AI Slop.

The term became popular from 2024 onwards, when AI-generated images and texts began to fill search results and social networks. 'Slop' is waste or mush, something produced in volume with nobody paying attention to quality. It does not mean every piece of AI content, only the careless kind: texts that say a lot and claim nothing, images with six fingers, articles that merely rephrase existing articles. For search engines this is not a new problem; Google has long rated such content as unhelpful, regardless of who created it.

In web design AI slop has a recognisable face. Ask a language model to build a landing page without guidance and you often get the same thing: purple-blue gradients, a centred headline with one word in a gradient, three equally sized cards with icons, the same default font everywhere, emojis as icons and phrases like 'We revolutionise your business'. None of it is wrong, but it is interchangeable. Because many tools are trained on similar data, they arrive at the same average solutions.

One thing helps above all: constraints. A design system with fixed design tokens, deliberately chosen typography, real photos instead of stock motifs and copy with concrete statements give an AI tool a frame. Dedicated design skills are emerging on GitHub for this purpose (see agent skill); they forbid a coding agent from using typical slop patterns and push it towards distinct decisions.

For SMEs, AI slop is primarily a brand problem. A website that looks like a thousand others builds no trust and is not remembered. Copy without first-hand experience and evidence gives neither Google nor AI search engines a reason to cite this particular page, see information gain. AI tools are therefore not forbidden, but they belong in the hands of people who know what the result should look like and who check it before it goes live.