LLM grooming

LLM grooming is the deliberate flooding of the open web with propaganda or false content so that AI language models absorb it and later repeat it to users as if it were fact.

How to recognise

The strategy — named by the American Sunlight Project — targets machines rather than people. Its sites often have almost no human audience; the goal is simply to be indexed by search engines and ingested by the crawlers that feed AI systems.

The Pravda network is the most-documented example, having published millions of articles aimed largely at AI ingestion.

Its real-world effect is genuinely debated, and an honest reading holds both sides. Independent studies suggest the influence concentrates in "data voids" — narrow topics where little reliable information exists — rather than across the board, and that headline figures can overstate the impact.

The more serious long-term risk is not today's citations, which can be filtered, but the slow contamination of future AI training data, which is far harder to reverse.