OpenAI's launch of ChatGPT in November 2022 set off a cascade that is reshaping the public internet. Within months, the volume of human‑driven queries on platforms such as Stack Overflow fell dramatically, and the share of automated traffic now exceeds that of real users.

Sharp decline in human‑generated activity

Stack Overflow, once the go‑to forum for programmers working late‑night fixes, recorded 109,152 new questions in November 2022. By August 2026 that figure had dwindled to just 1,146, a drop of more than 99 per cent that analysts link directly to the rise of AI‑powered code assistants.

Social media sites are reporting a similar pattern. Users complain that feeds are increasingly populated with generic images and captions produced by large language models, a trend that feels especially jarring on platforms like Reddit and Pinterest where personal curation used to be the norm.

Search engines and bots tilt the balance

Google's AI‑generated overviews, which appear above traditional search results, have reduced click‑through rates to original websites. A Pew Research Centre study shows only 8 per cent of users click through after seeing an AI summary, compared with 15 per cent when the summary is absent.

Cloudflare, the company that routes a substantial portion of global web traffic, confirmed that automated agents now generate 58.5 per cent of all requests, leaving humans with just 41.5 per cent. The proportion of bot traffic is still climbing, raising concerns that the internet could become a "dead" space dominated by machine‑to‑machine dialogue.

Grassroots resistance: fonts that fool machines

In response, a wave of developers has created tools that exploit the current limits of image‑based AI. Designer Eric Lu unveiled Ghost Font, a static typeface that appears as visual noise until it moves, revealing the phrase "Hello Human". Without explicit instructions, models such as ChatGPT, Claude or Grok cannot decipher the hidden text.

Similar efforts have emerged from a Brazilian studio and a Danish type‑foundry with ShieldFont, an open‑source typeface that swaps roughly a quarter of the words in a sentence with synonyms that preserve grammatical structure. Scrapers see a coherent sentence and therefore do not flag it, yet the intended meaning is lost to the reader.

Both fonts illustrate a growing DIY movement that seeks to protect human expression from being overwritten by automated summarisation tools. While a simple screenshot can still capture the rendered page, the techniques raise the cost of large‑scale data harvesting.

Implications for European workers and policy

For European content creators, journalists and small‑scale developers, the shift threatens revenue streams that depend on organic traffic. Advertising models built on page views are being eroded as AI answers replace the need to click through to original sources.

Trade unions representing digital workers have begun to voice concerns that the unchecked spread of generative AI could lead to job displacement in sectors ranging from web design to copywriting. At the same time, industry bodies argue that AI can boost productivity and lower costs for consumers.

European regulators are watching the trend closely. The European Commission's recent proposals on AI transparency require that automated content be clearly labelled, a measure that could help users distinguish between human‑authored and machine‑generated material. However, enforcement remains a challenge, especially when developers embed anti‑AI tricks at the typographic level.

As the balance tilts further toward automated traffic, the debate over how to preserve a human‑centric internet is likely to intensify. Whether grassroots tools like Ghost Font and ShieldFont can scale to protect the broader digital ecosystem remains an open question, but they signal a growing willingness among creators to push back against a tide that threatens to marginalise genuine human voices.