Guides
Is dead internet theory supported by US platform data?
US data show bots are substantial, not a majority. Meta puts fake accounts at 5 percent. Imperva puts bad bots at 32 percent. Read the proxies before acting.
What to take away
US data do not back the strong dead internet theory. Meta's 2023 annual report put fake accounts at 5 percent of monthly active users. Imperva's 2024 Bad Bot Report put bad bots at 32 percent of web traffic, with humans at 50 percent.
X stopped publishing a bot share after 2022, and unattributable session share remains a workable proxy but overstates automation.
What the US platform numbers say
Meta's 10-K for 2023 estimated fake accounts at 5 percent of monthly users. The same filing put duplicate accounts at about 11 percent. That covers accounts, not sessions or posts. It is the largest platform disclosure.
X has not published a successor number. Elon Musk cited a 20 percent bot claim in a 2022 dispute, but the company did not issue a formal update. Reddit's 2024 registration statement said bots and spam are a small single-digit share of daily active users. All of these are voluntary.
Imperva's 2024 Bad Bot Report offers a traffic-level check. Bad bots were 32 percent of web traffic in 2023. Humans were 50 percent, and good bots filled the rest. Across US retail and social sites, no majority bot share appeared.
That data does not rule out a platform where bots dominate. It shows the platforms that disclose do not support the theory. Wikipedia's entry on dead internet theory agrees that empirical support is missing.
How to read the numbers you can get
How to read the numbers
| Metric | Source | What it shows |
|---|---|---|
| Fake account share | Meta 10-K | 5 percent of monthly accounts |
| Bad bot traffic share | Imperva 2024 | 32 percent of web requests |
| Bot management decisions | DataDome network data | Percent of traffic blocked or challenged |
| US platform bot ratio | X shareholder letters | Not disclosed since 2022 |
| DAU bot share | Reddit registration filing | Small single-digit percentage |
No public dataset measures dead internet theory directly. Unattributable session share is the share of logged sessions a platform cannot tie to a verified human. A session counts when identity checks fail, automation flags fire, or a device fingerprint links to automation.
Unattributable session share triggers
Is unattributable session share above 30% for two consecutive quarters?
treat as trigger for deeper review
is it between 15% and 30%?
Above 30 percent for two consecutive quarters is a trigger. Between 15 and 30 percent, keep quarterly tracking. Below 15 percent, annual review is enough. This proxy plus the disclosures above gives a practical read.
What the proxy cannot tell you
Anonymity sits inside unattributable share. Readers who block trackers, use a VPN, or share a tablet register as automation. The metric overstates bots by a margin that tracks privacy habits. It says nothing about content quality.
Automated accounts can post original writing, and humans can churn out spam. Tracing where a meme originated is content provenance work, not session counting. No federal rule requires bot disclosure.
47 U.S.C. 230 shapes filtering, not disclosure. Our page on what Section 230 protected explains why takedown counts are policy artifacts.
Attribution and its limits
Platforms choose what to disclose under definitions they write themselves. Consent orders add reporting duties. The FTC case page for the 2012 Facebook order documents the compliance reporting that followed.
Historical counts carried the same flaw. Platform figures from the 2000s were assembled by hand and cited for years afterward. The early social networks statistics we keep show how much rests on voluntary disclosure.
Five public tools for third-party checks
DataDome sells per-request bot management, publishes a quarterly bot traffic dashboard, targets ecommerce and SaaS. Cloudflare Bot Management sits inside Cloudflare's security plans, with a free tier for basic analytics.
Akamai's Bot Manager is enterprise only and often requires a multi-year contract. HUMAN Security's Bot Insights offers an annual report and per-incident data, while Fastly's bot management integrates with its CDN and charges per request after a base plan.
These tools do not answer dead internet theory for a US platform. They give a request-level check that can anchor an internal proxy. Use one, and record its definition before comparing quarters. A 2024 ownership change on X made year-over-year comparison harder, as Twitter vs X user retention shows.
