Plain-English definitions for terms that show up across Cited Brands’ research, blog and Index Report. Includes AEO methodology, statistical concepts, SA publishers, and the named bench mechanics. 35 terms.
Answer Engine Optimisation — the discipline of making sure AI assistants (ChatGPT, Claude, Gemini, etc.) name, cite, and describe your brand accurately when users ask questions about your category.
The core practice SA-AEO-Bench measures.
What is AEO? — full definition post →Google's AI-generated summary that appears above the standard search results on many queries. AI Overviews now appear on around 89% of branded searches and contribute to the zero-click shift.
Not measured directly by SA-AEO-Bench v1 — bench measures the three frontier LLMs, not Google AI Overview specifically.
A user-facing AI system that returns a single composed answer rather than a list of links. ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews are all answer engines.
Whether an AI engine names your brand unprompted, with no brand names planted in the question. Measured by asking the model 'which medical aid should I pick' rather than 'how does Discovery compare to Bonitas'.
The headline metric per brand in SA-AEO-Bench v1.
A statistical model for estimating item strengths from pairwise comparisons. Given a series of "A vs B" preference outcomes, it produces a single strength score per item that ranks them.
Used in SA-AEO-Bench to convert per-prompt A/B citation wins into a per-brand strength score per LLM.
The most-cited SA publisher by AI engines in SA-AEO-Bench v1, with thousands more citations than the next-largest SA source. National business press. A near-mandatory placement target for any SA-focused AEO programme.
Where SA brands need to get cited — the publisher list →A URL referenced or linked by an AI engine's answer. Each citation maps to a source domain (e.g. businesstech.co.za) and gets classified into one of 12 source categories during SA-AEO-Bench analysis.
The percentage of citations attributed to a given domain or category, out of total citations for a brand or industry. The flagship measure of 'how often does AI read X when describing Y'.
A consumer-reviews site whose primary content is grievances about brands. Hello Peter is the dominant SA example. AI engines cite complaint platforms heavily, which is why complaint exposure is one of the three numbers in the 90-minute audit.
A schema.org JSON-LD shape for declaring a glossary or term list. Search engines and AI engines use it to surface definitions in featured snippets and answer responses. This page emits one.
The current generation of leading-edge large language models — currently OpenAI GPT-5, Anthropic Claude Sonnet 4.5, and Google Gemini 2.5 Pro. SA-AEO-Bench measures all three on every prompt.
Generative Engine Optimisation — a synonym for AEO. The term emerged in 2024 academic papers; AEO is the more common practitioner label. Same field, different name.
AEO vs GEO — the terminology question →The seven pre-registered hypotheses for SA-AEO-Bench v1. Filed on OSF (osf.io/w4az2) before any LLM was queried. Each has a numeric threshold and a confirmation/rejection verdict tracked in the pre-completion report.
All H1–H7 verdicts in the research docs →South Africa's dominant consumer complaint platform. The only complaint platform in the top 10 SA citation sources in SA-AEO-Bench v1. Brand-level Hello Peter ratings correlate strongly with how defensively AI describes the brand.
The Hello Peter threshold finding →Cited Brands' quarterly synthesis of SA-AEO-Bench data — per-brand findings, industry deep-dives, leaderboards, and the cost-of-inaction frame for CFO conversations. Gated download.
Request the Q2 2026 Index Report →Aggregate SA-source-share per industry in the bench. Top: short-term insurance at 72%. Bottom: restaurants at 30%. The gap is wider than the disagreement between any two LLMs on the same brand.
A measure of overlap between two sets, calculated as the size of the intersection divided by the size of the union. Used in SA-AEO-Bench H3 to compare citation sets across Latin Square position swaps.
An experimental-design technique for cancelling out position bias in comparison queries. When asking an LLM "is A better than B", the answer can flip depending on which name appears first. Latin Square counterbalancing runs all orderings and aggregates, isolating preference from position.
Applied to every A-vs-B comparison query in SA-AEO-Bench v1.
Large language model. The class of AI systems that power answer engines — GPT-5, Claude, Gemini, Llama, Mistral, etc. SA-AEO-Bench v1 measures three frontier LLMs from three vendors.
A third synonym for AEO/GEO, popularised by SEO agencies repositioning their service deck. Same practice as AEO; the naming difference is positioning, not substance.
The public protocol describing how a bench was run — sample selection, prompt set, scoring rubric, statistical analysis plan, budget ceiling. SA-AEO-Bench's was filed on OSF before the run started.
Read the SA-AEO-Bench v1 protocol →The degree to which a brand's owned messaging shapes how AI describes it, vs how much of the description comes from complaints, third-party reviews, or news coverage the brand did not commission.
A free academic platform for pre-registering studies. SA-AEO-Bench v1 was pre-registered at osf.io/w4az2 before any LLM was queried — meaning the hypotheses, methods, and analysis plan were locked publicly first.
View the SA-AEO-Bench pre-registration on OSF →The practice of publishing a study's hypotheses, methods, and analysis plan before collecting data. Prevents p-hacking and HARKing. SA-AEO-Bench v1 is pre-registered at osf.io/w4az2.
How a brand's ranking by AI shifts when the prompt is qualified — adding "affordable", "premium", or "trustworthy" can swing the named-leader by 30 points. Most brands have inherited a prompt-conditional narrative they didn't mean to.
The short list of domains AI engines actually read when composing answers about SA brands — currently dominated by businesstech, Hello Peter, news24, mybroadband, techcentral, rateweb, and a long tail. Earning placement on the top six is the most leveraged AEO PR move.
The pre-registered SA-AEO-Bench finding that Gemini cites Reddit hundreds of times on SA queries while GPT-5 and Claude cite it zero times. One of the largest cross-model citation disagreements in the dataset.
The pre-registered benchmark of how frontier LLMs cite South African brands. Run by Cited Brands. v1 covered 100 brands across 10 industries with ~189,000 classified citations. Re-run quarterly.
All SA-AEO-Bench v1 documents →Cited Brands' free brand-level AEO snapshot. Returns blind visibility, complaint-platform exposure, and top three citation sources for any SA brand in under 60 seconds.
Get your free Scorecard →The taxonomy SA-AEO-Bench uses to bucket citations by domain type: sa_local_press, sa_review, sa_directory, sa_gov, sa_edu, sa_other, wikipedia, reddit, youtube, social, intl_review, intl_press, intl_other. Twelve mutually-exclusive categories.
The TypeScript constant in @citedbrands/brand-data that holds the canonical SA-AEO-Bench v1 stats: total citations, brand count, industry count, completion percentage, and the snapshot date. Single source of truth for every number shown on the marketing site.
A methodological adjustment for the tendency of LLMs to agree with the framing of a question. If you ask "isn't X the best?" an LLM is more likely to say yes than if you ask "is X the best?". SA-AEO-Bench applies counterbalanced phrasing to neutralise this.
An editorial post format that ranks brands in a category ('Best banks in South Africa, ranked by AI'). High consumer search volume, ranks well, gets AI-cited on the corresponding query, builds the authority that helps every other post.
A public URL that lets anyone confirm a cert was actually issued by Cited Brands. For example /verify/cb-aeo-prac-2026-x7q9. Anyone with the link can view the holder name, course, and issue date. Used on the Cited Brands AEO Certified badge.
A search that ends without a click to any result. Around 60% of Google searches now end zero-click thanks to AI Overviews and featured snippets. The shift is the macro driver of the AEO field.