Weekly index · week 13 edition
YouTube in AI search: week 13
How much each AI model cites YouTube in its answers, week by week, and where that presence is changing.
This is the week 13 edition, kept as published. See the current edition →
In short
In week 13, Grok was the model that cited YouTube most, in 76.0% of answers, followed by Google AI Overviews (66.3%). Significant change from the previous week: Google AI Overviews (+12.7 pp), Grok (+9.0 pp), Perplexity (+4.7 pp), Google AI Mode (−1.7 pp). Since week 1, the largest drop is ChatGPT's (−4.0 pp) and the largest rise is Grok's (+56.6 pp).
In week 13, by model
Share of answers citing YouTube and change from week 12.
Timeline
Weekly presence since week 1. Each panel has its own scale; compare the shape of the curves, not heights across panels. Hollow circles mark weeks with a reduced base; blue markers are events, click to see them.
All weeks
Week 13 ranking
| Model | Presence | Week change | 4-week avg. | Since W1 | Change |
|---|---|---|---|---|---|
| 1. Grok | 76.0% | ↑ +9.0 pp | 67.1% | ↑ +56.6 pp | Significant |
| 2. Google AI Overviews | 66.3% | ↑ +12.7 pp | 61.4% | ↑ +16.1 pp | Significant |
| 3. Google AI Mode | 55.8% | ↓ −1.7 pp | 54.1% | ↑ +4.9 pp | Significant |
| 4. Perplexity | 33.3% | ↑ +4.7 pp | 28.9% | ↑ +10.1 pp | Significant |
| 5. ChatGPT | 0.5% | ↓ −0.3 pp | 0.4% | ↓ −4.0 pp | Within noise |
| 6. Copilot | 0.5% | ↓ −0.2 pp | 0.7% | ↑ +0.2 pp | Within noise |
| 7. Gemini | 0.3% | ↑ +0.1 pp | 0.2% | ↑ +0.1 pp | Within noise |
Presence: share of the model's answers that cite youtube.com among their sources. Changes in percentage points (pp).
Events
Large changes in the series and what is known about them. We only log explanations with a source; when there is none, we say so.
- W11 · Mar 9–15 Observed by third parties Google AI Overviews goes from 50% in January to 65–71% between weeks 13 and 18, and Google AI Mode from 33% (week 3) to 55–57% between weeks 11 and 14. Tinuiti, with Profound data, recorded YouTube's share of AI Overviews citations more than doubling, and of AI Mode citations more than quadrupling, between January and April. Google has not commented on the cause. Source: How YouTube became the most cited platform in LLMs ↗
- W10 · Mar 2–8 Possible ChatGPT falls from 4.2% (week 9) to 0.2% and stays below 1.8% through week 18. The same week, on March 3, OpenAI switched the default model to GPT-5.3 Instant, described as less likely to lean on web results. No one has linked this change to YouTube. Source: OpenAI releases GPT-5.3 Instant for smoother everyday conversations and better search ↗ · Source: GPT-5.3 Instant: ChatGPT Shows Fewer Links in Web Search — SEO Impact ↗
- W6 · Feb 2–8 No explanation found ChatGPT falls from 5.1% (week 5) to 0.6%, swings between 1.1% and 2.0% over the next two weeks and returns to 4.2% in week 9. We found no public change to ChatGPT in the week of the drop.
- W4 · Jan 19–25 No explanation found Grok rises from 18.7% (week 3) to 38.2%, 49.6% (week 6), 66.1% (week 11) and 76–82% between weeks 13 and 16. We found no xAI announcement or public measurement of this rise.
- W4 · Jan 19–25 Possible Google AI Overviews falls from 48.3% (week 3) to 35.1%, 24.6% and 31.4% (week 6), and returns to 51.7% in week 7. On January 27, Gemini 3 became the default model for AI Overviews, and on February 4 Google acknowledged a bug that left some answers without links; SE Ranking measured the missing sources in those weeks. The drop starts earlier, on January 19, and no one has linked these changes to YouTube. Source: Gemini 3 Replaces 42% of Previously Cited Domains in AI Overviews ↗ · Source: Google AI Overviews Bug Removes Links & Citations ↗
- W2 · Jan 5–11 No explanation found Google AI Mode falls from 50.9% (week 1) to 43.0% and 33.4% (week 3), and only rises above 40% again in week 8. We found no public change to AI Mode in that window.
Method, in short
Source: data observed and catalogued by First Answer. Each week, we measure by model the share of answers that cite youtube.com among their sources and compare it with the previous week. A change is flagged as significant only when it exceeds the 95% margin and is at least 1 pp. Weeks with a smaller-than-usual base are flagged and left out of the comparison.
Limitations
The index measures what First Answer's base observes, not every search made on the models. We do not disclose volumes, questions, topics, categories, brands, channels or clients; only aggregate measures by model. Claude is not included in this series. Explanations under Events come from external sources and state how confirmed they are.
How to cite
Beckmann, R. (2026). YouTube in AI search: weekly index (week 13, Mar 23–29, 2026). First Answer. https://rodrigobeckmann.com/en/research/indices/youtube-in-ai-search/2026-w13