The Web Is Filling With AI Content. Your Notes Are Where You Keep What's Real.
In November 2024, the volume of AI-generated articles on the web passed the volume of human-written ones1. The open web is not ending. But it is getting harder to rely on, and harder to find inside — and you do not own any of it. The durable place to keep what you know is a body of notes written for yourself.
That last clause is the whole argument. The reflex, when a channel gets noisy, is to publish into it harder: more posts, better keywords, a tool that promises to win the feed back. This essay makes the opposite case. The feed is not yours to win. What is yours is the writing you keep on your own device, in a format no platform can revoke — a private layer that survives every algorithm change because it never depended on one.
What share of the web is AI-generated now?
A large and rising share. Graphite, analyzing articles published between January 2020 and May 2025, found that AI-generated articles surpassed human-written ones in November 20241. Ahrefs sampled roughly 900,000 new pages and reported that 74.2% contained AI-generated content2. These are detector estimates, not a census. But the direction is not in dispute.
Two honest caveats belong here before the numbers do any work. First, those percentages come from AI-detection tools, and detection is contested. Graphite says so plainly in its own study: there is "considerable disagreement about the accuracy of AI detection algorithms, and many argue that detecting AI is impossible, or at best, highly inaccurate."1 Read 74.2% as an estimate, not a verdict.
Second, AI content is not the same as bad content, and a flood is not a takeover. Graphite found the speed startling — "After only 12 months, AI-generated articles accounted for nearly half (39%) of articles published"1 — but it also found those articles "largely do not appear in Google and ChatGPT."1 The web is not dead. It is louder, and the signal is harder to trust at a glance.
Why do most Google searches no longer send a click?
Because the result page increasingly answers the question itself. In 2024, SparkToro and Datos measured that "58.5% of American Google searches resulted in zero clicks"3. By early 2026, SparkToro put that figure at "68.01% of Google searches ended without a click," attributing the acceleration to AI Overviews4. The page you publish to keeps more of its own traffic each year.
The non-vendor evidence is the sharpest part. The Pew Research Center watched the browsing of 900 U.S. adults in March 2025 and found a clean before-and-after: "Users who encountered an AI summary clicked on a traditional search result link in 8% of all visits. Those who did not encounter an AI summary clicked on a search result nearly twice as often (15% of visits)."5 An AI summary roughly halves the clicks the open web receives. And people rarely click the summary's own sources — Pew recorded that in "just 1% of all visits to pages with such a summary."5
There is a structural point underneath the figures, and Rand Fishkin stated it well: "for every 1,000 searches on Google in the United States, 360 clicks make it to a non-Google-owned, non-Google-ad-paying property."3 The open web is shrinking as a destination. Publishing into it is less and less a way to reach anyone.
Why does the same handful of sites keep winning?
Because the surfaces that still send clicks concentrate on a few aggregators. Semrush tracked Reddit's presence in Google's AI Overviews and reported a jump that is hard to overstate: "The site appeared in just 2.3K AI Overviews in November 2024. Now it appears in more than 8.3 million of them."6 When the web answers itself, it answers from a narrowing set of sources.
This is the shape worth naming. The villain in this story is not Reddit, or Google, or any AI vendor — each is doing what its incentives reward. The villain is the shape of a web you publish into but do not control: where the rules change weekly, where a summary can intercept the click, where reach is rented and never owned. Cory Doctorow gave that arc its name. "Here is how platforms die," he wrote: "first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves."7
He called the pattern enshittification — "a seemingly inevitable consequence arising from the combination of the ease of changing how a platform allocates value, combined with the nature of a 'two sided market,' where a platform sits between buyers and sellers, holding each hostage to the other, raking off an ever-larger share of the value that passes between them."8 If that is the shape of every channel you publish into, the question is what you keep off it.
What is the durable alternative to publishing harder?
A body of writing kept for yourself. Andy Matuschak states the principle directly: "Write notes for yourself by default, disregarding audience."9 His own published notes describe the practice: "These notes are mostly written for myself: they’re roughly my thinking environment"10. The audience is not the point. The thinking is, and the record of it is yours.
This is not a retreat from the public. It is a different good. Writing for yourself and publishing for the world are two distinct acts, and the second does not replace the first — Matuschak, after all, publishes the notes he wrote for himself. The argument here is not "go silent." It is "have a durable private layer underneath whatever you publish," a place where what you have judged worth keeping does not depend on a feed to exist.
And a private knowledge base is not a magic filter. It does not make the web quieter or the search results truer. It does one thing: it gives you somewhere to keep what you have already decided is worth keeping, where no ranking change can bury it and no platform can revoke it. That is a smaller claim than "own your data" in the privacy sense — a distinction worth keeping straight, and one we draw out separately11.
How do you keep notes that outlast the tools?
Keep them in a format you control, on a device you control. Steph Ango, who leads Obsidian, put the test plainly: "Apps are ephemeral, but your files have a chance to last."12 His standard is exacting — "if you want to create digital artifacts that last, they must be files you can control, in formats that are easy to retrieve and read."12 Plain Markdown clears that bar. A proprietary database does not.
Three properties make a note durable, and they are worth being concrete about:
- It lives on your own device. Stored locally, not on a server whose terms can change. It works offline, with no account standing between you and your own writing.
- It is plain Markdown. A text format you can open in any editor, today and in a decade. Portable by default — there is nothing to export from, because it was never locked in.
- You decide who sees it. Sharing, when you choose it, is end-to-end encrypted; the default is that the note is yours and stays where you put it.
None of that is exotic. It is the inversion of the platform shape: instead of renting reach and hoping the rules hold, you keep the writing somewhere the rules cannot reach. The feed will keep churning. Your notes will not.
What this actually changes tomorrow
Less than you'd think, and more than you'd hope. You will still read a noisier web and run searches that send you nowhere. What changes is where the things you find, and the things you figure out, come to rest. Not in a draft folder on someone else's platform, one terms-of-service update from disappearing. In a file you can read, move, and keep.
The reflex the noisy web provokes is to publish louder. The quieter, more durable move is to write for yourself first, in a format that outlives the tool, and to treat publishing as something you do from that private layer — not instead of it. The web will fill with whatever the incentives reward. What you keep for yourself is the part that stays real.
Frequently asked questions
What percentage of web content is AI-generated?
Estimates are high and rising. Graphite found AI-generated articles surpassed human-written ones in November 20241; Ahrefs sampled ~900,000 new pages and found 74.2% contained AI-generated content2. Both rely on AI detectors, which are contested; Graphite itself notes "considerable disagreement about the accuracy of AI detection algorithms"1. Treat the figures as estimates of a real trend, not a precise census.
What percentage of Google searches are zero-click?
Most of them, and the share is climbing. SparkToro and Datos measured 58.5% of U.S. Google searches as zero-click in 20243; SparkToro put it at 68.01% across the first four months of 2026, citing the growth of AI Overviews4. "Zero-click" means the searcher got their answer on the results page and never visited a website.
Do AI summaries reduce clicks to websites?
Yes, by roughly half. The Pew Research Center found that users who saw an AI summary clicked a search result in 8% of visits, versus 15% for users who did not see one5. Clicks on the links inside the summary were rarer still, about 1% of visits to pages with a summary5.
Should I write notes for myself or for an audience?
For yourself, by default. Andy Matuschak's rule is "Write notes for yourself by default, disregarding audience"9, because notes written to be useful to your future self stay useful regardless of who else reads them. This does not bar you from publishing — it means the thinking comes first and the audience, if any, comes later.
Is it worth keeping my own knowledge base?
Yes, as a durable private layer — with one honest limit. A personal knowledge base does not replace publishing, and it does not filter the web for you. What it does is hold what you have judged worth keeping in a form no ranking change can bury and no platform can revoke. Kept in plain Markdown on your own device, that record outlasts the tools you used to make it.
When the channel you publish into can change its rules every week and intercept the click before it reaches you, the writing worth trusting is the writing you kept for yourself. The web will fill with whatever pays. Your notes are where you keep what's real — and a knowledge base you own, in plain Markdown stored on your own device, is one place to keep it: mnmnote.com.
Footnotes
-
Paredes, J. L., Smith, E., Druck, G., & Benson, B. "More articles are now created by AI than humans." Graphite, May 15, 2025. https://graphite.io/five-percent/more-articles-are-now-created-by-ai-than-humans (dataset January 2020 – May 2025; November 2024 crossover). Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084558/https://graphite.io/five-percent/more-articles-are-now-created-by-ai-than-humans ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7
-
Law, R., Guan, X., & Soulo, T. "What Percentage of New Content Is AI-Generated?" Ahrefs, May 19, 2025. https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/ (~900,000 new pages sampled, April 2025; 74.2% contained AI-generated content). Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084631/https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/ ↩ ↩2
-
Fishkin, R. "2024 Zero-Click Search Study." SparkToro (data: Datos, a Semrush company), July 1, 2024. https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084742/https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/ ↩ ↩2 ↩3
-
Fishkin, R. "In 2026, Less Than One Third of Google Searches Still Send a Click." SparkToro (data: Similarweb, January–April 2026, US), June 8, 2026. https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084719/https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/ ↩ ↩2
-
Chapekis, A., & Lieb, A. "Google users are less likely to click on links when an AI summary appears in the results." Pew Research Center, July 22, 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/ (n=900 U.S. adults; browsing data March 2025). Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084839/https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/ ↩ ↩2 ↩3 ↩4
-
Silva, C. "Reddit SEO: 5-Step Guide to Search Engine & AI Visibility." Semrush (data: Semrush Organic Rankings tool), December 17, 2025. https://www.semrush.com/blog/reddit-seo/. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084746/https://www.semrush.com/blog/reddit-seo/ ↩
-
Doctorow, C. "Tiktok's enshittification." Pluralistic, January 21, 2023. https://pluralistic.net/2023/01/21/potemkin-ai/. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084518/https://pluralistic.net/2023/01/21/potemkin-ai/ ↩
-
Doctorow, C. "Tiktok's enshittification." Pluralistic, January 21, 2023. https://pluralistic.net/2023/01/21/potemkin-ai/. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084518/https://pluralistic.net/2023/01/21/potemkin-ai/ ↩
-
Matuschak, A. "About these notes." notes.andymatuschak.org. https://notes.andymatuschak.org/About_these_notes. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084816/https://notes.andymatuschak.org/About_these_notes ↩ ↩2
-
Matuschak, A. "About these notes." notes.andymatuschak.org. https://notes.andymatuschak.org/About_these_notes. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084816/https://notes.andymatuschak.org/About_these_notes ↩
-
MNMNOTE. "Owning Your Data Is About More Than Privacy." https://blog.mnmnote.com/posts/own-your-data-beyond-privacy. See also "What You Type Into AI Leaves Your Walls" (https://blog.mnmnote.com/posts/what-you-type-into-ai-leaves-your-walls) and "Export Before You're Forced Out" (https://blog.mnmnote.com/posts/export-before-youre-forced-out). ↩
-
Ango, S. "File over app." stephango.com, July 1, 2023. https://stephango.com/file-over-app. Accessed 2026-06-22. Wayback: https://web.archive.org/web/20260622084830/https://stephango.com/file-over-app ↩ ↩2