On August 8, OpenAI changed how ChatGPT runs its background searches. Within nine days, Reddit’s share of citations inside those searches fell from 3.8% to 0.52%. Over the same window, site-specific queries, meaning ChatGPT going straight to a particular domain instead of searching broadly, jumped from 0.37% to 16.8%. That’s an 86% drop in citation share in just nine days, with no public announcement to the sources affected.
The numbers only land, though, once you know what Reddit’s position actually was. A study analyzing 30 million citations back in March had already identified Reddit as the single most-cited domain across AI engines, ahead of every other source measured. That’s the baseline worth holding onto. A domain that had been the most-cited source in a 30-million-citation study saw its citation share collapse inside ChatGPT‘s background searches just nine days after the reported retrieval change. If any source looked stable, it was this one.
The reported shift wasn’t really about Reddit as a platform. It was about how ChatGPT decided to search in the background. Instead of casting a broad net across many sources, background searches increasingly went straight to specific, named domains, which is the jump from 0.37% to 16.8% in site-specific queries. Following the reported change, broad, cross-source retrieval seems to have mattered less, and being a named, trusted destination seems to have mattered more. The data doesn’t establish exactly why OpenAI made this change, and it doesn’t prove the change is the sole cause of Reddit’s decline. What it illustrates is how much retrieval behavior can move, and how fast, without any warning to the sources affected.
None of this is really a story about Reddit losing relevance. It’s a story about concentration risk. Reddit’s citation share fell because the system reading it changed how it retrieves and weighs information. The platform itself didn’t get worse, or less useful, or less popular. A strategy built on being cited broadly across many searches lost most of its value in nine days, without Reddit doing anything differently at all.
That’s the risk for any visibility strategy anchored to one platform, one directory, one community, or one type of citation. The value of that position depends entirely on a system that isn’t fixed, and the sources built on top of it rarely get advance notice when it moves.
Translate this to a medspa’s AI visibility, and the lesson holds. A clinic that’s built a strong presence on one review platform, gotten a handful of mentions on one forum, or leaned hard on a single directory listing has a version of Reddit’s problem. It looks like visibility, right up until the system reading it changes how much weight it assigns. Nine days is an extreme example, but the direction is the same. The platforms deciding what AI systems see and cite aren’t standing still, and a visibility strategy built on just one of them is exposed the day that platform’s role changes.
Durable AI Visibility comes from spreading credibility across more than one kind of signal, so no single retrieval change can undo it. In practice that means owned content on the clinic’s own site that actually answers real patient questions, not just service pages. It means provider and clinic information that reads the same everywhere it appears, on the website, in directories, on review platforms, because inconsistency erodes an AI-generated answer’s confidence. It means reviews, genuine third-party mentions, and local information the clinic doesn’t fully control, because those are harder to fake and harder to lose overnight. And it means useful educational content that answers the questions patients are actually asking, which gives AI systems something specific and current to draw from instead of thin service-page copy.
Not every AI platform weighs these signals identically, and how much each one matters can shift over time, the way ChatGPT’s background search behavior reportedly just did. The point isn’t to guess which signal matters most this month. It’s to not depend on any single one of them.
Open ChatGPT, Claude, Gemini, and Perplexity and ask each one what it knows about your clinic. Look for consistency: the same specialties, the same location, roughly the same reputation. Where the answers diverge, or where one platform clearly knows less than the others, that’s a gap in your ecosystem, not a platform problem worth waiting out. It’s a fifteen-minute exercise, and it tells you more about where your clinic actually stands than any single platform’s citation numbers would.
If your self-check turns up more questions than answers, that’s a reasonable thing to bring to BOWO. Reach out if you want a second read on where your clinic’s visibility actually stands today.
According to the analysis behind this piece, Reddit's share of citations inside ChatGPT's background searches fell from 3.8% to 0.52% within about nine days of a reported change to how ChatGPT retrieves information in early August.
A study analyzing 30 million citations in March had identified Reddit as the single most-cited domain across AI engines, which is part of why the nine-day shift is worth paying attention to.
Not stop entirely. Diversify instead. The goal is a broader ecosystem of consistent, credible signals across owned content, reviews, and third-party mentions, so no single platform change can undo your visibility.
Ask ChatGPT, Claude, Gemini, and Perplexity directly what they know about your clinic, and compare the answers for consistency and accuracy.
Stop chasing keywords and start building topical authority. Learn how to position your brand to be the primary source for AI-generated answers.
