TL;DR
- Meltwater data reported by Digiday on 23 September 2026 put YouTube as the most-cited platform across eight AI products in August 2026, ahead of Reddit.
- YouTube's share of citations in Google AI Mode more than quadrupled from January to April 2026, and more than doubled in AI Overviews, per a Q2 report built on Profound tracking.
- OtterlyAI data puts 94% of YouTube citations on long-form videos and 5.7% on Shorts; 10-20 minute videos take 32.1%.
- Scrunch found YouTube citations fell about 43% within a month (18.4% to 10.5%) across sponsored videos tracked May to July 2026; no figure is broken down by country.
YouTube was the most-cited platform across eight AI answer engines in August 2026, according to Meltwater data published by Digiday on 23 September 2026, overtaking Reddit, which had held that position. Other datasets in the same report add three details that matter for anyone trying to be cited: long-form videos take about 94% of YouTube citations, YouTube's share of citations in Google AI Mode more than quadrupled between January and April 2026, and citations fade quickly, falling about 43% within a month of first appearing.
The report, published by Digiday and written by Krystal Scanlon, pulls together data from several AI-visibility and monitoring tools. None of them is a complete census of AI answers, and they use different prompts, time windows and definitions of a citation. Read together, though, they point the same way. This article sets out each finding, what it does and does not show, and what a brand in Thailand can do with it.
What the data shows
YouTube leads citations across eight AI platforms
Meltwater's data showed YouTube as the most-cited platform in August 2026 across eight AI products: Claude, ChatGPT, Copilot, Google AI Mode, Google AI Overviews, Gemini, Perplexity and Grok. Digiday noted that YouTube continues to surpass Reddit, which used to lead after Google struck a deal with Reddit to make its content available for AI products. The breadth matters. This is not one model with a preference for its parent company's video site. Models from OpenAI, Anthropic, Microsoft, Perplexity and xAI are in the same list.
The rise is steepest inside Google's own AI products
A Q2 2026 citation-trends report, built on a consistent set of commercial-intent prompts tracked through the AI visibility platform Profound, found that YouTube's share of citations in Google AI Mode more than quadrupled between January and April 2026. Its share in AI Overviews more than doubled over the same period. Google owns YouTube, so a larger share inside Google surfaces is not surprising. The size of the change over four months is the notable part.
A citation is not always a source
Digiday cited analysis from Weglot comparing YouTube citations in Claude and Gemini. Both models frequently embed YouTube videos as recommended content. That means a YouTube link in an AI answer can be a pointer for further viewing, rather than evidence that the model drew on the video to write its answer. This is a useful caution for anyone reading AI-visibility dashboards: a citation count mixes "the answer came from here" with "watch this next".
Long videos get cited, Shorts mostly do not
Data from OtterlyAI, which Digiday said was cited on the developer site DEV Community by a user called Watson Foglift, puts 94% of YouTube citations on long-form videos and 5.7% on Shorts. Videos between 10 and 20 minutes accounted for 32.1% of citations. The data reached Digiday second-hand, so treat the exact figures with some care. Digiday's suggested explanation is structural: videos of that length are more likely to have chapter markers and timestamps, which give AI systems signals about where each topic appears, and descriptive titles and descriptions add further context.
Citations fade within a month
Scrunch, which Digiday described as an agent experience platform, analysed thousands of sponsored YouTube videos and millions of citation events across AI platforms between May and July 2026. It found YouTube citations faded by around 43% within a month, with the citation rate falling from 18.4% to 10.5% in the month after a video was first cited. The likely reason, per the report, is that AI systems keep refreshing retrieval and ranking, so newer videos displace older ones and the topics people search for shift.
The findings in one table
Each figure below comes from a different tool, with different methods, so the rows should be read as separate signals rather than a single dataset.
| Finding | Figure | Data source (via Digiday) |
|---|---|---|
| Most-cited platform across 8 AI products | YouTube, August 2026, ahead of Reddit | Meltwater |
| YouTube share of citations in Google AI Mode | More than 4x, January to April 2026 | Q2 2026 report built on Profound tracking |
| YouTube share of citations in AI Overviews | More than 2x, January to April 2026 | Q2 2026 report built on Profound tracking |
| Long-form vs Shorts share of YouTube citations | 94% long-form, 5.7% Shorts; 10-20 minute videos 32.1% | OtterlyAI, via DEV Community |
| Citation decay after first citation | About 43% within a month (18.4% to 10.5%) | Scrunch, May to July 2026 |
What the report does not show
It is worth being precise about the limits, because AI-visibility data is easy to over-read.
- No market breakdown. None of the figures is split by country or language. There is no data here on Thai-language prompts or Thai YouTube content.
- No traffic figures. A citation is not a click. The report does not say how many people follow a YouTube citation, or whether it leads to purchases.
- Different prompt sets. The AI Mode figure uses commercial-intent prompts. The decay figure uses sponsored videos. Neither describes all AI answers.
- Causation is a hypothesis. The link between chapter markers and citations is Digiday's explanation, not a tested finding in the data.
Why YouTube suits AI answers
Several mechanics make YouTube easy for AI systems to use, and none of them depends on the data above. YouTube can generate automatic captions for many videos, so spoken content becomes text that can be retrieved. Titles, descriptions and chapters give a structure that maps well to a question. A single long video often covers a full how-to process, which suits the step-by-step answers AI tools like to give. For Google's own products there is a further factor: Google has direct access to YouTube's content and metadata.
This connects to a separate announcement the same day. At Made On YouTube on 23 September, YouTube said Ask YouTube will show comparison tables for product review searches. So video content is becoming an input to AI answers both outside YouTube, in the engines above, and inside it.
What brands should do with this
- Check what AI answers cite for your category. Run your main commercial questions in Google AI Mode, AI Overviews, ChatGPT and Gemini. Note which YouTube videos appear, and whether they are yours, a creator's or a competitor's.
- Prioritise long-form explainers over Shorts for citation goals. Shorts have their place for reach, but the OtterlyAI split suggests they are rarely what AI tools point to. A 10 to 20 minute explainer with chapters is a better fit.
- Add chapters and write descriptive titles. Chapter markers, accurate titles and a description that names the product and the question make it easier for any system to understand what the video covers.
- Check the transcript. Auto-generated transcripts can mangle brand and product names, especially in Thai. If the name is wrong in the transcript, a model working from the text may not connect the video to your brand.
- Plan for decay. If citations fall by around 43% within a month, a single video is not a lasting asset for AI visibility. Refreshing and updating content on a schedule matters more than one launch.
- Measure citations separately from traffic. Given the Weglot caution, track whether you are the source of an answer or only a recommended link, and do not treat either as a sales result without further data.
This is the same discipline as generative engine optimisation for web pages, applied to video. It also overlaps with how you run YouTube advertising, because sponsored and brand-owned videos that earn citations are the ones worth supporting with paid media.
What this means for Thai marketers
The report contains no Thailand-specific data, so the honest reading is directional. YouTube is a major platform in Thailand, and much Thai how-to and product review content already lives there, often in long-form formats. If AI answers draw on YouTube more, Thai brands that publish clear long-form explainers in Thai, with correct product names in titles, descriptions and transcripts, are better placed than brands that only post Shorts. The practical risk for Thai content is transcription: Thai auto-captions can be less reliable than English ones for brand names and technical terms, so checking and correcting captions is a small task with a real payoff. Brands already working on AI SEO for their websites should add their YouTube channel to the same audit.
Frequently asked questions
Which platform do AI answer engines cite most?
YouTube was the most-cited platform across eight AI products in August 2026, according to Meltwater data reported by Digiday on 23 September 2026. The eight were Claude, ChatGPT, Copilot, Google AI Mode, Google AI Overviews, Gemini, Perplexity and Grok, and YouTube was ahead of Reddit.
Do YouTube Shorts get cited by AI tools?
Rarely, according to OtterlyAI data cited by Digiday, which put Shorts at 5.7% of YouTube citations against 94% for long-form videos. Videos of 10 to 20 minutes made up 32.1% of citations.
Does a YouTube citation mean the AI used the video to answer?
Not necessarily. Weglot's analysis, cited by Digiday, found that Claude and Gemini often embed YouTube videos as recommended viewing, so a citation can be a pointer rather than the source of the answer.
How long do YouTube citations last in AI answers?
Not long, based on Scrunch's data: citations fell about 43% within a month, from a citation rate of 18.4% to 10.5%. The analysis covered sponsored YouTube videos between May and July 2026.
Is there data for Thailand?
No. The report does not break any figure down by country or language, so it says nothing directly about Thai prompts or Thai-language videos.
The short version
In Meltwater's August 2026 data, the eight AI engines tracked cited YouTube more than any other platform. Separate datasets suggest long-form videos with clear structure are what they cite, and those citations fade fast. For brands, that makes YouTube part of AI visibility work rather than a separate channel. If you want your video content and your website reviewed together for how AI tools see them, Relevant Audience's GEO team can help.
