AI and Writing Statistics 2026: What the Research Actually Says

You want one solid number for your next post about AI and writing. You search, and you get "90% of online content will be AI by 2026", "57% of the internet is AI-generated" and "AI detectors are 99% accurate". None of those holds up when you trace it back to the original source.

This page collects the AI writing statistics that do hold up. We checked every figure against its original survey, study, company announcement or court filing on 19 October 2026, and we give the sample, the date and a link for each one so you can cite the source directly. Where a figure comes from a company with something to sell, we say so.

Key AI Writing Statistics for 2026

• 49% of US adults now use AI chatbots, up from 33% in 2024 (Pew Research Center, June 2026).
• 42% of fiction authors use AI at least sometimes, against 61% of other professional writers (Gotham Ghostwriters and Josh Bernoff, November 2025).
• 51% of published UK novelists think AI is likely to replace their work entirely, and 39% say it has already cut their income (University of Cambridge, November 2025).
• 74% of authors who use AI don't tell their readers (BookBub, 2025), while 56% of Americans want to be told about any AI contribution to a book (YouGov, August 2025).
• About 40% of work-related ChatGPT messages are about writing, and two-thirds of those ask it to edit the user's own text (NBER working paper, September 2025).
• 49.9% of new English-language web articles are mainly AI-generated, a share that has stayed near 50% since early 2025 (Graphite, May 2026).
• Google users click a result 8% of the time when an AI summary appears, against 15% without one (Pew Research Center, July 2025).
• AI detectors flagged 61.22% of essays by non-native English writers as AI-written (Stanford, 2023).
• Anthropic's $1.5 billion settlement with authors, about $3,000 per book, got final court approval on 20 July 2026 (Authors Guild).

How Many People Use AI Chatbots?

About half of American adults now use AI chatbots. Pew Research Center's February 2026 survey of 5,119 US adults found 49% use them, up from 33% in 2024, and 24% use one at least once a day. ChatGPT is far ahead: 44% of US adults use it, against 24% for Gemini, 17% for Copilot, 14% for Meta AI, 8% for Grok and 6% for Claude.

• Use falls sharply with age. 66% of 18 to 29 year olds use chatbots, against 61% of 30 to 49s, 42% of 50 to 64s and 23% of people 65 and over (Pew, June 2026).
• Looking things up is the top use. 42% of Americans use chatbots to find information, 38% of employed adults use them for work tasks and 13% for news (Pew, June 2026).
• ChatGPT reached 900 million weekly users in February 2026, with 50 million paying subscribers, according to OpenAI as reported by TechCrunch. This is a company figure, not an audited one.
• Writing is the biggest work use of ChatGPT. About 40% of work-related messages are writing tasks, and about two-thirds of writing requests ask it to edit, critique, translate or summarise text the user supplied. Writing's share of all ChatGPT use fell from 36% to 24% between July 2024 and July 2025 as people used it more for looking things up (Chatterji and others, NBER Working Paper 34255). The paper uses OpenAI's own data and has not been peer reviewed.
• Weekly use of generative AI nearly doubled in a year across Argentina, Denmark, France, Japan, the UK and the US, from 18% to 34% (Reuters Institute, October 2025).

How Many Authors Use AI?

Fewer than half, and the answer depends on who is asked. Across the most recent surveys, between a third and just under half of authors use AI in any part of their work, mostly for research and admin rather than writing prose. There is no single "authors" figure, so quote the survey that matches your readers.

What authors use it for, and how they feel about it:

• Research comes first. 81% of authors who use AI in BookBub's survey use it for research. Among BookBub's non-users, 84% avoid it because they consider it unethical.
• Very little AI text gets published. Only 7% of professional writers in the Gotham survey had published AI-generated text, although 63% of AI users generate text that they then edit. The most common tasks were suggesting titles (72% of users), search (71%) and brainstorming (68%).
• Fiction authors who don't use AI are united against it. In the same survey, 100% of fiction authors who don't use AI said training AI on their work is unfair, and only 1% saw AI as positive for the profession.
• Novelists are worried about money. 39% of UK novelists say AI has already cut their income, 85% expect it to push their future income down, and 59% say their work was used to train AI without permission or payment (Cambridge).
• Translators and illustrators were hit first. By January 2024, 36% of translators and 26% of illustrators in a Society of Authors survey of 787 UK creators had already lost work to generative AI (Society of Authors, April 2024).
• Few authors tell readers. 74% of authors who use AI don't disclose it (BookBub). 58% of ALLi members have no written AI policy and 10% publish one on their website.

How Many Bloggers and Content Marketers Use AI?

Almost all of them. Surveys of bloggers and marketers in 2025 and 2026 put AI use above 90%, much higher than among authors. The difference is in how the output is used: most marketers still edit what AI produces, and few report better results because of it. If you are weighing how much to write, our study of top-ranking pages found the median page on Google's first page is 997 words.

• 92% of bloggers use AI, and the average blog post now takes 3 hours 20 minutes to write, down from 4 hours 10 minutes in 2022 (Orbit Media, September 2026; 1,042 content marketers; Orbit is a web agency). Orbit found that AI use "correlates with nothing" in the results bloggers report.
• 93% of marketers use AI to help create content, up from 87% a year earlier. 7% of AI-using marketers say more than three-quarters of what they publish is unedited AI output (Ahrefs, September 2026; 301 respondents; Ahrefs sells SEO software).
• 89% of B2B marketers use AI for content, but only 39% say it improved content performance, and 12% say quality got worse (Content Marketing Institute, October 2025; 1,015 B2B marketers).
• 82% of journalists use at least one AI tool, up from 77%; ChatGPT is the most used at 47% (Muck Rack, March 2026; 897 journalists; Muck Rack sells PR software).
• About 9% of new US newspaper articles are partly or fully AI-generated, and only 5 of 100 flagged articles disclosed it (Russell and others, 2025 preprint; 186,000 articles from 1,500 newspapers, scored with one AI detector).
• Four in ten freelance and small-agency writers say AI has reduced their income, and 73% of professional writers expect fewer opportunities over the next five years (Gotham Ghostwriters and Josh Bernoff).

How Much of the Internet Is Written by AI?

About half of newly published English articles, by the best current measure. Graphite's May 2026 study of 55,400 random English web articles found 49.9% were mainly AI-generated in early 2026, and the share has stayed close to 50% since early 2025 rather than rising further. Almost all new pages contain at least some AI text, but very few are pure AI.

• Half, and flat. AI articles were 49.6% of new articles in Q1 2025, peaked at 50.9% in Q4 2025 and were 49.9% in Q1 2026 (Graphite, May 2026). Graphite averaged three detectors and measured a false-positive rate of 1.36% to 1.84% on articles written before ChatGPT existed. Graphite is an SEO agency, and the study counts what gets published, not what people read.
• Mostly mixed. 74.2% of 900,000 new web pages found in April 2025 contained some AI-generated text, but only 2.5% were purely AI and 25.8% purely human (Ahrefs, May 2025; in-house detector).
• LinkedIn is the extreme case. 81.2% of 5,000 long LinkedIn posts were classified as likely AI in July 2026 (Originality.ai), a company that sells the detector it used.
• Amazon caps uploads. Kindle Direct Publishing limits each account to three new titles a day, a rule added in September 2023 as AI-generated books flooded the store (Authors Guild).
• Amazon's disclosure rule is narrower than many authors think. KDP requires disclosure of AI-generated text, images or translations, even after "substantial edits", but says of AI-assisted work such as editing and brainstorming: "It is not necessary to inform us." Readers aren't shown the disclosure (KDP Content Guidelines, checked 19 October 2026).

What Do Readers Think About AI-Written Content?

Readers want to know, and most say they like AI-written work less when they find out. That is the consistent pattern across the US and UK surveys below, and it sits uncomfortably next to the 74% of AI-using authors who don't disclose.

• 76% of Americans say it is very or extremely important to be able to tell AI content from human content, but 53% aren't confident they can (Pew, September 2025; 5,023 US adults).
• 53% of Americans think AI will make people worse at thinking creatively; 16% think it will help (Pew, September 2025).
• 56% of Americans want to be told about any AI contribution to a book, 28% say AI shouldn't be used at any stage of making a book, and 61% say they would feel less fulfilled if they learned a book was written by AI (YouGov, August 2025; 1,000 Americans). 40% accept AI for line or copy editing.
• 72% of Britons would be less likely to read an article if they knew AI helped write it, against 3% who would be more likely, and 91% want disclosure when an article is published without a human editor (YouGov, March 2026; 2,014 British adults).
• Only 12% of people are comfortable with news made entirely by AI, against 62% for news made entirely by humans (Reuters Institute, October 2025).
• In blind tests, readers can prefer AI that imitates an author. When AI models were fine-tuned on an author's books, both expert and general readers preferred the AI passages over ones by expert human writers, and detectors flagged the fine-tuned text only 3% of the time. With ordinary prompting, expert readers strongly preferred the human writing (Chakrabarty, Ginsburg and Dhillon, 2025 preprint; short excerpts, not whole books).

How Accurate Are AI Detectors?

Not accurate enough to judge one writer's work. Independent tests have found every major detector making serious errors, and the worst errors fall on people who write in English as a second language. Vendor claims of 98% or 99% accuracy come from the companies selling the tools.

• OpenAI withdrew its own detector. It caught 26% of AI-written text and wrongly flagged 9% of human text, and OpenAI shut it down on 20 July 2023 "due to its low rate of accuracy" (OpenAI, as reported by Decrypt).
• Non-native writers get flagged most. Seven popular detectors wrongly flagged an average of 61.22% of TOEFL essays by non-native English writers as AI-written, and 97.8% were flagged by at least one detector, while US students' essays were classified almost perfectly (Liang and others, Stanford, published in Patterns, 2023).
• Polishing the wording fixes it. When those same essays were run through ChatGPT with an instruction to sound more like a native speaker, the false-positive rate fell from 61.22% to 11.77%.
• None of 14 tools passed 80%. A peer-reviewed test of 14 detection tools, including Turnitin, found all scored below 80% accuracy and only five scored above 70%; editing and paraphrasing made results worse (Weber-Wulff and others, International Journal for Educational Integrity, 2023).
• Human markers miss it too. 94% of GPT-4 exam answers secretly submitted at the University of Reading went undetected, and they scored higher than real students' work on average (Scarfe and others, PLOS ONE, June 2024).
• Small error rates add up. Even in a favourable 2026 test, current commercial detectors wrongly flagged 1.36% to 1.84% of human articles (Graphite). Across millions of pages, that is a large number of human writers marked as AI.

How Are AI Overviews Affecting Blog Traffic?

They reduce clicks when they appear, but they haven't ended search traffic. Behavioural data shows people click far less on searches with an AI summary, while Google says total clicks are stable without publishing figures. For bloggers, the effect is real and uneven: informational posts that answer a simple question are the most exposed. Forums compete for the same clicks too: in our own October 2026 measurement, a Reddit thread appeared on page one for half of 400 everyday searches.

• Half the clicks. In browsing data from 900 US adults in March 2025, people clicked a regular result on 8% of Google visits with an AI summary, against 15% without one. Only 1% clicked a link inside the summary, and 18% of searches produced a summary (Pew, July 2025).
• Position 1 loses most. When an AI Overview appears, the top-ranked page gets a 58% lower click-through rate, and position 3 gets 46.4% lower (Ahrefs, February 2026; 300,000 keywords).
• Some recovery. Across 53 brands, click-through on searches with an AI Overview fell from 3.19% in January 2025 to a low of 1.31% in December 2025, then recovered to 2.36% by February 2026 (Seer Interactive, April 2026).
• 68% of US Google searches ended without a click in January to April 2026, up from about 60% in 2024 (SparkToro, June 2026).
• News sites lost a third of their Google traffic. Google search traffic to more than 2,500 news sites fell 33% globally and 38% in the US between November 2024 and November 2025, though the Reuters Institute says it is "not clear how much of this is down to AI overviews" (Chartbeat data in Reuters Institute Trends and Predictions 2026).
• Publishers expect worse. News executives expect search traffic to fall 43% over the next three years (same report).
• AI chatbots are still a small share of searching. AI tools handled 3.2% of US desktop searches in late 2025; Google handled 73.7% (SparkToro, March 2026).
• Google's own figures. Google says AI Overviews have more than 2.5 billion monthly users (Google I/O 2026) and that total organic clicks have been "relatively stable year-over-year" (Google, August 2025), without publishing supporting data.

AI Copyright Lawsuits and Licensing Deals: The Numbers

The money is now concrete. A $1.5 billion settlement paid authors for pirated books used in AI training, publishers are licensing backlists, and US courts and the Copyright Office have said AI output without human authorship can't be copyrighted.

• Bartz v. Anthropic: $1.5 billion. Anthropic agreed to pay $1.5 billion plus interest to settle the authors' class action, about $3,000 per book, which the court called four times the $750 statutory minimum. Final approval came on 20 July 2026 (Authors Guild; Association of American Publishers). The $3,000 is per work, before fees, and is shared between rights holders such as author and publisher.
• Nearly half a million books. The settlement covers nearly half a million works; by 19 March 2026, owners of 264,809 works (54%) had filed claims (final approval motion).
• Lawyers' fees cut. The court reduced fees to about 6.8% of the fund ($101.56 million), against the 12.5% requested (Authors Guild).
• Why it settled. In 2025 the court ruled that training AI on books was fair use, but that downloading pirated copies to build a library was not, which is what the settlement paid for.
• Kadrey v. Meta. Meta won on fair use against 13 authors in June 2025, but the judge wrote that the ruling "does not stand for the proposition that Meta's use of copyrighted materials to train its language models is lawful" and that the authors "made the wrong arguments" (ruling).
• HarperCollins: $2,500 to the author. HarperCollins offered $5,000 per nonfiction backlist title for AI training, split 50/50 with the author, opt-in only (Authors Guild, November 2024).
• Wiley: $40 million. Academic publisher Wiley earned $40 million from AI licensing in its 2025 fiscal year, up from $23 million (Wiley results via STM Publishing News).
• Prompts aren't authorship. The US Copyright Office says human expression in AI-assisted work can be protected, "but not the mere provision of prompts" (Copyright Office, January 2025).
• Human author required. On 2 March 2026 the US Supreme Court declined to hear Thaler v. Perlmutter, leaving in place the ruling that the Copyright Act requires a human author (Supreme Court docket 25-449).

AI Writing Myths That Keep Getting Quoted

These figures appear in thousands of articles. Each one is wrong, out of date or about something else. The same problem runs well beyond AI: our attention span statistics page traces the 8-second goldfish figure that many articles still repeat as fact.

"90% of online content will be AI-generated by 2026." This came from a 2022 Europol report saying unnamed "experts estimate" up to 90% of online content "may be" synthetic by 2026. It was a forecast about all media, not text, and the figure is no longer in the revised January 2024 version of the report. Measured figures are about 50% for new web articles and about 9% for US newspaper articles.

"57% of the internet is AI-generated." This misreads an Amazon Web Services research paper that found 57.1% of sentences in a translation dataset had been translated into three or more languages, much of it likely by older machine translation. It says nothing about ChatGPT-style writing.

"More than half of new articles are written by AI, and rising." The study behind it has been updated: the share has been flat at about 50% since early 2025 (Graphite, May 2026).

"AI detectors are 98 to 99% accurate." These are vendor claims. Independent tests found no tool above 80% accuracy and 61% false positives on non-native writers.

"69% of all Google searches are zero-click." Similarweb's 56% to 69% rise was for news searches only, as reported by Press Gazette. The all-search figure is SparkToro's 68.01% for early 2026. The numbers look alike but measure different things.

"Anthropic is paying each author $3,000." It is about $3,000 per book, before fees, shared between rights holders. An author with several books claims for each one.

"HarperCollins pays authors $5,000 per book for AI training." The $5,000 is split 50/50, so the author gets $2,500.

"ChatGPT has 1 billion weekly users." The latest OpenAI figure we could verify is 900 million weekly users in February 2026. The circulating "1 billion" is a third-party estimate of monthly app users, a different measure.

"23% of authors use AI." That is the Authors Guild's May 2023 figure. Current surveys range from 33% to 45% depending on who was asked.

"AI Overviews killed blog traffic" and "Google says clicks are stable." Both overreach. Clicks drop sharply on searches that show an AI Overview, Google publishes no data for its stability claim, and click-through partly recovered in early 2026.

How to Cite These Statistics

Cite the original source linked next to each figure, not this page, so your readers can check the method. If this page helped you find them, a link to "AI and Writing Statistics 2026, BestWebsiteTemplate.com" is appreciated. We re-check these figures when new editions of the main surveys come out; every source was last checked on 19 October 2026.