Live data on how ChatGPT, Claude, Gemini and Perplexity discover, cite and recommend brands across the web.
BrandGhost continuously analyzes AI responses, citations and recommendations to understand how the major AI engines see the web — and how differently they arrive at their answers.
AI engines don't see the same web
Even when answering similar questions, major AI engines frequently rely on different parts of the web. This compares how often two engines cite the same domains for the same prompts.
Across shared questions, the major engines overlap on only a small fraction of cited domains — a reminder that AI visibility is fragmented across the web, not concentrated in one ranking.
Overlap is measured as the average share of shared cited domains between each engine pair across comparable prompts. Gemini is excluded from this comparison because its current citation representation is not directly comparable.
How old is the content AI cites?
The distribution of cited-page age across mutually exclusive buckets, for citations where a reliable publication date could be detected.
More than 44.2% of dated citations in the current Observatory run point to content more than one year old. Established, durable content still earns AI citations.
Median cited age is 308 days. Only citations with a detectable page date are included (848 citations).
Citation freshness by AI engine
Median age of cited pages with detectable publication dates, by engine. Engines without sufficient date coverage are omitted.
Median age of cited pages with detectable publication dates.
Where AI citations come from
Broad, identifiable source categories from the latest run. Source classification is still expanding, so a large share of citations remains unclassified.
Reddit accounts for 4.7% of all citations in the latest run among sources we can currently classify.
Source classification is an expanding capability. 91.4% of citations in this run are not yet assigned to an identifiable category and are excluded from the bars above rather than shown as a dominant "other" bucket.
How providers differ
The identifiable source mix varies by engine. Where an engine's citations can't yet be reliably classified, we say so rather than presenting misleading comparisons.
Top public citation sources
A limited leaderboard of the most-cited recognizable domains in the latest run. This is a benchmark sample, not the full citation database.
| Domain | Citations | Share of run | Engines |
|---|---|---|---|
| reddit.com | 118 | 4.7% | 2 |
| consumerreports.org | 33 | 1.3% | 3 |
| youtube.com | 24 | 1% | 1 |
| forbes.com | 20 | 0.8% | 3 |
| alibaba.com | 17 | 0.7% | 2 |
| yahoo.com | 15 | 0.6% | 3 |
| laptestpro.com | 14 | 0.6% | 3 |
| edmunds.com | 13 | 0.5% | 3 |
| easybear-appliancerepair.com | 13 | 0.5% | 2 |
| rtings.com | 12 | 0.5% | 3 |
| paristourism.org | 11 | 0.4% | 2 |
| nytimes.com | 11 | 0.4% | 1 |
| usnews.com | 10 | 0.4% | 2 |
| timeout.com | 9 | 0.4% | 3 |
| santorinidave.com | 9 | 0.4% | 3 |
| restaurantsforkings.com | 9 | 0.4% | 1 |
| tripadvisor.com | 8 | 0.3% | 2 |
| cnet.com | 8 | 0.3% | 1 |
| bgr.com | 8 | 0.3% | 2 |
| tomsguide.com | 7 | 0.3% | 2 |
AI discovery is a representation problem, not just a ranking problem
The Observatory data suggests that AI engines frequently rely on different domains and can recommend different brands for similar questions.
For businesses, visibility increasingly depends on how consistently their products, expertise and reputation are represented across the web — not simply whether a single page ranks in traditional search.
How does AI see your brand?
BrandGhost Launchpad analyzes how your company is represented across search and AI discovery, identifies visibility gaps, and turns those gaps into an actionable content strategy.
About the data
The BrandGhost AI Discovery Observatory runs a cross-section of consumer and business discovery questions through major AI providers and analyzes the resulting recommendations and cited sources.
Metrics shown on this page reflect the latest completed Observatory run.
Publication-age metrics only include pages where a reliable date could be identified.
Source and content classifications may have different coverage levels. Metrics with incomplete classification are explicitly labeled.
AI systems, models and retrieval behavior change frequently, so Observatory statistics should be treated as a snapshot of the current AI discovery environment.