Google’s Reddit problem is really an incentives problem

A few weeks ago I wrote about how Google was treating the opinions of Reddit users as expert advice. The platform’s share price has soared in recent months thanks to other AI models widely treating it as a source of factual information. We’re now starting to see the predictable consequences of that decision.

The alarm has been raised by some of Reddit’s moderators. In communities addressing particularly sensitive topics such as health, supplements or hormone treatments, there are reports of a significant increase in interference by companies trying to influence conversations. This is not traditional spam, the kind we can all spot at a glance, but something much more sophisticated: seemingly legitimate accounts, conversations designed to seem spontaneous, along with “recommendations” for products or services seamlessly woven into discussions.

The reason? Very simple: Reddit is no longer simply Reddit. For years, appearing in Google’s top results justified the creation of a gigantic industry dedicated to manipulating search algorithms. SEO ended up turning a large part of the web into a huge collection of pages written for machines rather than people. Google tried to combat that problem by looking for signals that were more authentic, more human, and harder to fake. And it found Reddit.

But in doing so, it has created the same incentive it was trying to avoid: if Google-generated responses use Reddit as a source, if ChatGPT, Claude, Gemini, or any other conversational system consider Reddit conversations relevant, and if millions of users start making decisions based on responses partially constructed from that content, then influencing Reddit becomes economically valuable. And when something acquires economic value, someone will always find a way to manipulate it.

The paradox is that Google turned to Reddit because the web had been filled with content optimized for Google. But by making Reddit a privileged source for its AI systems, it has set in motion the birth of a new industry dedicated to optimizing Reddit for Google and AI. The difference is that now web pages are no longer manipulated. Conversations are manipulated. And if you have corporate responsibilities related to that area, you will already know it, because your long-time agency will already be offering it to you.

In addition, the problem is much more complex to detect. A page created to rank for a keyword often leaves obvious traces, but a seemingly authentic conversation between users is much harder to distinguish from a genuine interaction. Large language models don’t help much, either: they’re extraordinarily good at finding patterns but far less effective at distinguishing between real experience and theater carefully designed to look like real experience. And as more and more people begin to use them to find what they need, the levels of manipulation of their answers through increasingly refined techniques are growing.

All this should lead us to rethink some ideas about the authenticity of online communities. Reddit remains one of the most interesting spaces on the internet precisely because it hosts real conversations, personal experiences and practical knowledge accumulated over years. But the more important a direct source becomes for search engines and AI, the more attractive it becomes for those who want to influence those same tools.

We’ve been here before. Every time a platform becomes a relevant intermediary for attention, visibility or influence, an industry dedicated to exploiting its mechanisms appears. It happened with search engines. It happened with social networks. It happened with recommendation systems. And it is happening now with the sources used by generative AI.

The question, therefore, is not whether Reddit is a source of valuable information. Of course there are nuggets buried in there. The question is how long it will remain reasonable to assume that the views we find there spontaneously reflect the collective experience of a community. The decline of Reddit was inevitable.

Reddit’s problem isn’t Reddit. The problem is incentives. When a seemingly human conversation can end up influencing the responses that hundreds of millions of people receive, those incentives are enormous. So enormous that it is naïve to think that they will not be exploited systematically.

Google thought it had found an escape route from a website contaminated by the SEO plague it had created. What it may not have anticipated is that it was creating the perfect conditions to industrialize something even more valuable: authenticity.

(En español, aquí)