AEOForged
Get in touch

Original research · measured 2026-07-26

Reddit and YouTube Lead AI Answer Engine Citations on AEO Topics, Appearing in 35.5% and 32.7% of 110 Checks Respectively

Original research · 110 engine checks across 30 questions, measured on 2026-07-26 · Methodology · Dataset

Executive Summary

Across 110 recorded engine checks spanning 30 category questions on answer engine optimization and AI search visibility, measured on 2026-07-26 across three engines (ChatGPT, Gemini, and Perplexity), the single most striking finding is that reddit.com topped the domain leaderboard — cited in 39 of 110 checks (35.5%, 95% CI 27.1%–44.7%). A user-generated discussion platform outranking every purpose-built editorial property signals a clear pattern in how AI answer engines source responses on this topic.

Source Economy

Despite reddit.com's headline position, the broader citation economy is decisively editorial. Editorial sources accumulated 966 citations — 88% of all citations recorded. Reddit and social sources each contributed 3.6% (39 citations apiece), YouTube 3.3% (36 citations), documentation sources 1.2% (13 citations), and review sites 0.5% (5 citations). Volume-wise, editorial content dominates; reddit.com's high domain-level citation rate reflects breadth across questions rather than sheer citation count.

Top-Cited Domains

YouTube ranked second, cited in 36 of 110 checks (32.7%, 95% CI 24.7%–41.9%). LinkedIn followed at 20 of 110 checks (18.2%, 95% CI 12.1%–26.4%). Among editorial properties, semrush.com led at 18 of 110 checks (16.4%, 95% CI 10.6%–24.4%), followed by searchengineland.com and medium.com (each 15 of 110, 13.6%, 95% CI 8.4%–21.3%), blog.hubspot.com, frase.io, and hubspot.com (each 12 of 110, 10.9%, 95% CI 6.4%–18.1%), and tryprofound.com at 11 of 110 checks (10.0%, 95% CI 5.7%–17.0%).

Per-Engine Behavior

The three engines behaved markedly differently. ChatGPT ran 30 checks, but only 13 returned any sources — an average of 2 sources per answer. Its top cited domains were schema.org, support.google.com, and techradar.com, each appearing twice. Gemini ran 30 checks with 29 returning sources, averaging 11.2 sources per answer; reddit.com (9 appearances) and medium.com (8 appearances) led its citations. Perplexity ran 50 checks, all 50 returning sources, at an average of 14 sources per answer; youtube.com (32 appearances), reddit.com (30 appearances), and linkedin.com (20 appearances) were its most-cited domains. Perplexity's higher check count and 100% source-return rate make it the most citation-dense engine in this sample.

Stability

Citation sets are unstable. In the 10 same-day repeat groups tested, 9 of 10 changed their cited-domain set — a 90% flip rate. Practitioners should treat any single citation observation as a snapshot, not a steady state.

Implications

Editorial content dominates the citation economy by volume, but community platforms and video sources appear with a frequency that pure traffic or domain-authority logic would not predict. Engine behavior varies substantially: ChatGPT is sparse and inconsistently sourced, Gemini is mid-range, and Perplexity is consistently high-volume in its citations. Any strategy aimed at improving AI answer engine visibility should account for this engine-level heterogeneity and the high day-to-day citation variance documented here.

Key findings

  • reddit.com was cited in 39 of 110 checks (35.5%, 95% CI 27.1%–44.7%), making it the single most-cited domain across all three engines on the AEO topic.
  • Editorial sources accounted for 966 citations — 88% of all citations recorded — despite individual editorial domains each appearing in fewer than 17% of checks.
  • ChatGPT returned sources in only 13 of 30 checks at an average of 2 sources per answer, while Perplexity returned sources in all 50 of its checks at an average of 14 sources per answer.
  • YouTube ranked second overall, cited in 36 of 110 checks (32.7%, 95% CI 24.7%–41.9%), appearing primarily via Perplexity where it accumulated 32 appearances.
  • 9 of 10 same-day repeat groups changed their cited-domain set entirely, yielding a 90% flip rate that underscores citation instability across engines.
  • semrush.com led all editorial properties, cited in 18 of 110 checks (16.4%, 95% CI 10.6%–24.4%), while coursera.org ranked 15th at 8 of 110 checks (7.3%, 95% CI 3.7%–13.7%).

Findings

Citation Concentration and the Domain Leaderboard

Across 110 recorded engine checks conducted on 2026-07-26 — spanning 30 category questions and 3 engines (ChatGPT, Gemini, Perplexity) — a clear citation hierarchy emerged. No checks failed; all 110 are included in the analysis. The leaderboard below reports Wilson 95% confidence intervals throughout.

reddit.com ranked first, appearing in 39 of 110 checks (35.5%, 95% CI 27.1%–44.7%). This is the single highest individual-domain citation rate in the sample. youtube.com ranked second at 36 of 110 checks (32.7%, 95% CI 24.7%–41.9%). The confidence intervals for these two domains overlap substantially, meaning their ordering should not be treated as definitive at this sample size, but both sit well clear of the remainder of the leaderboard.

linkedin.com ranked third at 20 of 110 checks (18.2%, 95% CI 12.1%–26.4%), followed closely by semrush.com at 18 of 110 checks (16.4%, 95% CI 10.6%–24.4%). These two domains represent the upper tier of what might be called the mid-citation band.

Five domains clustered between 10.9% and 13.6% citation rates: medium.com (15 of 110, 13.6%, 95% CI 8.4%–21.3%), searchengineland.com (15 of 110, 13.6%, 95% CI 8.4%–21.3%), blog.hubspot.com (12 of 110, 10.9%, 95% CI 6.4%–18.1%), frase.io (12 of 110, 10.9%, 95% CI 6.4%–18.1%), and hubspot.com (12 of 110, 10.9%, 95% CI 6.4%–18.1%). The overlapping confidence intervals across this cluster indicate that rank ordering within it carries limited statistical weight.

The lower portion of the top 15 contains four domains in the 8.2%–10% band: tryprofound.com (11 of 110, 10%, 95% CI 5.7%–17%), surferseo.com (10 of 110, 9.1%, 95% CI 5%–15.9%), and three domains tied at 9 of 110 checks each — authoritytech.io, conductor.com, and ziptie.dev (all 8.2%, 95% CI 4.4%–14.8%). coursera.org closes the leaderboard at 8 of 110 checks (7.3%, 95% CI 3.7%–13.7%).

A notable structural feature of this leaderboard is how quickly citation rates fall after the top two domains. reddit.com and youtube.com each clear 32%, while the fifteenth-ranked domain sits at 7.3%. That gap is large relative to the sample, suggesting that citation volume is meaningfully concentrated at the top even within a competitive category.

The Source-Type Economy

When citations are aggregated by source type rather than individual domain, the concentration becomes even more pronounced. Editorial sources — defined in this study as purpose-built content and marketing industry publications — accounted for 966 citations, representing 88% of all citations recorded. No other source type comes close.

Reddit as a source type contributed 39 citations (3.6%), equal in raw count to the social category (39 citations, 3.6%), which in this dataset is anchored primarily by linkedin.com. YouTube as a category accounted for 36 citations (3.3%). Docs-type sources produced 13 citations (1.2%), and review sites generated 5 citations (0.5%).

The editorial dominance — 88% of the citation economy — is the single most structurally important finding in this dataset. It establishes that, for this category of questions asked on 2026-07-26, answer engines drew overwhelmingly from long-form written editorial content rather than community platforms, video, or official documentation. Reddit and YouTube, despite holding the top two individual-domain positions, together represent only 6.9% of total citation volume when counted at the source-type level, because the editorial category is populated by a large number of contributing domains each accumulating moderate individual citation counts.

Per-question data reinforce the heterogeneity beneath the aggregate: the most-cited domain shifts substantially by question, with hubspot.com leading for "What is answer engine optimization?" (4 citations), frase.io leading for "What are the best AI visibility tracking tools?" (4 citations), and reddit.com leading for "What is the difference between SEO and AEO?" (4 citations). No single domain dominated universally across question types.

Engine behavior

Across 110 recorded checks conducted on 2026-07-26, the three engines produced measurably different behavior in how often they returned sources, how many they returned per answer, and which domains they favored.

ChatGPT returned sources in 13 of its 30 checks, meaning 17 checks produced zero cited sources — a measured negative that reflects how the engine handles this topic category. When sources did appear, the average was 2 per answer, the lowest of the three engines. The domains that appeared most frequently were schema.org (2 citations), support.google.com (2), and techradar.com (2). The low source rate and low per-answer count suggest ChatGPT applies more selective or restrictive citation behavior for answer engine optimization queries, at least as measured in this sample.

Gemini returned sources in 29 of its 30 checks, with one check producing no cited sources — again recorded as a measured negative. The average was 11.2 sources per answer. Its most-cited domains were reddit.com (9 citations), medium.com (8), hubspot.com (5), frase.io (4), and llmpulse.ai (4). The source mix skews toward community discussion platforms and mid-tier web publications, with no single authoritative institutional source dominating.

Perplexity returned sources in all 50 of its checks, with zero zero-source answers in this sample. It produced the highest average source count at 14 per answer. Its most-cited domains were youtube.com (32 citations), reddit.com (30), linkedin.com (20), semrush.com (14), and searchengineland.com (13). The presence of youtube.com at the top of this list is a structurally notable finding: Perplexity draws heavily from video content URLs as citeable sources, a pattern not visible in either of the other two engines. Reddit appeared prominently across both Gemini and Perplexity but was absent from ChatGPT's top-cited set entirely.

Test-Retest Stability

The 10 same-day repeat groups reveal a citation instability problem that cuts across engines. In 9 of 10 repeat groups, the cited-domain set changed between runs — a 90% flip rate. This means that even within a single day, the same question directed to the same engine does not reliably return the same sources. Practitioners treating any single citation observation as a stable measurement should account for this variance. The flip rate does not mean sources are random, but it does mean the citation set for any given query is not fixed. Longitudinal tracking across repeated measurements is necessary to distinguish signal from session-level noise.

What this means

The citation economy measured across 110 checks on 2026-07-26 is heavily concentrated in one source type. Editorial content — structured articles, guides, and blog posts — accounts for 966 citations, or 88% of all citations recorded. Every other source type trails by a wide margin: Reddit at 3.6%, social (LinkedIn, Medium) at 3.6%, YouTube at 3.3%, docs at 1.2%, and review sites at 0.5%. For teams working in the answer engine optimization and AI visibility category, this distribution carries a direct implication: AI answer engines, when queried on this topic, draw almost exclusively from editorial content. Other formats contribute, but they are marginal in aggregate.

The domain leaderboard reinforces this. Reddit leads all individual domains at 39 of 110 checks (35.5%, 95% CI 27.1%–44.7%), and YouTube places second at 36 of 110 checks (32.7%, 95% CI 24.7%–41.9%). Both are notable given that the overall citation share for their respective types is small — which means a handful of high-performing URLs within those platforms are doing most of the work. Teams should not interpret category-level type data as meaning all Reddit or YouTube content is equally likely to be cited.

Among editorial domains, semrush.com, searchengineland.com, blog.hubspot.com, frase.io, hubspot.com, tryprofound.com, surferseo.com, authoritytech.io, conductor.com, and ziptie.dev all appear in the leaderboard, but none exceeds 18 of 110 checks. The top editorial domain reaches 16.4% citation frequency. That ceiling is low relative to the editorial category's dominance overall, which suggests citation share is distributed across many editorial sources rather than locked up by a few. That distribution pattern points to open ground: no single editorial voice has saturated AI retrieval on this topic.

For teams seeking AI visibility in this category, the data recommends prioritizing publication of well-structured editorial content. The consistent appearance of mid-tier and specialist editorial domains in the leaderboard suggests that domain authority relative to major platforms is not a prerequisite for citation. Format and topical fit appear to matter. Video and community content can contribute to citation presence, as YouTube and Reddit demonstrate, but editorial remains the dominant surface by a factor of roughly 24 to 1 over any individual non-editorial type. Teams should treat editorial production as the primary lever and monitor citation frequency over time to assess whether specific content earns placement.

Methodology and data

This study was produced with a deterministic measurement instrument (110 planned engine calls, 110 recorded, 0 failed — recorded as negatives). The complete instrument description, sample, refusal conditions, and limitations are in the methodology; every underlying measurement is in the dataset.