/ THE SHORT ANSWER
- 01For well-known brands, earned-source share ranged from 63.4% on Gemini to 93.5% on ChatGPT.
- 02Perplexity used 23.8% social sources in the well-known brand sample, while ChatGPT used none.
- 03For niche brands, ChatGPT cited 95.1% earned sources and 4.9% brand sources in this experiment.
- 04The study used ranking-style consumer prompts, so the mix should not be generalized directly to every B2B query.
/ dotSuper point of view
A brand needs an authority portfolio, not one preferred channel copied across every engine.
The source mix for well-known brands
The distribution below is for the paper’s well-known brand query set.
It shows why a single content channel cannot stand in for an engine strategy.
SOURCE COMPOSITION
Earned sources lead, but the engines do not agree on the rest.
Share of cited domains by source category for well-known brand prompts.Percentages describe this experiment’s cited-domain mix. They are not market-wide citation rates.
View the chart data
| Engine | Earned | Brand | Social |
|---|---|---|---|
| ChatGPT | 93.5% | 6.5% | 0% |
| Claude | 87.3% | 6.8% | 5.9% |
| Perplexity | 67.4% | 8.8% | 23.8% |
| Gemini | 63.4% | 25.1% | 11.5% |
Niche brands face an even stronger authority problem
Claude remained earned-heavy at 86.3%.
Perplexity and Gemini still included more brand and social sources, but earned content remained the largest category.
The paper also reports lower cross-engine answer agreement for niche brands, at 71% to 76%, compared with 76% to 81% for well-known brands.
Less established entities therefore face both a visibility problem and a consistency problem.
| Engine | Earned | Brand | Social |
|---|---|---|---|
| ChatGPT | 95.1% | 4.9% | 0% |
| Claude | 86.3% | 10.6% | 3.2% |
| Perplexity | 73.4% | 9.1% | 17.5% |
| Gemini | 66.4% | 21.2% | 12.7% |
The practical response: build an authority portfolio
Google explicitly warns against inauthentic mentions.
The better route is to produce evidence that credible third parties can inspect, challenge, and reference.
That portfolio can combine founder-led research, documented event learning, customer-approved case evidence, association contributions, expert interviews, and original datasets.
Owned pages remain the canonical evidence layer, while authentic external coverage supplies independent validation.
- Publish a methods note beside every benchmark or quantified claim.
- Give partners a concise evidence pack with source links and reusable charts.
- Track which external domains are actually cited for dotSuper’s target queries.
- Separate social distribution from earned validation in reporting.
What this page cannot conclude
- 01Both papers are preprints. Their findings should be treated as evidence to test, not as a settled ranking formula.
- 02The GEO engine study collected data in August 2025. Models, retrieval systems, interfaces, and citation behavior can change quickly.
- 03Observed citation patterns do not prove that changing one page element will cause an engine to cite that page.
- 04Google states that there is no special schema, file, or content format required for its generative AI features. Core SEO and useful, original content remain the foundation.
Sources
- 01Generative Engine Optimization: How to Dominate AI SearcharXiv · accessed Sep 1, 2026
- 02Optimizing your website for generative AI features on Google SearchGoogle Search Central · accessed Sep 1, 2026
Our editorial standard · Found an error? Send a correction with its source.
/ CITE OR SHARE THIS GUIDE
Make the evidence easy to verify.
When you reference this guide, link to its canonical URL. That gives readers one stable place for the evidence, limitations and future updates.
dotSuper Research Desk. (September 1, 2026). AI search source mix by engine: earned, brand, and social. dotSuper. https://dotsuper.net/feeds/search-discovery/ai-search-source-mix-by-engine