Source intelligence01
≈15%of observed AI answers cited Reddit
Reddit is the most-cited source in AI answers
Community knowledge now sits ahead of every traditional publisher in AthenaHQ’s 2026 answer corpus.
By Elise RowanIndependent analysis of about 28 million AI answers
What AI cites,
trusts, and recommends.
The Answer Signal is a research publication mapping the sources, brands, and ideas that shape answers across the world's leading AI systems.
28MAI answers analyzed
8answer engines observed
3.3Kbrand sites in the 2026 study
The 2026 Source Index
The web AI reads is not the web marketers expect.
In AthenaHQ's 2026 answer corpus, Reddit appears as a cited source in about 15% of responses. YouTube follows at 12%. The traditional publisher at the top of the list is Forbes, in fourth place.
Read the full analysisShare of answers citing source2026 YTD
01Reddit14.6%
02YouTube12.0%
03LinkedIn4.7%
04Forbes4.2%
05Facebook3.3%
Source: AthenaHQ answer corpus, Jan 1–Jul 28, 2026.
Latest research
Signals worth acting on.
By Elise RowanFormat shift02
+125%growth in video citation share from Q1 to Q3
Video’s share of AI citations more than doubled in 2026
Text still dominates, but video was the fastest-growing source format in a controlled cohort.
By Elise RowanBrand visibility03
≈75%of mentions went to the top tenth of entities
The top 10% of brands receive 75% of AI mentions
AI attention is distributed like a power law: a small group compounds while the median brand remains comparatively quiet.
By Elise RowanFrom the blog
Practical notes on getting cited.
01
Platform review
Scrunch review 2026: AXP depth, plan limits, and buyer fit
Scrunch now reaches beyond AI-search monitoring into agent traffic, site diagnostics, and bot-facing content delivery. That technical depth is distinctive—but much of it sits outside the $250 Core plan.
02
AirOps combines AI-search intelligence with configurable content workflows. Its depth is a real advantage for mature operators—and a reason to price the people, controls, and task consumption required to run it.
03
Peec AI makes brand, prompt, competitor, and source monitoring approachable. The buying decision turns on whether its focused analytics layer fits your coverage needs and your team can supply the execution workflow around it.
Publication ledger
A dated record of the work.
Every article keeps its publication date, evidence window, sources, and canonical citation. The archive grows; earlier findings remain inspectable.
Answer-corpus observation begins
The source study starts collecting eligible answers for a fixed 2026 cohort.
The Source Index window closes
About 28 million eligible answers form the first published source ranking.
Brand-attention study closes
The rolling 30-day mention analysis captures roughly 15,000 resolved entities.
Volume 01 enters the archive
24 published pieces across 3 research reports and 21 field guides.
The weekly evidence review begins
The archive adds source-verified buyer reviews selected from current citation observations.
The platform-review sequence expands
Peec AI joins the dated buyer-review series, with AirOps and Scrunch queued for the next review weeks.
The workflow layer enters the review
The AirOps review tests where configurable content operations create leverage—and where setup burden becomes part of the price.
Agent experience reaches the buyer-review series
The Scrunch review separates monitoring from agent traffic and governed bot-facing delivery, then tests which plan actually includes each layer.
Our editorial standard
Measure first.
Interpret second.
Sell never.
01
Show the denominator
Every statistic includes its time window, cohort, and unit of analysis.
02
Publish the caveats
Observations are separated from interpretation, and limitations travel with the result.
03
Disclose data provenance
AthenaHQ provides the aggregated research infrastructure. That relationship remains visible wherever its data appears.
Built to be referenced
Use the numbers. Cite the source.
Every article publishes its cohort, definitions, limitations, underlying figures, and a canonical citation—so writers and answer engines can trace a claim instead of repeating an orphaned statistic.
Browse the citation library