More Than 50% of Articles Online Are Now AI-Generated 🤯

Welcome to another edition of Horizon AI,

A new study says AI now writes the majority of newly published articles online. Are we edging closer to the scenario imagined by the “dead internet” theory?

Let’s find out!

Read Time: 4.5’ min

Here's what's new today in the Horizon AI

• Chart of the week: The Compute, Cash, and Contracts that Power OpenAI

• Over 50% of New Online Articles Are Now Written by AI

• Free Resources

• AI tools to check out

• Video of the week

TOGETHER WITH RIPPLING

Software sprawl? That’s SaaD.

Software was supposed to make work easier. Instead, most teams are buried under it.

That’s SaaD – Software as a Disservice. Dozens of disconnected tools waste time, duplicate work, and inflate costs.

Rippling changes the story. By unifying HR, IT, and Finance on one platform, Rippling eliminates silos and manual busywork.

• HR? One update applies to payroll, benefits, app access, and device provisioning instantly.

• Finance? Close the books 7x faster with synced data.

• IT? Manage hundreds of devices with a single click.

Companies like Cursor, Clay, and Sierra have already left outdated ways of working behind – gaining clarity, speed, and control.

Don’t get SaaD. Get Rippling.

Is SaaD hurting you?

Chart of the week

The Compute, Cash, and Contracts that Power OpenAI

• A visualization that maps OpenAI’s infrastructure network across three flows: compute, cash, and contracts, highlighting the increasingly circular nature of AI development funding.

Source

AI News

AI RESEARCH

Over 50% of New Online Articles Are Now Written by AI

A Graphite study of 65,000 English-language articles found that AI-written content has now slightly overtaken human-written work on the internet, even if you rarely see them.

Details:

• Researchers sampled articles of 100+ words from January 2020 to May 2025 using Common Crawl data and applied Surfer’s AI detector to determine whether they were written by humans or AI.

• AI-written articles rose rapidly after the launch of ChatGPT in November 2022, reaching about 39% within a year. By November 2024, the number of AI-generated articles published on the web had slightly surpassed those written by humans.

• The study does not measure whether AI-written articles are actually viewed by real users; however, it found that most AI-generated pieces receive little visibility on platforms like Google or ChatGPT.

• The authors noted that “the proportion of AI-generated articles has remained relatively stable over the last 12 months.” They hypothesize that this is mainly because people have found that such articles do not perform well in search.

While these findings may seem concerning to some, they show that despite the rapid rise of AI-written articles, the trend now appears to have plateaued, making it unlikely that predictions such as Europol’s 2022 report (which estimated that 90% of online content would be generated by AI by 2026) will come to pass.

Source

AI Findings/Resources

👀 Why AI startups are taking data into their own hands

🔮 The 14 next big things in applied AI for 2025

📺 AI Filmmaker PJ Ace shares his complete process for creating million-dollar-quality ads for brands using Veo 3.1 and Nano Banana

🤯 Using AI to track cartel activity from grainy videos posted on Telegram

🌍 Why world models are the next big thing in AI

AI Tools to check out

⚡ Justcopy: Ship production apps in minutes.

👀 Voxcruit: 24/7 AI interviewer. Let AI handle first-round interviews while you focus on top candidates.

👨‍💻 Kilo Code: Open source AI coding assistant for planning, building, and fixing code.

⭐ Shazi VideoGen AI: Long-form video generator for faceless YouTube channels.

💥 Jasper: AI platform that unifies the brand experience, accelerates content velocity, and automates marketing processes—at scale.

Video of the week

Andrej Karpathy on AGI, Reinforcement Learning, and the Next Decade of AI

Andrej Karpathy joins Dwarkesh Patel to discuss why AGI is still ten years away, why reinforcement learning is “terrible,” and what today’s models are still missing.

He breaks down why early attempts at agents were premature, how pretraining acts as “crappy evolution,” and why true intelligence will come from building cognitive cores rather than just bigger models.

Karpathy also reflects on his nanochat project, the limits of coding AIs, and why progress will feel gradual, not explosive, over the next decade.

Timestamps:

00:00:00 – AGI is still a decade away

00:30:33 – LLM cognitive deficits

00:40:53 – RL is terrible

00:50:26 – How do humans learn?

01:07:13 – AGI will blend into 2% GDP growth

01:18:24 – ASI

01:33:38 – Evolution of intelligence & culture

01:43:43 - Why self driving took so long

01:57:08 - Future of education

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻