Video vs text content statistics at a glance
| Category | Statistic | Source |
|---|---|---|
| Preference | 55% prefer to read the news, 31% prefer to watch it, 15% prefer to listen, across 48 markets | Reuters Institute, June 2025 |
| Behaviour | 77% of people consume online news video each week; in 45 of 48 markets that beats broadcast TV news | Reuters Institute, June 2026 |
| Behaviour | 91.1% of adult internet users watched some online video in the past seven days | DataReportal Digital 2026, October 2025 |
| Preference | 34% of US adults prefer TV for news, 21% news sites and apps, 14% social media, 5% print | Pew Research Center, 2025 |
| Attention | The average YouTube visit runs 19 minutes 10 seconds worldwide; the average Wikipedia visit runs 3 minutes 14 seconds | Similarweb, August 2026 |
| Comprehension | In a 48-person study, novices comprehended better with multimedia while experts ignored the video and used the text | Song et al., 2023 |
| Comprehension | Adults read English non-fiction at 238 words per minute, from a meta-analysis of 190 studies | Brysbaert, 2019 |
| AI search | YouTube is the most cited domain in Google AI Overviews at about 23.3%, ahead of Wikipedia at 18.4% | Surfer, October 2025 |
| AI search | YouTube is 0.99% of all ChatGPT citations; Reddit is 2.4% | Profound, February 2026 |
| AI search | Long-form video is 94% of AI citations to YouTube; Shorts are 5.7% | Otterly.ai, March 2026 |
| Cost | Writing an average 1,312-word blog post takes 3 hours 20 minutes | Orbit Media survey of 1,042 bloggers, 2026 |
| Cost | Almost 40% of companies spent under $5,000 on video in 2025; just over 30% spent more | Wistia vendor data, 2026 |
| Shelf life | A third of webinars still draw plays three months after the live event | Wistia vendor data, 2026 |
| Distribution | 54% use social media and video networks for news, above traditional sources for the first time | Reuters Institute, 2026 |
Do people prefer video or text content?
People say they prefer text: 55% of respondents across 48 markets prefer to read the news, 31% prefer to watch it and 15% prefer to listen, in the Reuters Institute Digital News Report 2025. That is a stated preference, and it is the most misquoted figure in this argument.
| Statistic | Source |
|---|---|
| Across 48 markets, 55% prefer to read the news, 31% prefer to watch and 15% prefer to listen. | Reuters Institute, June 2025 |
| In India, Mexico and the Philippines, more people now say they prefer to watch the news than read it. Norway, Germany and the UK stay firmly text-first. | Reuters Institute, June 2025 |
| 34% of US adults prefer television for news, 21% prefer news websites or apps, 14% prefer social media, 10% prefer search engines and 5% prefer print. | Pew Research Center, August 2025 |
| Among US adults aged 18 to 29, 31% prefer social media for news against 12% who prefer television. | Pew Research Center, August 2025 |
| Fewer than 1% of US adults name an AI chatbot as their preferred news platform, though 9% use one often or sometimes. | Pew Research Center, August 2025 |
Stated preference is the weakest evidence in this field and the most quoted. Ask a vendor survey whether consumers want more video from brands and the answer is always yes; nobody has ever told a survey they want less of anything. Reuters asks the public rather than the sellers, in 48 markets every year. Text preference is real, and it erodes fastest at the bottom of the age range.
What do people watch and read in practice?
Measured behaviour tilts far harder to video than stated preference does: 77% of people worldwide consume online news video each week, against the 31% who say they prefer watching to reading. That 46-point gap is the most useful number on this page.
| Statistic | Source |
|---|---|
| 77% of people globally consume online news video each week, and in 45 of 48 markets more people watch online news video than broadcast TV news. Total sample 97,520. | Reuters Institute, June 2026 |
| Weekly consumption of any online news video rose from 67% in 2020 to 75% in 2025. Social video news rose from 52% to 65% over the same five years. | Reuters Institute, June 2025 |
| In the US, weekly news video consumption rose from 55% in 2021 to 72% in 2025. Of that, 61% happens on third-party platforms and 29% on news sites and apps. | Reuters Institute, June 2025 |
| 91.1% of adult internet users watched some online video in the past seven days, and they spend more than two and a half hours a day with online video and social feeds. Streaming is 50.4% of global TV watch time and 59.4% in the US. | DataReportal Digital 2026, October 2025 |
| Worldwide in August 2026, the average YouTube visit lasted 19 minutes 10 seconds, against 9 minutes 54 seconds on Google and 3 minutes 14 seconds on Wikipedia. A visit is a session on the domain, so YouTube's figure includes background listening and autoplay. | Similarweb, August 2026 |
| 38% of US adults regularly get news on Facebook, 35% on YouTube, 20% on Instagram and 20% on TikTok. 55% of TikTok users get news there, up from 22% in 2020. | Pew Research Center, August 2025 |
People prefer reading and they watch anyway, because video is what the feed hands them. Preference is what someone picks when both options sit in front of them; weekly reach is what the feed pushes, and it is the number to plan against. The 5.9x gap between an average YouTube visit and an average Wikipedia visit shows which medium holds a session, even if some of those YouTube minutes run in a background tab.
Which format do people understand and remember better?
Neither format wins outright on comprehension: expertise decides it. In a 48-participant study published in 2023, novices given multimedia read faster, finished tasks faster and comprehended better, while experts ignored the video and worked from the text.
| Statistic | Source |
|---|---|
| Among 48 participants, novices given multimedia showed better comprehension (p=0.0161), faster task completion (p=0.0156) and reduced reading time (p=0.0427). Experts largely ignored the multimedia and used the text. | Song, Healey, Siu, Wigington and Stasko, 2023 (DATED) |
| Adults read English non-fiction silently at 238 words per minute and fiction at 260, from a meta-analysis of 190 studies covering 18,573 participants. | Brysbaert, Journal of Memory and Language, 2019 (DATED) |
| Nielsen Norman Group's usability testing found users turned to video when a process was long and multistep or the topic unfamiliar, and skipped it otherwise. Videos at the bottom of a page were often ignored. | Nielsen Norman Group, 2020 (DATED, 8 participants) |
| 26% of views on Nielsen Norman Group's own video channel use subtitles, a far wider audience than hearing loss alone accounts for. | Nielsen Norman Group, 2020 (DATED) |
I have marked these four sources dated because nobody has replaced them. Brysbaert's reading rate meta-analysis is still the only hard anchor for how fast text gets consumed, and it is seven years old.
The expert-novice split is the finding I would build a content plan on. Text is random access: a reader who knows the subject skims to the paragraph that matters and leaves. Video is sequential, so it serves the person who does not yet know what to look for. That is why a setup tutorial works as video and a pricing comparison does not, and why the 26% subtitle figure matters: a quarter of your video audience quietly turns your video back into text.
Where does each format get found in 2026?
Discovery splits three ways: organic search still favours indexed text pages, social feeds favour video by default, and AI answer engines disagree sharply about which to cite.
How much of Google's AI Overview citation pool goes to video?
YouTube is the single most cited domain in Google AI Overviews at about 23.3% of citations, which still leaves 3.3 citations from other domains for every YouTube citation.
| Statistic | Source |
|---|---|
| YouTube takes about 23.3% of Google AI Overview citations, ahead of Wikipedia at 18.4% and Google.com at 16.4%, across 36 million AI Overviews and 46 million citations sampled March to August 2025. | Surfer, October 2025 |
| YouTube's citation share reaches about 93% in gaming, 39.1% in SEO, 32.4% in ecommerce and 23.5% in travel. | Surfer, October 2025 |
| The share of Google queries triggering an AI Overview moved from 6.49% in January 2025 to 24.61% in July and 15.69% in November, across more than 10 million keywords. | Semrush, 2025 |
Do AI chatbots cite video the way Google does?
No, and the gap is enormous: YouTube takes 0.99% of all ChatGPT citations against 23.3% of Google AI Overview citations, so its citation share is about 23.5 times larger in Google's engine.
| Statistic | Source |
|---|---|
| YouTube accounts for 0.99% of all ChatGPT citations and Reddit for 2.4%, from roughly 700,000 social platform citations in US English responses collected October to December 2025. Profound corrected this analysis in February 2026 after its privacy filter stripped YouTube links as false positives. | Profound, February 2026 |
| 85% of ChatGPT's YouTube citations point to a specific video rather than a channel or playlist. | Profound, February 2026 |
| Of all YouTube citations observed across six AI engines, 38.7% came from Perplexity, 36.6% from Google AI Overviews, 19.6% from Google AI Mode and 4.4% from ChatGPT. Sample: more than 100 million citations over 30 days. | Otterly.ai, March 2026 |
| Long-form video is 94% of AI citations to YouTube. Shorts are 5.7% and playlists, channels and livestreams 0.3%. | Otterly.ai, March 2026 |
| Views, likes and subscriber counts show near-zero correlation with citation frequency (r about -0.03 for all three). | Otterly.ai, March 2026 |
| 10% of people worldwide use AI chatbots for news each week, up from 7% a year earlier, rising to 16% of under-35s. | Reuters Institute, June 2026 |
Which format gets distribution in social feeds?
Social and video networks passed traditional sources as a news channel for the first time in 2026, at 54%, and the growth is all video-native.
| Statistic | Source |
|---|---|
| 54% now use social media and video networks for news, above traditional sources for the first time. 30% call them their main news source, up from 22% five years earlier, and among 18 to 24 year olds 52% do, while 56% never regularly read newspapers. | Reuters Institute, June 2026 |
| Video on publishers' own sites and apps fell 5 percentage points year over year to 23%. | Reuters Institute, June 2026 |
The AI citation split is the most actionable thing here and almost nobody acts on it. Google's AI Overviews treat a YouTube video as a first-class source; ChatGPT treats it as a rounding error. If your buyers research in ChatGPT, a well-structured page does the work and video does close to nothing. If they use Perplexity or Google, the opposite. Otterly's near-zero correlation between views and citations should also end the habit of judging a video by its view count when the goal is being quoted. One caveat on my 23.5x figure: Surfer and Profound are different vendors measuring different engines, query mixes and months, and each counts citations its own way, so treat it as an order of magnitude, not a decimal. Our content decay statistics page covers what happens to that traffic afterwards.
What does video cost to make compared with written content?
Video costs more money and text costs more of the resource nobody tracks: 40% of companies spent under $5,000 on video in 2025, while one blog post takes 3 hours 20 minutes of a writer's time.
| Statistic | Source |
|---|---|
| 76% of companies make at least one video a month. Almost 40% spent under $5,000 on video in 2025 and just over 30% spent more than $5,000. Wistia vendor data from a survey of 900-plus professionals plus 13 million videos and 79 million hours of viewing. | Wistia, 2026 |
| Two-thirds of the 382 marketers surveyed spend under $10,000 a year on video creation and 36% spend under $5,000. 69% named production time as a major challenge. | Content Marketing Institute survey of 382 marketers, August 2023 (DATED) |
| 67% of marketers said video had become more important to their business, while 7% said they used it to its full potential. | Content Marketing Institute, August 2023 (DATED) |
| Writing an average blog post takes 3 hours 20 minutes, down from 4 hours 10 minutes in 2022. 92.4% of bloggers now use AI in their process, and 13.9% report strong results, an all-time low. | Orbit Media survey of 1,042 bloggers, 2026 |
| Vidyard users averaged 37 videos each in 2024, up 241% year over year, and total creation rose 88%. Vendor data covering 1 January to 15 December 2024. | Vidyard Business Video Benchmarks, 2025 |
| Creation of videos over 20 minutes grew 420% year over year, education up 436% and consulting up 280%. Vendor data. | Vidyard Business Video Benchmarks, 2025 |
The cost comparison everyone runs puts a video budget against a writer's salary, and those sit in different columns. Wistia and Vidyard both measure their own paying customers, which is why the volume growth looks so big; people who buy video software make video. The direction is what counts: video spend stays under $5,000 for most companies while writing time per post fell 50 minutes since 2022, with AI now in 92.4% of blogging workflows. Cheap text is about to stop being a moat.
How long does each format keep working?
Video holds value longer than its first week suggests: a third of webinars still draw plays three months after the live event, while an article's engaged time barely moves with length.
| Statistic | Source |
|---|---|
| A third of webinars still draw plays three months after the live event. Replays running 31 to 45 minutes earned more than double the engagement of replays under 30 minutes, against a typical 15% to 20% on-demand engagement rate. Wistia vendor data. | Wistia, 2026 |
| Articles around 6,000 words averaged 80 seconds of engaged time, with the best and worst performers separated by roughly 6 seconds. Most articles across the network run under 500 words. | Chartbeat, millions of articles, January 2019 to April 2022 (DATED) |
| Long-form video is 94% of AI citations to YouTube, and citation frequency has no meaningful correlation with views, so a low-view reference video can keep earning citations indefinitely. | Otterly.ai, March 2026 |
Chartbeat's engaged time data predates AI Overviews, so read it as a floor on attention rather than a current benchmark. Adding words does not reliably add reading time.
Shelf life is where video quietly wins and almost nobody measures it. A webinar recording that still gets plays in month three is doing the job a pillar page does, and the Otterly correlation means a 400-view reference video can still be the source an AI engine reaches for. The catch is that video decay stays invisible unless you instrument it. There is no equivalent of watching a page slide from position 4 to position 11. Our short-form vs long-form content statistics page covers length effects in both mediums.
How we calculated the original numbers
Four figures here appear nowhere else. The arithmetic is below, so you can check it or change the inputs.
- 23.5x: video's citation edge in Google AI Overviews over ChatGPT. Surfer put YouTube at about 23.3% of Google AI Overview citations across 36 million overviews and 46 million citations sampled March to August 2025. Profound's corrected February 2026 analysis put YouTube at 0.99% of all ChatGPT citations, drawn from roughly 700,000 social platform citations in US English responses collected October to December 2025. 23.3 / 0.99 = 23.5. Both figures are one domain's share of all citations in one engine, which is what makes the comparison possible, but they come from two vendors with different query mixes, months and citation rules. Treat 23.5x as an order of magnitude, or just take the two facts on their own: 23.3% and 0.99%.
- 3.3 to 1: other domains per YouTube citation in an AI Overview. Both shares come from the same Surfer citation base of 46 million citations. If YouTube is 23.3%, every other domain together is 76.7%. 76.7 / 23.3 = 3.29. YouTube leads the domain list and non-YouTube domains still supply roughly three citations in four.
- 46 points: the preference-behaviour gap. 31% of people across 48 markets say they prefer to watch the news (Reuters Institute, 2025). 77% consume online news video every week (Reuters Institute, 2026). 77 - 31 = 46. Two different questions, one on preference and one on weekly reach, asked a year apart in the same research programme. The gap is the point: weekly reach runs roughly 2.5 times stated preference.
- 5.9x: time per visit, YouTube against Wikipedia. Similarweb's worldwide ranking for August 2026 gives YouTube an average visit of 19 minutes 10 seconds, or 19.17 minutes, and Wikipedia 3 minutes 14 seconds, or 3.23 minutes. 19.17 / 3.23 = 5.93. These are destination domains, not content formats, and a visit means different things on each: a YouTube session can run in a background tab or roll on through autoplay, while a Wikipedia session ends when the reader has the answer. It sizes how long a medium holds a session, not how long one asset holds one person.
What the 2026 numbers say
Preference and behaviour point in opposite directions, and behaviour is winning. A stable 55% still say they would rather read. Weekly video reach hit 77% anyway, social and video networks passed traditional sources at 54%, and 56% of 18 to 24 year olds have never regularly read newspapers. Plan a 2027 content mix off the preference number and you are planning for an audience that is aging out.
The AI citation split is the biggest unexploited asymmetry in the data. Google's AI Overviews cite YouTube more than any other domain at 23.3%. ChatGPT cites it at 0.99%. Same web, two engines that disagree by more than an order of magnitude. Until that converges, format choice is a distribution choice: pick based on where your buyers ask their questions.
Text is cheap and about to get cheaper, which makes it a worse moat. Writing time per post fell from 4 hours 10 minutes to 3 hours 20 minutes in four years while 92.4% of bloggers adopted AI, and the share reporting strong results hit an all-time low of 13.9%. Video still costs money and time, and that friction is most of its defensibility now. Our AI-generated video statistics page tracks how fast that last barrier is falling. Browse the rest of the statistics library for the neighbouring numbers.
FAQ
Do people prefer video or text content?
55% of people across 48 markets say they prefer to read the news against 31% who prefer to watch it (Reuters Institute Digital News Report 2025). Behaviour goes the other way: 77% consume online news video every week (Reuters Institute, 2026). Stated preference favours text, weekly reach favours video, and the 46-point gap is the honest answer.
Is video better than text for learning something?
It depends on the learner. In a 48-participant study by Song and colleagues (2023), novices given multimedia comprehended better (p=0.0161) and finished tasks faster (p=0.0156), while experts ignored the video and used the text. Nielsen Norman Group (2020) found users switched to video for long multistep processes and skipped it otherwise.
How much of Google's AI Overview citations go to video?
About 23.3% of Google AI Overview citations go to YouTube, the largest share of any domain, ahead of Wikipedia at 18.4% (Surfer, October 2025, from 36 million overviews). That leaves 3.3 citations from other domains for every YouTube citation, so non-YouTube domains still supply roughly three citations in four.
Does ChatGPT cite YouTube videos?
Rarely. YouTube accounts for 0.99% of all ChatGPT citations, against 2.4% for Reddit, in Profound's corrected February 2026 analysis of roughly 700,000 social platform citations. YouTube's citation share is therefore about 23.5 times larger in Google AI Overviews than in ChatGPT, though the two figures come from different vendors and months.
How long does it take to make a blog post versus a video?
Writing an average 1,312-word blog post takes 3 hours 20 minutes (Orbit Media, 2026, 1,042 bloggers surveyed). Video has no comparable public time benchmark, but 69% of 382 marketers named production time a major challenge (Content Marketing Institute, August 2023) and almost 40% of companies spent under $5,000 on video in 2025 (Wistia vendor data).
Do more views mean an AI engine will cite your video?
No. Views, likes and subscriber counts all show near-zero correlation with citation frequency, at r of about -0.03, in Otterly.ai's March 2026 study of more than 100 million AI citations. Long-form video takes 94% of AI citations to YouTube while Shorts take 5.7%, so structure and reference value matter more than popularity.
Sources
- Reuters Institute for the Study of Journalism, Digital News Report 2026 executive summary (June 2026)
- Reuters Institute for the Study of Journalism, Digital News Report 2025 executive summary (June 2025)
- Pew Research Center, News Platform Fact Sheet (August 2025)
- Pew Research Center, Social Media and News Fact Sheet (August 2025)
- DataReportal, Digital 2026 Global Overview Report (October 2025)
- Similarweb, Top Websites Ranking (August 2026)
- Surfer, AI Citation Report (October 2025)
- Profound, How ChatGPT cites social media (February 2026)
- Otterly.ai, YouTube AI Citation Study 2026 (March 2026)
- Semrush, AI Overviews Study (2025)
- Orbit Media, Blogging Statistics and Trends (2026)
- Wistia, State of Video Report (2026)
- Vidyard, Video in Business Benchmark Report (2025 report, 2024 data)
- Content Marketing Institute, Why Most Video Content Fails To Reach Its Full Potential (August 2023)
- Nielsen Norman Group, Videos as Instructional Content (January 2020)
- Hayeong Song, Jennifer Healey, Alexa Siu, Curtis Wigington and John Stasko, Experts prefer text but videos help novices (2023)
- Marc Brysbaert, How many words do we read per minute? A review and meta-analysis of reading rate, Journal of Memory and Language (2019)
- Chartbeat, Is there an optimal article length? (2022)
Methodology. Numbers were collected in September 2026 and checked against the original publisher's page, not against statistics roundups. Vendor claims without a stated sample or method were excluded, and vendor data from Wistia, Vidyard and Surfer is labelled as vendor data in the sentence that uses it. Stated preference and measured behaviour are kept apart throughout, because conflating them is the most common error in this topic. Sources older than 24 months are marked DATED. Last updated 22 September 2026.