# Speaker video: what YouTube as the most-cited AI source means

Author: Marin Niehues, Stage Starter · Published: 2026-07-08
Topic: Visibility & Personal Brand (https://www.stage-starter.de/en/blog/topic/visibility-and-personal-brand/)
Source: https://www.stage-starter.de/en/blog/speaker-video/

A speaker video, meaning the recording of a talk or a showreel of excerpts from several appearances, can become a source that answer engines name in their answers. According to the [Meltwater AI Search Visibility Report](https://meltwater.com/en/blog/ai-search-visibility-report-june-2026), which the company published on 3 July 2026, YouTube was the most-cited source in AI answers in May 2026 with 188,863 citations, an increase of 56.4 per cent on April. Meltwater says it analysed more than 8 million citations from eight major language models, and the report is a study by the vendor Meltwater itself. According to Meltwater, the systems read above all the structured information around a video, that is, title, description and transcript. For speakers it follows that a published recording becomes easy for AI systems to read only when suitable text accompanies it.

**Key points at a glance**
- On 3 July 2026 Meltwater published its AI Search Visibility Report, whose data comes from May 2026 and covers more than 8 million citations from eight major language models.
- YouTube was the most-cited source with 188,863 citations, ahead of Reddit with 111,860 and NIH with 91,227.
- According to Meltwater, YouTube citations rose by 56.4 per cent compared with April, which works out at roughly 120,800 citations in April.
- Meltwater recommends transcripts, descriptions, titles and metadata, because the systems read the information around the video.
- The report shows no effect on individual speakers, because it counts citations of YouTube as a whole.

## Speaker video and AI answers: YouTube leads in May 2026

Meltwater published its report on 3 July 2026. It is titled “June 2026” and contains data from May, and the page names Nathaniel Pallen, Senior Director of Product Management at Meltwater, as its author. What is counted are citations, that is, sources that AI systems name in their answers.

**Most-cited sources in AI answers, May 2026 (Meltwater)**
- 188,863 | YouTube (+56.4 % on April)
- 111,860 | Reddit (+13.0 %)
- 91,227 | NIH (+41.3 %)
- 83,191 | Wikipedia (+55.2 %)
- 74,875 | Statista (+11.8 %)

The report also breaks down where the YouTube citations arise: according to Meltwater, Google AI Mode produced 73,813 of them, Google’s AI Overviews 38,002, Gemini 37,924 and Perplexity 25,949. Together that is 175,688 citations, or about 93 per cent of the 188,863, and the three Google products alone account for 149,739, or about 79 per cent. We calculated this from Meltwater’s figures. For speakers who want to be found, this means that it is above all Google’s AI features that point to YouTube.

## What the Meltwater report shows and where its limits lie

The report comes from a vendor, its page does not explain how the visibility score is calculated, and it names neither the number of prompts analysed nor the sample, so the results can hardly be recalculated from outside. The introduction speaks of eight major language models, while seven are named, from Claude to Grok. A graphic on the page gives YouTube growth of minus 2.29 per cent, whereas the text states plus 56.4 per cent. We rely on the text and the table, because the two agree, and we read the figures as a direction for May.

The subject of the count also calls for caution. Meltwater counts citations of the YouTube domain as a whole, across tutorials or product videos as much as across talks. How many of them concern conference talks is not apparent from the report, and neither does it disclose which videos the systems select.

## Why a speaker video works as a source for AI answers

According to Meltwater, AI platforms read the metadata around a video beyond its popularity, in the words of the report “the structured information around the video”, and as measures Meltwater names the transcript, the description, the title and the metadata. In our assessment, a video on its own contains little that a text-based system can read, whereas a transcript supplies every statement of the talk as text.

That gives us a calculation. Suppose you give six talks a year, three of them are recorded and you are allowed to publish them, and you cut a showreel from them. You then have four videos with three text elements each, which makes twelve pieces of text in which your name can appear together with your topic and your core message. If you allow 80 minutes per video for the description and transcript (an assumption of ours), you arrive at 320 minutes, a good five hours a year. Without those hours, recordings sit online that, in our assessment, an AI system can read only with difficulty.

A recording of a talk also shows people how you come across on stage. Where a call for papers allows or requires a link to a video sample, a well-kept channel with clearly labelled recordings is a better answer than a phone video without text. How appearances show up in AI answers more generally is covered in our article [Getting mentioned in AI answers through conference talks](/en/blog/get-mentioned-in-chatgpt-as-a-speaker/).

## How to make your speaker video readable for AI systems

We derive the following steps from Meltwater’s recommendation, although Meltwater does not show that exactly these steps lead to more citations, and the order and the examples come from our own assessment.

1. Clear the rights: Ask the organiser before the talk whether a recording will be made, whether you may use it on your channel and your website, and at what length. The rights to the recording belong in the agreement, as our article [The speaker agreement: what it covers and what is missing](/en/blog/speaker-agreement/) shows.
2. Title with name and topic: Put your topic first in the video title and then your name, the name of the conference and the year.
3. Description with a core message: The first paragraph summarises the message of the talk in two sentences and links to the conference programme page and your website.
4. Publish a transcript: Have the automatic transcript corrected, above all technical terms and names, and place it with the video, in the description or as text on your website.
5. Cut a showreel: Combine two or three excerpts from different appearances into one short video and label it like a recording.
6. Check the result: Ask ChatGPT, Gemini and Perplexity for specialists on your topic and note whether YouTube videos appear as sources and whether yours are among them.

How recordings behave on online stages is described in our article [Virtual and hybrid conferences: what changes for speakers](/en/blog/virtual-and-hybrid-conferences/), and which questions to put to the organiser before the talk is set out in [What do I need to clarify with the organiser as a speaker?](/en/blog/what-to-clarify-with-the-organiser/).

**How Stage Starter gets you onto conference stages**
Stage Starter is a speaker agency for experts from B2B fields. We research the conferences that suit your topic and your target clients, develop the talk with you and write several versions of the abstract for each call for papers. We then submit in your name, keep track of the deadlines and follow up with the organisers. Our founder Marin Niehues brings the experience of more than 100 invitations on six continents. In the [first conversation](/en/erstgespraech/) we clarify which stages suit you.

**Meltwater counts YouTube as the most-cited source in AI answers in May 2026 and recommends text around the video. If you have your talks recorded and give them a title, description and transcript, you give answer engines readable material about your expertise.**

## Frequently asked questions

### How important is a speaker video for visibility in AI answers?

According to the Meltwater AI Search Visibility Report, YouTube was the most-cited source in AI answers in May 2026 with 188,863 citations. The report counts citations of the platform as a whole and says nothing about individual speakers. A speaker video helps you above all when the text around the video names your topic and your name.

### What belongs in a speaker video?

A speaker video shows you on a stage, either as the recording of a whole talk or as a showreel with excerpts from several appearances. People should be able to recognise your voice and presence as well as the topic. The text that describes the video is part of it.

### May I publish the recording of my talk myself?

That depends on your agreement with the organiser. Ask before the talk whether a recording will be made and whether you may use it on your own channels, and have the answer confirmed in writing.

### What is a showreel for speakers?

A showreel is a short video that combines excerpts from several appearances. It shows organisers in a few minutes how you perform, and like a recording it can be given suitable text.

### How do I get my talk videos into AI answers?

No one can guarantee it, because Meltwater shows no effect on individual videos and merely recommends text around the video, since the systems read this information. Rely on clear titles, a core message in the first paragraph of the description and a corrected transcript.