Verify any claim · lenz.io
Claim analyzed
Tech“More than 50% of newly created online content produced in the past 12 months was produced with AI assistance or generated by AI.”
Submitted by Vicky
The conclusion
Available evidence does not support a claim that more than half of all newly created online content was made with AI in the last year. The strongest studies showing figures above 50% are limited to narrow slices such as SEO articles or newly indexed webpages, while higher-quality independent research points lower and says most viewed content remains human-made. The statement overgeneralizes and blurs AI-assisted with AI-generated.
Caveats
- The evidence cited for figures above 50% largely covers web articles or indexed pages, not all online content formats and platforms.
- AI-content detection methods have significant uncertainty, so precise global percentages should be treated cautiously.
- AI-assisted content and fully AI-generated content are not the same measure; combining them can make the share appear larger than the evidence supports.
Get notified if new evidence updates this analysis
Create a free account to track this claim.
Sources
Sources used in the analysis
The 2026 AI Index notes rapid adoption of generative AI tools but does not state that most online content is AI-generated. It reports instead on user adoption: “Generative AI reached 53% population adoption within three years, faster than the PC or the internet… the U.S. ranks 24th at 28.3% [adoption].” The report focuses on usage and economic value of generative AI, not on a measured percentage of web content created by AI.
An analysis of one month of browsing data finds that around six-in-ten respondents visited a search page with an AI-generated summary. At the same time, most of the online content that Americans see is still created by humans, but AI-generated material is increasingly visible via search results, recommendations and conversational assistants.
This preprint evaluates several AI detectors on a large corpus of web articles from 2023–2024 and finds substantial but not majority AI authorship for most domains: “Across news, blogs, and forum posts, between 18% and 35% of documents contain at least one segment classified as AI-generated by all three detectors at a conservative threshold… Even under generous assumptions, we estimate that at most 40% of new public web documents in late 2024 contain any AI-generated span.” The authors highlight “large uncertainty and detector bias” and explicitly caution against “claims that a majority of all new web content is AI-written based solely on a single detector’s output.”
The topic page compiles key figures about AI-generated content, including: "Number of AI tool users worldwide from 2020 to 2031 (in millions)" and "Generative artificial intelligence (AI) revenue worldwide from 2020 with forecasts to 2030" as well as statistics on "AI-generated online content (AIGC)". However, the publicly visible overview does not provide a quantified estimate of the share of all newly created online content that is AI-generated or AI-assisted; it focuses on user numbers, revenue, and volumes for specific platforms or use cases rather than a percentage of all new online content.
Key Takeaways: The quantity of AI-generated articles has surpassed the quantity of human-written articles being published on the web. However, the proportion of AI-generated articles has plateaued since May 2024. We find that in November 2024, the quantity of AI-generated articles being published on the web surpassed the quantity of human-written articles.
According to the report, "The global generative AI in content creation market size was calculated at USD 19.75 billion in 2025 and is predicted to increase from USD 24.08 billion in 2026 to approximately USD 143.09 billion by 2035, expanding at a CAGR of 21.90% from 2026 to 2035." It adds that the U.S. market "is expected to reach nearly USD 17.03 billion in 2035" and that North America held around 34% share in 2025. The analysis discusses revenue and growth by region, platform, and use case, but it does **not** provide or endorse any claim that more than 50% of all newly created online content in the past year was AI-assisted; it treats AI as a rapidly expanding segment rather than the dominant source of all new content.
VICE summarizes the Graphite research: "About a year ago, AI began outpacing human writers on the internet. For every one article written by a real-life, blood-bag of a meat puppet, slightly more than one was written by a machine." The piece explains: "The news was broken when Graphite published a study showing that AI-written articles surpassed human-written articles by a small margin in November 2024. 'We find that in November 2024, the quantity of AI-generated articles being published on the web surpassed the quantity of human-written articles,' reads the report." It continues: "After only 12 months, AI-generated articles accounted for nearly half (39%) of articles published."
This commentary discusses the Graphite study: "Now, in what may seem like a tipping point, the digital marketing firm Graphite recently published a study showing that more than 50% of articles on the web are being generated by artificial intelligence." It clarifies the dataset: "It’s important to clarify what’s meant by 'online content,' the phrase used in the Graphite study, which analyzed over 65,000 randomly selected articles of at least 100 words on the web." It notes that "A closer reading of the Graphite study shows that the AI-generated articles consist largely of general-interest writing: news updates, how-to guides, lifestyle posts, reviews and product explainers."
Reporting on an Europol document, the article notes: “Experts estimate that as much as 90 percent of online content may be synthetically generated by 2026, the report warned, adding that synthetic media ‘refers to media generated or manipulated using artificial intelligence.’” This is framed as a projection about future online content share rather than a measured current value.
According to an analysis by marketing agency Graphite, roughly half of all new online articles, blogs, and listicles are now produced mainly with the help of AI. However, that share has shown almost no growth since early 2025. The researchers concluded that AI is now producing roughly as much content as humans, though there is still no sign of machine-generated material fully dominating the web.
The Living Library quotes the VICE/Graphite reporting: "About a year ago, AI began outpacing human writers on the internet. For every one article written by a real-life, blood-bag of a meat puppet, slightly more than one was written by a machine." It continues: "The news was broken when Graphite published a study showing that AI-written articles surpassed human-written articles by a small margin in November 2024. 'We find that in November 2024, the quantity of AI-generated articles being published on the web surpassed the quantity of human-written articles,' reads the report." The snippet also notes: "After only 12 months, AI-generated articles accounted for nearly half (39%) of articles published."
Discussing content generation in enterprises, IBM reports: "More than 50% of marketing and customer-experience leaders in our survey say their organizations now use generative AI tools to assist in content creation, including emails, website copy, and social posts." At the same time, IBM notes that "human review and editing remains the norm" and that this survey "captures usage inside medium and large organizations, not the overall distribution of all content published on the public Internet." It emphasizes that the findings "should not be interpreted as meaning that most online content is AI-generated; rather, AI is now involved in some part of the workflow in many organizations."
We analyzed 900,000 newly created web pages in April 2025 and found that 74.2% of them contained AI-generated content. Each page was categorized according to the percentage of the page our model detected as being AI-generated… 25.86% showed moderate AI use (11%–40% of the page content was categorized as AI), 20.50% showed substantial AI use (41%–70%), and 15.51% showed dominant AI use (71%–99%). I was surprised that almost three-quarters of the pages we analyzed included AI content.
According to a study published by Graphite.io in 2025, the volume of AI-generated articles surpassed the volume of manually written articles by November 2024. In less than 12 months, almost 40% of articles published on the web were already generated by AI. In just two years, AI has become the leading author of the internet. While the proportion of automatically generated articles appears to have stabilized, the impact of this shift is profound and irreversible.
Most published estimates that a majority of new online content is AI-generated are based on specific datasets, such as web pages discovered by SEO crawlers or particular article types, and use AI detectors that have known error rates and biases. There is no globally accepted, independently audited metric that measures what fraction of all newly created online content across platforms, languages and formats is produced with AI assistance.
Summarizing other studies, the article states: “Ahrefs analyzed nearly a million new web pages published in April 2025 and found that 74.2 percent contained detectable AI generated content… Graphite, an SEO firm, examined more than 60,000 new articles from 2020 through 2025 and found that by late 2024, more than half of new English language articles were primarily AI written. That percentage grew into 2025… One widely cited analysis estimates that about 57 percent of all online text has been generated or translated using AI tools.”
The article claims: "More than 80% of social media content recommendations rely on AI algorithms. An estimated 70% of social media images may involve AI tools like Midjourney or DALL·E, reflecting their growing role in content creation across platforms." It further states in an "AI‑Generated Content Statistics" section that "71% of images shared on social media are now AI-generated" and that "LinkedIn shows over 50% of long-form posts likely created by AI, about 54%." These figures, if taken at face value, imply that in certain social media contexts AI-assisted or AI-generated material could exceed 50% of new content; however, the article does not provide methodology or a global cross‑platform estimate of all newly created online content.
A discussion thread references the Graphite research: "The study said ~50% of articles published in 2025 were produced primarily by AI, according to their AI detection tool: 'We classify an article as AI-generated if more than 50% of its text is likely to have been generated by AI.'" Participants discuss the implications of the finding that "~50% of articles" in the sampled dataset appear primarily AI-generated.
What do you think of the claim?
Your challenge will appear immediately.
Challenge submitted!
Continue your research
Verify a related claim next.
Expert review
3 specialized AI experts evaluated the evidence and arguments.
Expert 1 — The Logic Examiner
The claim asserts that more than 50% of ALL newly created online content was AI-assisted or AI-generated in the past 12 months. The supporting evidence (Sources 5, 7, 8, 13) is largely drawn from SEO-crawled articles and newly indexed webpages—a narrow, commercially motivated slice of online content—and cannot be logically extended to the full universe of online content including social media posts, emails, videos, audio, and user-generated content across all languages and platforms; Source 3 explicitly caps AI-generated spans at 40% of public web documents with large uncertainty, Source 2 states most content Americans see is still human-created, and Source 15 confirms no globally accepted metric exists for the broader claim. The Opponent's fallacy-of-composition argument is logically sound: the Proponent's inferential leap from 'majority of SEO-crawled articles' to 'majority of all newly created online content' is an overgeneralization, and the IBM source (Source 12) explicitly disclaims the interpretation the Proponent draws from it; therefore the claim as stated—covering ALL online content—is not logically supported by the evidence, making it misleading rather than true, though there is genuine evidence that AI-generated articles specifically may have crossed 50% in certain measurable domains.
Expert 2 — The Context Analyst
The claim uses broad language ('online content') but the supporting evidence is almost entirely limited to SEO-crawled articles and newly indexed webpages (Sources 5, 13), which represent a commercially motivated, narrow slice of all online content. The full universe of newly created online content includes social media posts, emails, videos, audio, user-generated content across all languages and platforms—none of which are captured by the Graphite or Ahrefs studies. Source 3 (arXiv) explicitly finds 'at most 40%' of new public web documents contain any AI-generated span, Source 2 (Pew) states most content Americans see is still human-created, and Source 15 confirms there is no globally accepted metric for all content formats. The claim's framing implies a universal majority that the evidence does not support; even the most favorable studies are domain-specific and methodologically limited, making the overall impression created by the claim misleading.
Expert 3 — The Source Auditor
The highest-authority sources in this pool — Stanford HAI (Source 1, high-authority), Pew Research Center (Source 2, high-authority), and the arXiv preprint (Source 3, high-authority) — all refute or fail to confirm the claim. Pew explicitly states 'most of the online content that Americans see is still created by humans,' and the arXiv study estimates 'at most 40% of new public web documents in late 2024 contain any AI-generated span,' cautioning against majority-AI claims. The supporting sources are predominantly secondary reports (Sources 7, 8, 10, 11, 14, 16, 17, 18) that largely recycle the Graphite.io study (Source 5), which is itself an SEO marketing firm's analysis of a narrow corpus of crawled articles — a dataset that cannot be extrapolated to all online content formats. The Ahrefs study (Source 13) is methodologically limited to newly indexed webpages and uses AI detection with known error rates. The claim uses 'online content' broadly, but the supporting evidence is confined to a specific, commercially biased slice of web articles; the most reliable independent sources do not confirm the >50% threshold for all newly created online content, making the claim misleading rather than true.
Expert summary
The arguments
Two AI advocates debated this claim using the research gathered.
Argument for
Multiple independent measurements of newly created web text already clear the “majority AI-involved” threshold: Graphite reports AI-generated articles surpassing human-written articles by November 2024 (Source 5), and Ahrefs' large crawl of 900,000 newly created pages in April 2025 finds 74.2% contained AI-generated content—directly implying that more than half of new online content had AI assistance within the past year (Source 13). This is reinforced by downstream reporting that characterizes the Graphite result as “more than half of new articles” being AI-written (Sources 7 and 8) and by enterprise workflow evidence that AI is now used to assist content creation in a majority of marketing/CX organizations (Source 12), making it more likely than not that the aggregate share of newly created online content with AI assistance exceeds 50%.
The Proponent's central reliance on Source 13 (Ahrefs) and Source 5 (Graphite.io) commits a fallacy of composition: both studies are explicitly limited to SEO-crawled articles and newly indexed webpages, which represent a narrow, commercially motivated slice of online content, and cannot be extrapolated to the full universe of newly created online content encompassing social media posts, emails, videos, audio, and user-generated material across all languages—a limitation Source 15 explicitly confirms. Furthermore, the Proponent's appeal to Source 12 (IBM) as corroborating enterprise-level 'AI assistance' is a non sequitur, as IBM itself explicitly states that its survey 'should not be interpreted as meaning that most online content is AI-generated' and that its findings capture only medium and large organizations, not the overall distribution of all publicly published content—making the Proponent's inferential leap from organizational tool adoption to majority AI content production logically unsound.
Argument against
The claim that more than 50% of all newly created online content is AI-generated is fundamentally undermined by the narrow scope of supporting evidence: Source 3 (arXiv) explicitly finds that 'at most 40% of new public web documents in late 2024 contain any AI-generated span' and cautions against majority-content claims, while Source 2 (Pew Research Center) directly states that 'most of the online content that Americans see is still created by humans.' The studies cited in support, such as Source 5 (Graphite.io) and Source 13 (Ahrefs), are limited to SEO-crawled articles and specific webpage types using AI detectors with known error rates and biases, as Source 15 (LLM Background Knowledge) confirms, meaning they cannot be extrapolated to all online content formats including social media posts, videos, emails, and user-generated content across all languages and platforms.
The Opponent's argument commits a scope fallacy by treating Source 3's detector-based estimate for “public web documents” in 2023–2024 as dispositive for “all newly created online content” in the past 12 months, even though Source 3 itself emphasizes domain variation and uncertainty and therefore cannot negate majority-AI involvement when other large-scale measurements of newly created pages/articles find majority AI presence (Sources 13 and 5). The Opponent also misuses Source 2 by substituting “what Americans see” (a consumption/exposure claim shaped by feeds and search) for “what is newly created,” while dismissing the best-available creation-side measurements (Sources 13 and 5) on generic detector-bias grounds from Source 15 without showing that any plausible correction would flip those results below 50%.