An anonymous, sourced record of how AI is changing the internet, written from inside the industry, with its conflict of interest stated.
- Three jobs. A quarterly record of published measurements, plain explanations of how AI is built and run, and practical defences against synthetic media and fraud.
- Not anti-AI, and not a prediction site. Failed forecasts stay on the page marked failed, and detector results are not used to accuse anyone.
- Anonymous, but checkable. Claims trace to named, dated sources, and mistakes are corrected in place, with a note where the change matters.
- The conflict of interest is disclosed. The site’s editor works in generative AI. No page carries affiliate links.
- Synthetic material is labelled. Generated imagery and fictional framings are marked, and figures are kept apart from invented material.
What the site does
Most writing about AI online is either panic or marketing. This site keeps receipts instead. It does three things, and turns down anything that is not one of them:
- The record. A quarterly gathering of the newest published measurements of the machine internet — bot traffic, the share of new writing that is AI-generated, detection reliability, fraud losses — each with a named source and a date, and each stated with what it does not prove.
- The machinery. Plain-language explanations of how AI systems are actually built, tested, and run: what a training run is, what the physical substrate looks like, who the labs are, and which questions remain genuinely open.
- The defence. Practical, hands-on skills for recognising synthetic text, voice, and video, and for surviving the fraud that now uses them — including the household habits that work regardless of how convincing the fake becomes.
What it is not
It is not anti-AI. Its editor works in generative AI, and the site treats the technology as often genuinely useful and says so on the pages where that is true. It is not a prediction site: forecasts quoted here that failed are kept on the page, marked failed. And it is not an accusation engine — detection tools are, by independent testing, too unreliable to say with confidence that a specific piece of work was machine-made, so this site does not claim otherwise about anyone.
Worry is warranted; hostility is a different thing. Automated traffic passed half of the web. Roughly half of new articles are machine-written. Nearly $900 million in losses was reported in a single year, in one country, in fraud complaints that mentioned AI. Those are the reasons this exists. They are also not the end of the internet, and the site says that just as plainly.Bots and articles checked at source 11 Sep 2026, fraud 17 Sep 2026. Bots: Imperva (Thales), 2026 Bad Bot Report — “Automated traffic continues to outpace human activity online, accounting for more than 53% of all web traffic in 2025” imperva.com. Articles: Graphite — “In Q1 2026, primarily AI-generated articles accounted for 49.9% of published content” graphite.io. Fraud: the FBI’s 2025 IC3 Annual Report, read at source 17 Sep 2026 in the PDF at ic3.gov, section “Artificial Intelligence (AI) Used in Cybercrime”: “In 2025, IC3 received more than 22,000 complaints reporting AI-related information. Adjusted losses of these complaints exceed $893 million.” An earlier version marked this not read at source because the PDF would not yield text on 11 Sep; it did on 17 Sep, and the figure matched. The same totals are reported by Abnormal AI abnormal.ai and Elder Law & Advocacy elaca.org.
Read each figure for what it measures, not for a side. The site is neither against AI nor in the business of forecasting, and it says what its numbers do not prove.
Who writes it
The site is published anonymously. Its editor works in generative AI, and the anonymity is stated rather than hidden.
Nothing here rests on anyone's authority. Every claim traces to a named institution with a publication date, so a reader never has to trust whoever wrote it — only to check the arithmetic. The sourcing rules are written out on the method page, and mistakes are corrected in place on the page concerned, usually with a note saying what it said before.
The perspective is the point: this is written from inside the industry making the content, not from outside it. That is also the conflict of interest, and it is disclosed in full here: no page on this site carries affiliate links, and no link here earns the site anything. No tool has ever paid for placement, for review wording, or to be left out. The earnings ranges the money-making guides once carried were removed rather than kept as illustrations.
Check the source, not the site. Each claim is meant to lead to a named, dated source you can open yourself, and no link here earns the site money.
How the site is made
The exhibits are built with machine assistance and edited by hand. The generated imagery, video and illustrative examples are synthetic and labelled as such wherever they appear. Fictional framings — the archetypes, the games — are marked as fiction. The figures are not fiction, and never share a page with invented ones without a label between them.
Using this material
Quoting, citing, and linking are welcome, including by journalists and researchers. Please cite the underlying source alongside this site where one is named — that is where the measurement came from. If a figure here is wrong, the correction belongs on the page, in the open, not in a quiet edit.
Dead Internet Theory: what it claims, and what the evidence supports · PRESS AND CITATION