What does strong AI presence mean?

Strong means the brand appears as a primary recommendation in more than 40% of relevant prompts on a given engine. Moderate means 15-40%. Weak means under 15%. Absent means the brand was not surfaced as a recommendation in our prompt set.

What is citation share?

Citation share is the percentage of all brand mentions across the full prompt set attributed to a given brand. It is an aggregate measure across all engines and all prompts. We always report it alongside per-engine breakdowns because the aggregate can hide dramatic engine-specific variation.

Which AI engines do you test?

We test ChatGPT (GPT-4o), Perplexity, Google AI Mode, Gemini Pro, and Claude. All prompts are run in fresh, anonymous sessions without signed-in accounts, prior search history, or location signals.

How often are studies updated?

We re-run a baseline subset of 20 prompts from each category study each quarter and publish changes as updates to the original study page. New category studies publish every two to four weeks.

What drives AI citation share?

The factors most correlated with strong AI citation share are: breadth of independent third-party mentions, volume and quality of practitioner-facing narrative content, and presence in training-data-adjacent sources such as analyst reports, academic papers, and respected trade publications. Self-published content contributes much less than an equal volume of independent, attributable third-party content.

Answer Engine Index Methodology

GEO Times

Stay ahead of GEO

Industry updates on Generative Engine Optimisation — straight to your inbox.

Subscribe on LinkedIn

GEO Times

GEO industry updates — straight to your inbox.

Subscribe on LinkedIn

GEOby

lilbigthings.com

Answer Engine Index/Methodology

The Answer Engine Index

How we measure who AI recommends.

Every study in the Index uses the same method. This page is the reference the studies link to.

We take the non-branded questions buyers actually ask, run them across six AI assistants repeatedly over several weeks, and count how often each brand is named. We inspect individual answers to see where each mention comes from, live web search or trained memory, and we record which sources the engines cite.

How each study runs

Define the category. We take a real category and its competitive set of tracked brands.

Write the questions. Thirty non-branded questions across the real buying themes and buyer personas, no brand names, from first exploration to shortlist.

Run them across six engines. ChatGPT, Claude, Gemini, Google AI Overviews, Copilot, and Perplexity, repeatedly, over the study window.

Count presence. Presence rate is the share of answers to those questions that name each brand.

Inspect the answers. We read individual responses to trace cited sources and to separate memory-based from retrieval-based mentions.

Publish in full. The leaderboard, per-engine and per-theme breakdowns, per-brand pages, and this method, all open.

Definitions

Presence rate

Share of answers to non-branded category questions that name a brand. Volume-weighted across the measured weeks.

Buying stage

Classification of each question as Awareness or Consideration based on the buyer's intent.

Citation ownership

The share of cited URLs owned by a brand, its competitors, or third parties.

Memory vs retrieval

Whether an engine answered from trained knowledge or from a live web search, judged from the presence or absence of citations in the answer.

The two doors

Memory (what the model recalls) and retrieval (what the model finds and cites). A brand can win one, both, or neither.

What the numbers are, and are not

These are directional estimates from third-party measurement, not audited results.
AI answers vary run to run. Presence is a sampled rate, not a census.
Leaderboards reflect the brands tracked in each category. Assistants also name untracked tools, so the real field is often wider than the ranking shows.
Per-brand shares reflect a category's question set and window, and should not be read as overall market share.
Theme matrices are a directional analyst read across topics, not a separately computed per-theme rate.

Who runs this

The Answer Engine Index is produced by Lil Big Things, a B2B marketing firm focused on AI search visibility. Measurement is powered by Scrunch AI, with category and organic context from Ahrefs and web research from Exa.

Want your category measured this way?

Talk to us Back to the Index

The Answer Engine Index, an ongoing study by Lil Big Things. This page is the canonical methodology reference for all Answer Engine Index studies.

Get each new study when it publishes.