YouTube AI Citations Flow to Google, Not ChatGPT. That Changes the Advice.

When practitioners discuss YouTube's role in AI search, the headline figure is usually the same: YouTube accounts for 38.7% of all social media AI citations, per the OtterlyAI March 2026 citation study covering 100 million-plus citation instances across six platforms. The implied conclusion is that YouTube is an AI optimization lever.

The number is accurate. The conclusion is platform-specific, and that specificity changes the fix plan.

The 38.7% figure needs a denominator

The OtterlyAI platform breakdown for YouTube citations in their March 2026 dataset:

- Perplexity: 38.7% of all YouTube citations - Google AI Overviews: 36.6% - ChatGPT: 4.4% - Gemini standalone: under 1% - Microsoft Copilot: under 1%

Perplexity at the top looks like the platform to optimize YouTube for. Our August 9, 2026 research correction (Scout session 102, `research-vault/knowledge/youtube-as-aeo-signal.md`) addresses that reading directly.

The 38.7% is an absolute citation count, not a platform-level share. It tells you how many YouTube citation instances originated from Perplexity across all platforms combined. It does not tell you what share of Perplexity's own citations are YouTube videos.

Tinuiti's Q1 2026 AI Citation Trends Report supplies the denominator: YouTube accounts for 3% of Perplexity's total citations as of January 2026. Both figures are accurate and compatible -- Perplexity cites many sources per response, and 3% of a large base produces a large absolute count. But as an optimization target, YouTube is a minor Perplexity signal. After Reddit citations dropped 86% on Perplexity following the October 2025 lawsuit, Perplexity did not replace Reddit with YouTube. It reduced all social citations.

ChatGPT drives the leads. YouTube barely touches it.

ChatGPT accounts for 87.4% of all AI referral traffic, per Conductor's 2026 data. For any business trying to increase AI-generated customers, ChatGPT is where almost all of the traffic originates.

YouTube's share of ChatGPT citations: 4.4% of all YouTube citations in the OtterlyAI dataset.

For local service queries specifically, the mechanism is not web retrieval. ChatGPT local recommendations flow through three confirmed data partnerships: Foursquare (December 2024 integration), Thumbtack for home services (October 2025 integration via OpenAI's Apps SDK, documented in Scout session 96, August 3, 2026), and Yelp (July 2026 data licensing agreement covering 330 million reviews). These are structured API feeds. Website schema, content quality, and video presence do not enter that pipeline.

A business creating YouTube content to improve ChatGPT local recommendations is working on a surface that represents 4.4% of YouTube citations, via a mechanism that does not use YouTube as an input for local queries. The fix path for ChatGPT gaps runs through Foursquare, Yelp, and Thumbtack -- not YouTube.

The timestamps finding applies to two platforms, not all of them

Our June 2026 research (Scout session 43, `research-vault/knowledge/youtube-as-aeo-signal.md`) documented that timestamped YouTube videos get cited more frequently by AI: 78% of timestamped videos cited by AI are cited from multiple distinct chapters. This was presented as a general AI optimization.

The August 9, 2026 session 102 correction is specific. Among all timestamped YouTube citations in the OtterlyAI dataset:

- 73% appeared in Google AI Overviews only - 27% appeared in Google AI Mode only - 0% appeared in ChatGPT, Perplexity, Copilot, or standalone Gemini

Adding timestamps to YouTube videos improves how Google AI Overviews and AI Mode treat that content. There is no measured citation effect on Perplexity, ChatGPT, Copilot, or standalone Gemini. The structuring advice is accurate -- for those two Google platforms.

Where YouTube actually works

Google AI Overviews cites YouTube more than any other domain. The 5WPR YouTube Citation Share Report (2026) puts YouTube at 23.3% of Google AI Overview citations. The OtterlyAI dataset shows 29.5% for AI Overviews from non-Google properties. Both figures place YouTube at the top of the AI Overviews source pool.

For instructional and explanatory content -- how HVAC maintenance works, what to look for when hiring an electrician, what happens during a roof inspection -- YouTube is the primary input Google AI Overviews draws from. Google's systems already know how to extract transcript content, chapter structure, and video metadata from YouTube. A 10-20 minute instructional video with accurate captions, chapter markers, and a description that includes the business name and city can appear in AI Overviews for relevant informational queries.

Google AI Mode also pulls YouTube into its Expert Advice and Community Perspectives sections for trade and service category queries. For businesses producing content about how their service actually works -- not just who they are -- YouTube is a legitimate signal in the informational layer of Google's AI surfaces.

One distinction that matters for local businesses: AI Overviews appears on approximately 15% of local-intent queries, compared to 92% for informational queries. YouTube citations in AI Overviews serve the informational layer -- "how does X work," not "who should I hire for X." If a homeowner wants to understand heat pump maintenance, your YouTube explainer might appear. If they are actively searching for the best HVAC company in their area, the citation surfaces driving those recommendations are different ones.

What this means for fix plan sequencing

Our August 9, 2026 research (session 102) establishes the standing recommendation: YouTube does not belong in Phase 1 or Phase 2 fix plans for home services clients. The primary AI citation mechanisms for home services recommendations are structured database integrations and directory quality. YouTube is a Phase 3 consideration for businesses that have the directory and schema layer complete and are looking for long-tail informational query coverage in Google's AI surfaces.

For businesses with documented Google AI Overviews gaps, YouTube is the action worth pursuing. The content format that gets cited: 5-20 minute instructional video, accurate transcript, chapter markers in the description, title and description that include both the service category and city. Our May 2026 research documented that 41% of AI-cited YouTube videos had under 1,000 views -- AI Overviews selects for topical clarity and structural quality, not audience size. A business does not need to build a YouTube following. One well-constructed video on a specific service question can be cited.

For ChatGPT and Perplexity gaps, the fix path does not run through YouTube. ChatGPT local recommendations come from database integrations that website content cannot reach. Perplexity draws from web content with strong directory and domain signals. For both platforms, the productive action after checking platform scores is directory infrastructure and content freshness -- not a recording setup.

A Signal Check at sourcepull.ca gives per-platform scores across ChatGPT, Perplexity, Claude, and Gemini. The breakdown shows which gap is actually largest and which fix applies. For some businesses, that answer is YouTube -- for the specific platforms and query types where YouTube is a genuine input. For most, it is not.