synthesized, conversational answers rather than lists of links. The major answer engines in 2026 are ChatGPT (OpenAI), Perplexity, Gemini (Google), Claude (Anthropic), Google AI Overviews, Microsoft Copilot, and (in commerce specifically) ChatGPT Shopping and Google's AI-organized search surfaces. They overlap with traditional search but are not the same: ChatGPT and Claude primarily synthesize from training data; Perplexity and Google AI Overviews are retrieval-first and cite source URLs.
Why this matters for SaaS: your buyer-intent prompts have different answers depending on which engine the buyer uses. Cross-engine visibility is the new cross-platform visibility.
Retrieval-augmented generation (RAG)
RAG is the architecture pattern that most answer engines use to ground their answers in current web content: the model first retrieves relevant pages from an index, then generates a response that cites those pages. Perplexity, Google AI Overviews, and ChatGPT's web-search-enabled mode all use RAG; ChatGPT's default training-only mode and Claude's standard chat do not. RAG explains why some answers cite your page directly while others paraphrase you without a link.
Why this matters for SaaS: RAG-friendly engines (Perplexity, Google AI Overviews) cite your published pages in near-real time. Training-only engines (default ChatGPT, Claude) cite based on what they learned during training — your pages can influence that only via ongoing publication.
LLM grounding
Grounding is the process of attaching an answer to verifiable external sources so the model is not relying on memorised training data alone. Perplexity grounds by retrieving and citing; Google AI Overviews ground by attaching inline sources; ChatGPT and Claude ground selectively — typically only on web-search-enabled or tool-use-enabled responses. The technical term matters less than the consequence: a well-grounded answer cites you; a poorly-grounded one paraphrases training-data presence, which may or may not include your brand.
Why this matters for SaaS: grounding is what makes a citation durable. If your brand first appears in well-grounded answers (Perplexity, Google AI Overviews), it carries through to less-grounded engines via training-data updates.
Passage retrieval
Passage retrieval is how modern answer engines find the specific sentence or paragraph on your page that supports the answer. Rather than ranking whole pages, the engine ranks individual passages (typically 100–300 word windows) and returns the one most aligned with the query. The implication for AEO is large: a single buried sentence on a high-authority page can earn a citation even if the page's overall topic is broader. Lead-with-an-answer pages give passage retrievers something concrete to find.
Why this matters for SaaS: rewrite pages so the most citation-worthy sentence is the first sentence, not buried in section three.
The Optimization Methods
What marketers actually do. These terms describe the disciplines inside the AEO category.
AEO (Answer Engine Optimization)
AEO is the discipline of making your pages, content, and brand citation-worthy inside answer engines. In practice, AEO means the same things as AI visibility or GEO, with one nuance: AEO implies optimizing for the answer itself, not just the source citation. Profound, AthenaHQ, Otterly AI, Peec AI, and Goodie AI are the most-cited tools in the AEO category.
Why this matters for SaaS: when you brief your leadership or write a budget request, 'AEO' is the most recognised category term in 2026 and the cleanest way to anchor the conversation.
GEO (Generative Engine Optimization)
GEO is the term most prominently used by Writesonic and several academic papers to describe the same practice as AEO — getting cited inside generative AI answers. The two terms are largely interchangeable in 2026, with AEO more popular in venture-funded tooling and GEO more popular in academic and SEO-platform circles. The tension is mostly branding: a tool that calls its product 'AEO software' and a tool that calls its product 'GEO software' are doing the same thing.
Why this matters for SaaS: if a vendor pitches 'GEO software,' do not pay more for it than for an 'AEO platform.' Same product class.
AI extractability
AI extractability is the degree to which an answer engine can pull structured, citable facts off your page when forming a response. The term is DataDab's: an extractable page leads with a single-paragraph answer, defines key terms in the first 200 words, uses comparison tables and FAQ blocks for the substantive claims, and uses Schema.org markup so engines see the structure before they read the prose. The AI Extractability Audit scores this on six dimensions.
Why this matters for SaaS: extractability is the closest thing the category has to a 'page-level score you can move.' Optimize for it explicitly.
Decision-stage content
Decision-stage content is DataDab's term for content written specifically for buyers who are actively evaluating solutions, not buyers at the top of the funnel. Decision-stage content converts 5–10x higher than top-of-funnel content because the buyer's intent is already purchase-shaped: comparison pages, ROI calculators, decision-friction audits, pricing explainers, migration guides. For AEO, decision-stage content is also more citation-worthy because the prompts that produce it ('best [category] for [use case]') are exactly the buyer-intent prompts answer engines cite from.
Why this matters for SaaS: the same content that wins conversions at the bottom of your funnel also wins citations inside answer engines. Optimise once.
The Measurement Vocabulary
The metrics. Be careful here: each tool measures these slightly differently. Use the right metric for the decision you are making.
AI citation
An AI citation is a single instance where an answer engine names your brand, your product, your page, or your research inside an AI-generated response to a buyer query. Citations come in three forms: named mentions (ChatGPT says 'Monday.com is a top work OS'), inline links (Perplexity and Google AI Overviews show your URL inline), and source panels (Perplexity's right-rail citation column). AI citation share is the share of relevant buyer-intent prompts that produce any of these for your brand.
Why this matters for SaaS: this is the unit of outcome. Everything else in this category is either an input (prompt set, content) or a correlation (engines, ranking).
Citation share
Citation share is the share of relevant buyer-intent prompts that produce any AI citation (mention, link, or source panel) for your brand. A 12% citation share in your prompt set means 12 of 100 buyer-intent prompts name you in the AI response. The DataDab SaaS AI Citation Index defines this as a composite of brand mention, prominence, sentiment, specificity, link/citation, and competitive position. Tools define it differently (Profound uses position-weighted citation share, Otterly uses mentions per prompt) but the headline concept is consistent across the category.
Why this matters for SaaS: ship a quarterly report to leadership that tracks citation share the way SEO programs track organic share of voice. Same conversation, different surface.
Brand mention
A brand mention is the simplest AI visibility metric: whether the AI names your brand at all in its response. It is necessary but not sufficient. Most well-known B2B SaaS brands score 8/10 on mention (the AI knows who they are) but score far lower on specificity (the AI cannot describe what they do, how they differ, or who they are for). The DataDab SaaS AI Citation Index weights mention at 25% and the other dimensions at 75% combined.
Why this matters for SaaS: if a vendor pitches 'mention rate' as the headline metric, you are talking to someone who has not read the data.
Mention prominence
Mention prominence is where in the AI's response your brand appears — first, second, or fifth in a recommendation list; named alongside competitors or after them. The first named brand in an AI recommendation typically receives a disproportionate share of the buyer's downstream attention; the third or fourth named brand may as well not be there. Mention prominence is why a comparison page that puts your product first in a structured 'best for X' table tends to be more cited than one that buries it in a feature checklist.
Why this matters for SaaS: every comparison page you publish should have a ranked recommendation table, not a feature matrix.
Brand representation
Brand representation is how accurately the AI describes your brand, beyond whether it names you at all. Representation is a separate problem from citation: the AI may consistently mention you but consistently describe you as 'an enterprise tool for large retailers' when you are actually a mid-market tool for SaaS. AthenaHQ's diagnostic depth is largely about measuring and correcting brand representation. The most common representation failures are wrong category, wrong audience, wrong use case, and stale feature claims.
Why this matters for SaaS: run a quarterly representation audit. The output is usually a handful of well-written third-party pages that re-categorise your brand — Wikipedia, Wikidata, G2, industry analyst reports.
Prompt volume
Prompt volume is the count of times a given question (or near-synonym) is asked across answer engines in a given month. A high prompt volume means a citation opportunity surfaces thousands of times; a low prompt volume means it surfaces dozens. Profound and Otterly publish prompt volume as a primary ranking signal — the prompts worth chasing are those asked often by your buyers, not those asked rarely by academics.
Why this matters for SaaS: optimise for the top 10 by prompt volume in your category. The long tail is not worth a publishing cycle.
Prompt research
Prompt research is the practice of building the set of buyer-intent questions that you want your brand to appear in answers to. The output of prompt research is a prompt set: a list of 10–100 questions, grouped by intent and stage, used as the input to AI visibility measurement. Profound's 'Prompt Research Reports' and Peec AI's prompt engine are the most visible tools in this category. Good prompt sets represent the actual questions your buyers ask, not the questions you wish they asked.
Why this matters for SaaS: your prompt set becomes your content roadmap. Every prompt in the top 10 is a page to write or refresh.
The Strategy Vocabulary
Concepts that show up in every vendor pitch and most leadership briefings. Use them deliberately.
Topical authority
Topical authority is the cumulative credibility an answer engine assigns to a domain on a specific topic, built through consistent publication depth over months and years. Sites that publish a single category-defining resource tend to be cited for queries in that category; sites that publish across many topics are cited less reliably. The strongest signal of topical authority is being cited as the reference source for definitional or category-wide queries, not being mentioned as 'one of many.'
Why this matters for SaaS: pick the category you can win (one specific ICP, one specific use case) and write 30 pages inside it before chasing adjacencies.
Entity
In AI search, an entity is a uniquely identified thing — a company, person, product, or concept — that an answer engine has reconciled across the web. Entity reconciliation is the process by which an AI determines that 'DataDab' on one page, 'DataDab, Inc.' on another, and '@datadab' on GitHub all refer to the same company. Strong entities have consistent schema markup (Organization + sameAs), a Wikipedia or Wikidata entry, and corroborating third-party profiles (LinkedIn, Crunchbase, G2). Weak entities are ambiguous — the AI hedges with 'a company called DataDab...' rather than naming you directly.
Why this matters for SaaS: ship Organization JSON-LD with sameAs links to every profile you control. Submit a Wikidata entry if you meet notability. This is the cheapest unseen AEO work you can do.
Structured data for AI visibility
Structured data is the machine-readable layer — typically Schema.org JSON-LD — that tells answer engines what your page is about, who published it, what organization is behind it, and what its key claims are. The most useful schema types for AI visibility are Organization (with founder, contactPoint, sameAs), Article (with author, datePublished, dateModified), FAQPage (each question as an extractable Q-A pair), Product (with offers), and BreadcrumbList. Pages with structured data extract more reliably than pages without it because the engine sees the structure before reading the prose.
Why this matters for SaaS: of every AI-visibility lever, structured data has the best effort-to-impact ratio. Most B2B SaaS sites are missing 60%+ of the schema their pages should have.
How To Use This Glossary
A short orientation on how to navigate from here based on the conversation you are having.
If You Are Evaluating A Vendor Pitch
Ignore the tool's own branding terms (AEO software, GEO platform, AI visibility suite) and read the methodology. Ask which citations they measure, which passage retrieval assumptions they make, and how they handle brand representation versus brand mention. The comparison page covers all of this for the major tools.
If You Are Briefing Leadership
Use AEO as the category term (most recognised in boardrooms in 2026), citation share as the headline metric (analogue to organic share of voice), and decision-stage content as the differentiation angle that explains why DataDab-style implementation beats vendor-only measurement.
If You Are Writing The Pages That Get Cited
Lead with the answer (one paragraph, one answer — gives passage retrievers something concrete). Use comparison tables and FAQ blocks (the most extractable structures). Add the right structured data. Refresh dated pages (mentioned in our how SaaS companies get cited guide).
If You Are Buying A Tool
Three questions worth asking every vendor: which prompt volumes are you indexing against, how do you score mention prominence rather than just mention, and how does your implementation arm work (or do you ship nothing and leave the rewrite to me)?
Sources & Verification
Where the term is a vendor's coinage, we have linked directly to the vendor's own materials. Where the term is widely used in the academic or SEO-platform literature, we cite the originating paper or platform post. Where the term is DataDab's — extractability and decision-stage content — we mark it as such and link to the work that grounds it.
- Profound research blog — primary research on prompt language, citation mechanics, "Parrot Problem," and AEO-category definitions.
- AthenaHQ — vendor source for brand-representation framing.
- Otterly AI — vendor source for AI Search rank-tracker framing.
- Peec AI — vendor source for prompt-engine framing.
- Writesonic blog — primary academic and SEO-platform source for the 'GEO' term.
- DataDab SaaS AI Citation Index — DataDab's composite citation-share methodology.
This glossary is re-verified quarterly alongside the SaaS AI Citation Index refresh. New terms get added when a major vendor ships a category-defining feature (e.g. Profound's 'Prompt Research Reports' is the most recent example). Deprecated terms get flagged but kept for historical cross-reference.