Fact check: What percentage of online news comments are generated by humans versus bots in 2025?

1. Summary of the results

The available analyses and cited pieces do not provide a definitive, empirical percentage of online news comments authored by humans versus bots in 2025; instead, they present indicative findings and experiments suggesting a substantial bot presence across internet traffic and on publisher sites. Multiple items note that bots make up a large share of overall internet traffic (Cloudflare estimate cited as nearly a third in 2025) and that AI-driven accounts have been used to post and influence discussions on platforms like Reddit ^{[1] [2]}. Other analyses report AI activity replacing or supplementing human visits to news sites and that AI-generated engagement can be persuasive, but none supply a specific human-versus-bot comment percentage for news comments in 2025 ^{[3] [4] [5]}.

2. Missing context/alternative viewpoints

Key missing elements across the materials include standardized definitions, measurement methods, and platform variation: what counts as a “bot” (spam bot, scripted crawler, or AI persona) and which platforms were measured strongly affect any percentage estimate. The summarized sources themselves acknowledge this absence: no direct survey or cross-platform audit producing a 2025 comment-share figure is cited, and lab or field experiments (e.g., researchers seeding AI accounts) demonstrate possibility rather than population-level prevalence ^{[2] [4]}. Surveys about news consumption and AI use show some public adoption (Reuters Institute-style findings of some AI use in news discovery), but these do not translate into comment-share metrics ^{[6] [7]}. The empirical gap means estimates from traffic-proxy data (like bot share of visits) cannot be straightforwardly converted into comment-origin percentages without platform-level comment-audit studies.

3. Potential misinformation/bias in the original statement

Framing the question as seeking a single percentage for 2025 risks implying a level of empirical certainty the sources do not offer; this benefits narratives claiming the “internet is dead” or that bots have fully supplanted human discourse, which some pieces implicitly push ^{[8] [1]}. Studies that deploy bots in experiments may be used to generalize influence more broadly than warranted—such experiments show capacity, not prevalence ^{[2] [4]}. Conversely, platform operators and some researchers may understate bot commenting to protect perceived legitimacy; the reviewed analyses note neither definitive platform audits nor consensus methodology, allowing opposing actors to select favorable framings ^{[3] [5]}. Given these gaps, policymakers, platforms, and journalists citing a single percentage should disclose methodology and limits to avoid misleading readers.