The numbers behind the benchmark
1,176,321 AI answers, 7,984,674 classified citations, 2025-08-18 to 2026-11-06. Pooled and anonymised, in files you can open in a spreadsheet or read with code.
Summary tables
Flat extracts of the most-used figures. Groups under 200 answers are left out, and a market is listed on its own only when its data draws on studies from at least 2 organisations, as on the market pages.
AI engine profiles (engines.csv)
One row per AI engine or model: answers, source use, source mix, platform reach, brand visibility and tone.
| Column | Meaning |
|---|---|
engine | AI engine or model |
answers | AI answers in the group |
pct_answers_citing_sources | Share of answers that cite at least one web source (%) |
sources_per_answer | Average distinct domains per answer that cites sources |
pct_citations_brand_and_corporate | Share of cited domains that are brand & corporate (%) |
pct_citations_earned_media | Share of cited domains that are earned media (%) |
pct_citations_social_and_ugc | Share of cited domains that are social & ugc (%) |
pct_citations_reference | Share of cited domains that are reference (%) |
pct_citations_government_and_public_sector | Share of cited domains that are government & public sector (%) |
pct_citations_academic_and_research | Share of cited domains that are academic & research (%) |
pct_citations_institutions_and_associations | Share of cited domains that are institutions & associations (%) |
pct_citations_marketplaces_and_aggregators | Share of cited domains that are marketplaces & aggregators (%) |
pct_citations_other | Share of cited domains that are other (%) |
pct_answers_citing_reddit | Share of source-using answers that retrieve or cite Reddit (%) |
pct_answers_citing_youtube | Share of source-using answers that retrieve or cite YouTube (%) |
pct_answers_citing_wikipedia | Share of source-using answers that retrieve or cite Wikipedia (%) |
pct_answers_citing_linkedin | Share of source-using answers that retrieve or cite LinkedIn (%) |
pct_unbranded_focus_brand_named | Unbranded questions: share of answers that name the focus brand (%) |
pct_unbranded_own_site_cited | Unbranded questions: share of source-using answers that cite the focus brand’s own site (%) |
pct_negative_framing | Share of scored answers that frame the brands they name negatively (%) |
Industry profiles (industries.csv)
One row per industry, with the same measures as the engine profiles.
| Column | Meaning |
|---|---|
industry | Industry |
answers | AI answers in the group |
pct_answers_citing_sources | Share of answers that cite at least one web source (%) |
sources_per_answer | Average distinct domains per answer that cites sources |
pct_citations_brand_and_corporate | Share of cited domains that are brand & corporate (%) |
pct_citations_earned_media | Share of cited domains that are earned media (%) |
pct_citations_social_and_ugc | Share of cited domains that are social & ugc (%) |
pct_citations_reference | Share of cited domains that are reference (%) |
pct_citations_government_and_public_sector | Share of cited domains that are government & public sector (%) |
pct_citations_academic_and_research | Share of cited domains that are academic & research (%) |
pct_citations_institutions_and_associations | Share of cited domains that are institutions & associations (%) |
pct_citations_marketplaces_and_aggregators | Share of cited domains that are marketplaces & aggregators (%) |
pct_citations_other | Share of cited domains that are other (%) |
pct_answers_citing_reddit | Share of source-using answers that retrieve or cite Reddit (%) |
pct_answers_citing_youtube | Share of source-using answers that retrieve or cite YouTube (%) |
pct_answers_citing_wikipedia | Share of source-using answers that retrieve or cite Wikipedia (%) |
pct_answers_citing_linkedin | Share of source-using answers that retrieve or cite LinkedIn (%) |
pct_unbranded_focus_brand_named | Unbranded questions: share of answers that name the focus brand (%) |
pct_unbranded_own_site_cited | Unbranded questions: share of source-using answers that cite the focus brand’s own site (%) |
pct_negative_framing | Share of scored answers that frame the brands they name negatively (%) |
Market profiles (markets.csv)
One row per market whose data draws on studies from at least two organisations, with the same measures.
| Column | Meaning |
|---|---|
market_code | ISO 3166-1 alpha-2 country code |
market | Market |
answers | AI answers in the group |
pct_answers_citing_sources | Share of answers that cite at least one web source (%) |
sources_per_answer | Average distinct domains per answer that cites sources |
pct_citations_brand_and_corporate | Share of cited domains that are brand & corporate (%) |
pct_citations_earned_media | Share of cited domains that are earned media (%) |
pct_citations_social_and_ugc | Share of cited domains that are social & ugc (%) |
pct_citations_reference | Share of cited domains that are reference (%) |
pct_citations_government_and_public_sector | Share of cited domains that are government & public sector (%) |
pct_citations_academic_and_research | Share of cited domains that are academic & research (%) |
pct_citations_institutions_and_associations | Share of cited domains that are institutions & associations (%) |
pct_citations_marketplaces_and_aggregators | Share of cited domains that are marketplaces & aggregators (%) |
pct_citations_other | Share of cited domains that are other (%) |
pct_answers_citing_reddit | Share of source-using answers that retrieve or cite Reddit (%) |
pct_answers_citing_youtube | Share of source-using answers that retrieve or cite YouTube (%) |
pct_answers_citing_wikipedia | Share of source-using answers that retrieve or cite Wikipedia (%) |
pct_answers_citing_linkedin | Share of source-using answers that retrieve or cite LinkedIn (%) |
pct_unbranded_focus_brand_named | Unbranded questions: share of answers that name the focus brand (%) |
pct_unbranded_own_site_cited | Unbranded questions: share of source-using answers that cite the focus brand’s own site (%) |
pct_negative_framing | Share of scored answers that frame the brands they name negatively (%) |
Monthly trends (fixed panel) (trends.csv)
Monthly web search, shopping, ad and platform rates on a fixed panel of long-running studies.
| Column | Meaning |
|---|---|
measure | web_search, shopping, ads or platforms |
series | AI engine, or platform for the platforms measure |
month | Calendar month (YYYY-MM); the latest month may be partial |
pct | Share of answers (%) |
Headline findings (findings.json)
The editorial findings published on the site (headline, branded, engines and industry insights) with their key statistics, plus dataset totals.
Source files
The JSON files the site itself runs on. Dimension values in the cube and the site lists are indexes into the label lists in meta.json.
- meta.json: Dataset totals and the labels for every dimension (industries, engines, months, markets, languages, question themes, source types, platforms) and the suppression thresholds.
- cube.json: Columnar cube: industry × engine × month × market × language × theme × branded, with additive measures. "dims" and "measures" name the columns; dimension values are indexes into the lists in meta.json, and -1 means not set.
- domains.json: Pre-aggregated most-cited domain lists. Each slice key is [industry, engine, market, theme, branded] (-1 = all); items are [domain index, answers citing], against n answers with sources.
- trends.json: Fixed-panel monthly series and model-version profiles, as charted on the Trends page.
Definitions and method
Source types, question themes and every measure are defined in the glossary. How the answers were collected, classified and anonymised is on the methodology page.
Using and citing the data
The data is © We Communications, all rights reserved. You may quote individual figures with credit; see the terms of use. Cite as: GEO Compass, We Communications. GEO Compass AI search benchmark, version 2026.11.07. wegeocompass.com
How is your brand showing up in AI answers?
We benchmark how ChatGPT, Gemini, Google AI Overviews, Perplexity, Claude and Copilot describe, cite and recommend brands, then build the content and earned-media plan to change it.
Want to understand your brand opportunities and risks on AI search? Contact We AI & Data Advisors for a sample audit: AIAdvisors@wecommunications.com