A word for it
Merriam-Webster picked "slop" as its 2025 word of the year and defined it as "digital content of low quality that is produced usually in quantity by means of artificial intelligence." Three weeks earlier, Australia's Macquarie Dictionary had chosen "AI slop," which won both the committee's vote and the public vote.
CBS News, "'Slop' chosen as Merriam-Webster's 2025 word of the year", 15 December 2025. cbsnews.com. Macquarie Dictionary, "Word of the Year 2025", 24 November 2025. macquariedictionary.com.au
How much of the web it now fills is harder to measure. The search marketing firm Graphite sampled 55,400 English-language articles from the Common Crawl archive and ran each through three AI detectors. About half of new articles were classified as mainly AI-generated: 50.9 percent in the last quarter of 2025 and 49.9 percent in the first quarter of 2026. Detectors make mistakes, and Graphite found that most of these articles rarely show up in Google or ChatGPT results, so the figure describes what is being published more than what people read.
Paredes, Druck, Benson and Smith, "AI Now Writes as Many Online Articles as Humans Do", Graphite, May 2026. Industry study, not peer reviewed. graphite.io
Teachers are using AI to make materials
In a spring 2025 survey of 2,232 US public school teachers, Gallup and the Walton Family Foundation found that 60 percent had used AI for their work that school year. A third of all teachers, 33 percent, used it at least monthly to make worksheets or activities, and 37 percent used it to prepare to teach.
Gallup and Walton Family Foundation, "Teaching for Tomorrow: Unlocking Six Weeks a Year With AI", June 2025. news.gallup.com
RAND's national survey panels show how fast this changed. In 2025, 53 percent of English, math, and science teachers said they used AI for school, an increase of more than 15 percentage points in one to two years.
Doss, Bozick, Schwartz, Chu, Rainey, Woo, Reich and Dukes, "AI Use in Schools Is Quickly Increasing but Guidance Lags Behind", RAND, September 2025. rand.org
Drafting a worksheet with a model and then correcting it is a reasonable use of a teacher's time. The research below looks at what those drafts contain before anyone corrects them.
What the drafts contain
In a 2025 study, researchers asked ChatGPT, Gemini, and Copilot for 310 civics lesson plans and sorted the 2,230 activities in them by Bloom's taxonomy. Forty-five percent asked students only to remember information. Four percent asked them to analyze, and 2 percent to evaluate. Overall, 10 percent of the activities were designed to promote higher-order thinking, and 94 percent included no multicultural content.
Trust, Maloy, Xu and Pelletier, "Civic Education in the Age of AI: Should We Trust AI-Generated Lesson Plans?", Contemporary Issues in Technology and Teacher Education 25(3), 2025. citejournal.org
Common Sense Media tested the teacher assistants in Gemini for Google Classroom, Khanmigo, Curipod, and MagicSchool, and rated them moderate risk in August 2025. Its reviewers found that tools which write behavior plans and individualized education plans don't have the information those documents need, and can show bias based on a student's perceived background. The report advises teachers to check generated material for accuracy and bias before using it in class.
Common Sense Media, "AI Teacher Assistants" risk assessment, August 2025. Product testing by an advocacy group, not a controlled study. commonsensemedia.org
Some of it is for sale
In August 2026, Chalkbeat reported AI-generated products on Teachers Pay Teachers, where teachers buy and sell classroom resources. Examples included a $1.99 alphabet coloring sheet with no letter F, and a $4 poster of Black inventors that illustrated the inventor of the potato chip with a firefighter in a gas mask. The CEO of IXL Learning, which owns the site, told Chalkbeat that AI products are "a minuscule fraction of a percent" of sales and that stores linked to low-quality AI content are demoted.
Altavena, "An alphabet without an F: AI slop seeps into Teachers Pay Teachers curriculum marketplace", Chalkbeat, 3 August 2026. chalkbeat.org
Children's books have the same problem. In June 2025, the newsletter Indicator identified 517 nonfiction children's books on Amazon, many of them biographies, that appeared to be partly or fully AI-generated. It pointed to errors on the covers and authors with synthetic profile photos. Amazon had removed 198 of them by the time the story ran. Two years earlier, Amazon had limited self-published authors to three new titles a day, a change it said was to help protect against abuse.
Mantzarlis, "Amazon is selling hundreds of seemingly AI-generated biographies for children", Indicator, 11 June 2025. indicator.media. Milliot, "KDP Will Limit Daily Number of New Titles", Publishers Weekly, 18 September 2023. publishersweekly.com
Students are using the same tools
Pew Research Center found that 26 percent of US teens had used ChatGPT for schoolwork in fall 2024, double the 13 percent from 2023. A year later, when Pew asked about any chatbot, 54 percent of teens aged 13 to 17 said they had used one for help with schoolwork, and 10 percent said they did all or most of their schoolwork with a chatbot's help.
Sidoti, Park and Gottfried, Pew Research Center, 15 January 2025. pewresearch.org. McClain, Anderson, Sidoti and Bishop, "How Teens Use and View AI", Pew Research Center, 24 February 2026. pewresearch.org
Few teachers have been trained to judge it
In Gallup's 2025 survey, 19 percent of teachers worked at a school with an AI policy, and 68 percent had received no AI training from their school or district. In Gallup's 2026 survey, 18 percent said they get formal guidance on how to use AI at work. RAND found that over 80 percent of students said their teachers had never explicitly taught them how to use AI for schoolwork.
Gallup and Walton Family Foundation, June 2025, full report (PDF). Gallup, "Most Teachers Receive No Formal Guidance on AI Use", May 2026, news.gallup.com. RAND, September 2025, as above.
What we take from this
The research doesn't show that AI makes classroom material worse by itself. Most of it describes drafts that went out unchecked, in schools where most teachers haven't been shown how to check them. Detection won't close that gap, for reasons we cover in our article on AI detectors. A teacher who reads the material, knows the class, and decides what stays can. That's what the training we're building with teachers is for.
