AEOForgedAEOForged
Get started

Original research · measured 2026-07-27

LinkedIn Leads AI Citation at 34.8% Across 92 Checks, With a 90% Same-Day Domain Flip Rate

Original research · 92 engine checks across 24 questions, measured on 2026-07-27 · Methodology · Dataset

Executive summary

This study measured which content formats and domains receive citations from AI answer engines when queried about building brand authority signals for AEO and GEO. Across 92 recorded engine checks spanning 24 category questions and 3 engines (ChatGPT, Gemini, Perplexity), conducted on 2026-07-27, the single most striking finding is that linkedin.com was cited in 32 of 92 checks — a 34.8% citation rate (95% CI: 25.8%–44.9%) — making it the most-cited domain in the category by a substantial margin over any other source.

Source Format Economy

The content format picture is unambiguous: editorial content dominated the citation pool, accounting for 821 of the total citations, representing 91.2% of all citation share. Social content followed at 43 citations (4.8%), with YouTube and Reddit each contributing 16 citations (1.8% each). Docs formats produced 3 citations (0.3%) and review sites 1 citation (0.1%). For practitioners seeking to earn citations in this category, editorial-format content represents the overwhelming structural preference of current AI answer engines.

Engine-Level Divergence

The three engines behaved very differently. Perplexity ran 44 checks and returned sources on all 44, averaging 14.6 sources per answer — the highest source density observed. ChatGPT ran 24 checks but returned sources on only 3, averaging 0.2 sources per answer. Gemini ran 24 checks, returned sources on all 24, and averaged 10.6 sources per answer. ChatGPT's near-absence of cited sources is not a data gap; it is a measured behavioral characteristic of that engine on this category as of the test date. Every question produced at least one source on some engine, and 0 of 92 planned calls failed.

Variance Panel: The Instability Finding

The test-retest variance panel — 10 same-day repeat groups querying identical questions — found that 9 of 10 repeat groups changed their cited-domain set, a 90% flip rate. This is the finding that most directly complicates any assumption of stable AI citation position. The leading cited domain on a given question can differ from one query to the next within the same day. Citation visibility in this category should be understood as probabilistic, not deterministic.

Domain Leaderboard Context

Beyond linkedin.com at 34.8%, the next tier includes reddit.com and youtube.com tied at 17.4% each (16 of 92 checks, 95% CI: 11.0%–26.4%), followed by forbes.com at 16.3% (15 of 92 checks) and searchengineland.com at 14.1% (13 of 92 checks). The top 15 domains span social, editorial, and platform formats, though editorial sources account for 13 of those 15 slots — consistent with the overall source economy.

These findings establish a measured baseline for citation behavior in the brand authority signal category as of the test date. The 90% same-day flip rate and ChatGPT's near-zero sourcing rate are findings that any strategy in this space should account for directly.

Key findings

  • linkedin.com was the most-cited domain across 92 checks, appearing in 32 checks at a 34.8% citation rate (95% CI: 25.8%–44.9%) on 2026-07-27.
  • Editorial content captured 91.2% of all citations (821 total), vastly outpacing social at 4.8%, YouTube at 1.8%, and Reddit at 1.8%.
  • ChatGPT returned sources in only 3 of 24 checks, averaging 0.2 sources per answer, while Perplexity averaged 14.6 sources per answer across all 44 of its checks.
  • 9 of 10 same-day repeat groups changed their cited-domain set, producing a 90% flip rate that indicates citation positions are probabilistic rather than stable.
  • reddit.com and youtube.com tied for second place on the domain leaderboard, each cited in 16 of 92 checks at 17.4% (95% CI: 11.0%–26.4%).
  • 0 of 92 planned engine checks failed, and every one of the 24 category questions produced at least one source on at least one engine.

Findings

Citation Economy and Domain Leaderboard

Across 92 recorded engine checks spanning 24 category questions and 3 engines (ChatGPT, Gemini, Perplexity), measured on 2026-07-27, a clear hierarchy of cited domains emerged. Zero of 92 planned calls failed, giving a complete sample on which all percentages below are based.

linkedin.com led all domains by a substantial margin, appearing in 32 of 92 checks (34.8%, 95% CI 25.8%–44.9%). This places LinkedIn in a citation tier of its own — its lower confidence bound (25.8%) exceeds the upper bound of every other domain in the leaderboard.

reddit.com and youtube.com tied for second, each cited in 16 of 92 checks (17.4%, 95% CI 11.0%–26.4%). forbes.com followed at 15 of 92 checks (16.3%, 95% CI 10.1%–25.2%), and searchengineland.com at 13 of 92 checks (14.1%, 95% CI 8.4%–22.7%).

The mid-tier cluster — authoritytech.io (12 of 92, 13.0%, 95% CI 7.6%–21.4%), moz.com (10 of 92, 10.9%, 95% CI 6.0%–18.9%), and a three-way tie among conbersa.ai, insidea.com, and medium.com (each 9 of 92, 9.8%, 95% CI 5.2%–17.6%) — shows substantially overlapping confidence intervals, meaning their rank order carries low statistical certainty.

The lower tier is similarly dense. searchatlas.com and semrush.com each appeared in 8 of 92 checks (8.7%, 95% CI 4.5%–16.2%), while 321webmarketing.com, lumar.io, and spawned.com each registered 7 of 92 (7.6%, 95% CI 3.7%–14.9%). The overlapping intervals across positions 6 through 15 indicate that the rank ordering below LinkedIn should be treated as approximate rather than definitive.

Source-Type Economy

The source-type breakdown reveals a pronounced concentration. Of the total recorded citations, editorial sources accounted for 821 citations, representing 91.2% of the entire citation pool. No other source type came close to this share.

Social sources contributed 43 citations (4.8%), driven substantially by LinkedIn's and Medium's citation totals. YouTube and Reddit each contributed 16 citations (1.8% each). Docs accounted for 3 citations (0.3%), and review_site for just 1 citation (0.1%).

This distribution means that nine out of every ten citations in this category went to editorial-classified domains. The social, video, and community formats together account for only 8.4% of all citations combined. The practical implication for researchers studying this category is that editorial publication formats — industry blogs, SEO tools with content hubs, and trade publications — dominate the citation surface area, while social and community formats represent a meaningful but secondary presence.

Per-Question Leading Domains

At the individual question level, no single domain dominates across all 24 questions. Instead, different domains lead on different questions: mailchimp.com led on "how to build brand authority" (4x), insidea.com led on both "GEO and brand authority" and "how do backlinks affect answer engine trust" (4x each), and forbes.com led on "what role does founder personal brand play in AI citations" (4x). Four of the 24 questions returned sources in only 2 of their allotted checks, indicating that certain narrower or more speculative queries produced fewer sourced responses from the engines tested. Notably, "how do case studies affect brand credibility in AI answers" produced a leading domain with only 1 citation (contagious.com, 1x), illustrating how thinly sourced some question-level results are.

Taken together, the citation economy in this category is structurally top-heavy at the source-type level (91.2% editorial) while remaining distributed at the domain level — a pattern where no single editorial domain breaks above a 34.8% check-level citation rate even at the top of the leaderboard.

Engine behavior

Measured 2026-07-27 · 92 recorded engine checks across 24 category questions and 3 engines · 0 of 92 checks failed

Source Volume Varies Dramatically by Engine

The three engines produced radically different citation behaviors across the same 24-question category inventory. Perplexity ran the most checks (44) and averaged 14.6 sources per answer, with all 44 checks returning sources. Gemini ran 24 checks and averaged 10.6 sources per answer, also returning sources on every check. ChatGPT ran 24 checks and averaged 0.2 sources per answer — only 3 of its 24 checks returned any sources at all. That means 21 of ChatGPT's 24 responses produced zero cited sources. These are measured negatives, not missing data: the engine was queried, it answered, and it returned no citable references. Any analysis of citation frequency must account for this structural difference between engines.

Domain Preferences Differ by Engine

Each engine favored a distinct domain profile. Gemini's citations concentrated on publisher and institutional domains: forbes.com appeared 7 times, medium.com 5 times, authoritytech.io 4 times, and conductor.com and digitalmarketinginstitute.com each 3 times across its 24 checks.

Perplexity drew from a broader and more platform-heavy set. linkedin.com led at 32 appearances across 44 checks, followed by youtube.com (16x), reddit.com (14x), searchengineland.com (10x), and authoritytech.io (8x). The presence of social and community platforms — LinkedIn, YouTube, and Reddit — at the top of Perplexity's citation list is a structural observation about what that engine indexes and surfaces, not an editorial judgment.

ChatGPT's citation set was sparse by definition. Across only 3 source-returning checks, the recorded domains were moz.com (1x), presenc.ai (1x), schema.org (1x), and visiblie.com (1x). No domain appeared more than once, and the sample is too small to draw frequency conclusions.

One domain, authoritytech.io, appeared in both Gemini (4x) and Perplexity (8x) citation sets, making it the only domain with cross-engine presence in this dataset.

Test-Retest Flip Rate Signals Low Answer Stability

A 10-group same-day variance panel tested whether the same question asked again on the same day returned the same cited domains. 9 of 10 repeat groups changed their cited-domain set, producing a 90% flip rate. This is not a margin-of-error artifact — it is a direct count of groups where the domain set differed on repeat. The finding indicates that citation presence on any given engine check is not a stable state. A domain appearing in one answer does not reliably appear in the next answer to the same question on the same day. Strategies oriented toward AI citation should account for this instability rather than treat any single observed citation as a durable outcome.

What this means

The 92-check measurement taken on 2026-07-27 produces a clear structural picture of how AI answer engines draw on sources when answering questions about building brand authority for AEO and GEO. The findings have direct implications for any team operating in this category.

Editorial content dominates the citation economy by a wide margin. Of all citations recorded, editorial sources claimed 821 citations — 91.2% of the total pool. Social platforms contributed 4.8% (43 citations), YouTube 1.8% (16 citations), and Reddit 1.8% (16 citations). Documentation and review sites together accounted for less than half a percent. Any team not maintaining a structured editorial presence — published articles, guides, or practitioner analyses on a crawlable domain — is operating almost entirely outside the formats that AI engines are currently drawing from in this category.

LinkedIn is the single most-cited domain, but its lead is narrower than it appears. LinkedIn.com appeared in 32 of 92 checks (34.8%, 95% CI 25.8%–44.9%), ranking first in the leaderboard. However, with confidence intervals this wide, that lead is not insurmountable. The social category as a whole produced only 4.8% of total citations, which means LinkedIn's ranking reflects frequency of appearance, not share of the overall citation pool. Teams should interpret LinkedIn's position as a signal that structured professional content on the platform is worth maintaining — not that social content alone builds citation authority.

No single editorial domain holds a commanding share. The top editorial domains — forbes.com (16.3%), searchengineland.com (14.1%), authoritytech.io (13.0%), and moz.com (10.9%) — are clustered closely enough that the category remains genuinely contested. Several domains with meaningful citation rates, including conbersa.ai, insidea.com, and spawned.com, each registered at 9.8% and 7.6% respectively, suggesting that mid-tier editorial players are actively competing for AI citation space.

Open ground exists. The category shows no dominant editorial authority with a citation rate above 35%, and the majority of cited domains sit well below 15%. For teams investing in original, well-structured editorial content, the data suggests this category has not yet consolidated around a small set of authoritative sources. That structural openness may represent an opportunity worth pursuing through consistent, substantive publication — though no outcome can be guaranteed.

Methodology and data

This study was produced with a deterministic measurement instrument (92 planned engine calls, 92 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.