To Fight AI Search Spam, Prioritize Real Human Voices
At some point in the future, AIs may create works of art, music, and writing that rival human-created ones. For now, what’s most impressive about most generative AIs is their capacity to produce lots of mediocre work very quickly. This ability is transforming many industries: In the world of higher education, where I work, we are discovering that it’s very hard to tell AI-produced mediocrity from something that indicates that a student is learning to produce good work. But no industry has been so transformed by AI than the shadowy world of search engine optimization.
Search engine optimization is the dark art of making a business more prominent in searches for a particular keyword. Because search engines rely on links, a popular form of search engine optimization involves creating thousands of pages of realistic-looking text with links to the page a client wants to promote. This search engine spam is a pervasive invasive species on the modern web, and generative AIs do a great job at creating it quickly. Indeed, search engine spam is so common that Google, Bing, and other search engines now offer AI-based assistants that promise to answer questions in humanlike fashion rather than pointing frustrated users to deceptive search engine spam. And thus, as internet scholars Judith Donath and Bruce Schneier document, we are beginning to see LLMO (large language model optimization) in an attempt to make new search engine AIs return specific data or recommendations to promote one site or product over another.
In other words, the web is increasingly becoming a cesspit of auto-generated content written by machines designed to be read by machines. As a result, authentic human voices are becoming a rare and desirable commodity. For years, those in the know have added “site:reddit.com” to searches in the hopes of getting an actual human opinion, but SEO has now come to Reddit, with AI-powered bots that promise to mention your product in comment threads in a “humanlike, authentic” way.
To find real human voices, in 2025 we will increasingly flock to the oldest corners of the web, where human moderation keeps the machines at bay. Consider MetaFilter, founded in 1999; it’s where 12,000 paying members, aided by a small team of moderators, feature the best websites and stories that they’ve found and answer one another’s questions in AskMeFi. Or Are.na, an ad-free social network where users curate collections of videoclips, images, and web links to document their interests and explore rabbit holes.
In 2025, we will also carve out our own human spaces using some of the newest social media tools available. Our research indicates that alongside the influencers and micro-celebrities that dominate your TikTok and YouTube feeds are millions of video makers who don’t want to go viral. Instead, they’re making short videos for family and friends, subverting the recommendation algorithms, and using social media not for celebrities but to maintain social ties.
Whether or not AI dominates our entertainment and knowledge futures, it is a safe bet that humans will still want to connect with one another. In 2025, as artificial voices keep drowning out human ones, the tools that help us find authentic voices may be as valuable as those that can convincingly pretend to be human.
