Last updated: September 3, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
The post-AI internet doesn't look great because generative AI adoption has pushed automated web traffic past human traffic, filled search results with AI-written summaries ahead of the sources they're pulled from, and spun up thousands of AI content farms chasing the same clicks that used to go to real writers. This isn't a mood — it shows up in bot reports, tracking centers, and search results you can check yourself.
Short answer: The "post-AI internet" complaint, popularized by an essay from developer Jordan Goodman, points to three measurable problems: automated bot traffic passed 51% of all web traffic in 2025, AI-generated "content farm" sites tracked by NewsGuard grew from 49 in 2023 to 3,749 by June 2026, and search engines now surface AI summaries ahead of source links.
I test AI tools and watch traffic logs for a living, so when Jordan Goodman’s essay started circulating on Hacker News, I treated it the way I treat any bug report: is this reproducible? In my testing, most of it held up. I ran the same three searches through Google on a normal Tuesday and got AI Overviews ahead of the source pages on all three. I pulled up my own site's Cloudflare analytics and watched the automated-request share climb higher than I expected before I turned on stricter bot rules. The complaint isn't manufactured nostalgia for a cleaner internet. It's a pattern with numbers behind it.
What the "post-AI internet" complaint actually says
Goodman's essay, which racked up a long Hacker News discussion thread, makes three specific claims: search increasingly buries source material under AI-generated answers, social platforms (he singles out X) are overrun with bot-posted "slop," and the near-zero cost of generating content is drowning out anything worth reading. He ends up doing what a lot of people I talk to have quietly started doing too — posting to his own site, subscribing to RSS feeds, and following a shorter list of newsletter writers he trusts.
This isn't a new theory. It rhymes with the older "dead internet theory," which argued that most of what you see online is now bot-generated rather than human. The difference in 2026 is that the theory doesn't need to be exaggerated anymore. The raw numbers are bad enough on their own.
How much of the internet is actually AI now
I pulled the two most-cited data sources behind this argument and checked them against their original reports rather than the summaries floating around.
| Metric | Before the AI wave | Now | Source |
|---|---|---|---|
| Share of all web traffic that's automated | Human traffic held the majority for a decade | 51% automated, human traffic in the minority | Imperva Bad Bot Report, April 2025 |
| Share of traffic that's "bad bots" specifically | — | 37% of all internet traffic | Imperva Bad Bot Report, April 2025 |
| AI-generated "content farm" news sites tracked | 49 sites | 3,749 sites, in 16 languages | NewsGuard AI Tracking Center, June 2026 |
| AI sites tracked, Feb 2024 checkpoint | — | 700+ | NewsGuard AI Tracking Center |
The bot number is the one that stuck with me. For the first time in more than ten years, automated requests outnumber human visitors on the average website, according to the 2025 Imperva Bad Bot Report. The report attributes most of the jump to AI lowering the cost of running scrapers and simple bot attacks. The NewsGuard count tells a parallel story on the content side: a tracker that found 49 AI-generated news sites in May 2023 was tracking 3,749 by June 23, 2026.
What you'll need
You don't need new hardware or a paid tool to act on any of this. You need an RSS reader (Feedly, NetNewsWire, or even a browser extension), a short list of newsletters or blogs you already trust, and ten minutes to change a few settings. If you run your own site or app, you'll also want access to your host or CDN's bot-management dashboard — Cloudflare's is free at the entry tier and is what I used for the numbers above.
Step-by-step: dealing with the post-AI internet
1. Audit where your information actually comes from for a week
For a few days, I kept a running note every time I clicked a link and it turned out to be an AI summary, an AI-written listicle, or a site I couldn't identify a human author for. It's a fast way to see your own exposure instead of guessing at it.
2. Move your regular reading to RSS or direct subscriptions
When I moved my daily reading list into an RSS reader instead of relying on search or social feeds, the AI-slop hit rate on my reading dropped close to zero, because I was pulling from named sources instead of a ranked feed optimized for engagement.
3. Turn on bot management if you run a site
I tested Cloudflare's "Block AI Bots" toggle on aisagely.com; obvious scraper requests dropped within a day of enabling it. My breakdown of Cloudflare’s newer AI traffic options covers what each setting actually blocks versus what it lets through.
4. Learn to spot AI-written pages before you cite them
Watch for generic author bylines, no clear publish history, and prose that restates a topic without adding a specific fact or number. My guide to spotting AI writing has the specific tells I use, tested against real AI output from several models.
5. Trace a claim to its primary source before repeating it
If a stat can't be traced past a listicle or an AI summary, treat it as unverified. That's the same rule I used to write this article — both numbers above link to the original reports, not a recap of them.
Example prompts you can copy
These work with any chatbot to check whether you're looking at a primary source or a repackaged one:
- "Is this page citing a primary source (official report, study, filing) for its claims, or is it summarizing another article? Point to the specific sentence that tells you which."
- "Find the original report behind this statistic: [paste the stat and its source]. Give me the direct link and the publication date."
- "List which sentences in this article are sourced facts and which are the author's own claims or predictions, and separate them."
They're useful because they force the model to trace a claim instead of just repeating the confident version back to you, which is exactly the failure mode driving the search-quality complaints in Goodman's essay.
Common mistakes to avoid
The biggest mistake is treating "an AI summary answered my question" as equivalent to "I found a reliable source" — the summary is often built from pages it never shows you. The second is overcorrecting into paranoia about anything AI-touched; plenty of well-edited, human-reviewed content uses AI somewhere in the process, and blanket distrust just costs you good information along with the bad. Third, don't rely on one AI detector's verdict as proof — when I tested three popular detectors on the same set of known-human and known-AI paragraphs, they disagreed with each other often enough that I now treat any single score as a hint, not a verdict.
Tools that make this easier
If you want to check whether your own content or research process is getting caught in this shift, a few of my other guides go deeper. AI Tool Reviews explains how I actually test AI products instead of taking a vendor's word for it, and AI Tool Ratings covers how to read any review, mine included, without getting fooled by a cherry-picked demo. If you manage a website and want to see how AI crawlers and chatbots are hitting your own pages, AI Search Console is built for exactly that. And if the "slop" side of this problem shows up in your own work — code, docs, or client projects — my take on AI slop flooding real projects is worth a read.
My take
Goodman's essay doesn't offer a fix, and I don't think there's a clean one — the bot-traffic and content-farm numbers aren't going to reverse because one blog post went viral on Hacker News. What changed for me after checking the numbers is how I spend attention day to day: fewer ranked feeds, more named sources I chose on purpose, and a habit of tracing a stat back before I repeat it. That's a smaller internet than the one search engines want to sell you, but in my testing it's a noticeably higher-quality one.
Frequently Asked Questions
Is the "post-AI internet" a real, measurable thing or just a vibe?
It's measurable. The 2025 Imperva Bad Bot Report found automated traffic at 51% of all web traffic, and NewsGuard's AI Tracking Center counted 3,749 AI-generated content farm sites as of June 2026, up from 49 in May 2023.
Who wrote the original "The Post-AI Internet Doesn't Look Great" essay?
Developer Jordan Goodman published it on his personal site; it was widely discussed on Hacker News after that.
How is this different from "dead internet theory"?
Dead internet theory is the older, broader claim that most online content and activity is bot-driven. The post-AI internet complaint is a narrower, more recent version backed by specific 2025-2026 bot-traffic and AI-content-farm data.
What's the fastest way to reduce how much AI slop I run into?
Move your regular reading from search and social feeds to RSS or direct newsletter subscriptions from sources you picked yourself. It's the single change that cut my own exposure the most.
Should I stop trusting AI search summaries entirely?
No — treat them as a starting point, not a source. Click through to the original page before you cite or repeat a specific fact or number.