Signal Through the Slop — Week 39, September 2026

AI search is changing how brands get found, recommended, and trusted. Each week, we pull signal from Reddit, LinkedIn, and industry research to track what's actually moving in the space. Here's what caught our attention this week.

What Reddit is saying this week

The story dominating SEO and GEO discussion spaces this week isn't a hot take, it's a correction. ChatGPT's citation rate for Reddit collapsed somewhere between 73% and 86% over a matter of days in mid-August, and the explanation gaining the most traction ties it directly back to something the community had been warning about since June: brands and agencies flooding Reddit with fake accounts and seeded threads, engineered specifically to get pulled into AI answers. One vendor was even caught advertising an "army of agents publishing blog content and Reddit posts" that promised to get brands "cited by AI" at scale. Reddit and the platforms serving its content noticed, and the shortcut got shut down hard.

If you were doing organic, honest community engagement on Reddit this whole time, the collapse doesn't touch you, but it's worth understanding why it happened. The pattern that got caught: reverse-engineering the question formats LLMs favor, seeding a thread the community would genuinely engage with, then slipping brand mentions into the exact spots that get pulled into an answer, all from accounts with real posting histories built up in advance. It worked for a while. Then it stopped working, fast and all at once, which is usually how platform corrections go once a manipulation pattern becomes visible at scale.

The open question nobody's answered cleanly yet: is Reddit still worth the investment for AI visibility after this, or did the whole channel just get burned for everyone? Early signs suggest it's the former; the platforms are correcting for manufactured presence, not authentic presence. But if you're managing AI search visibility for a brand and you outsourced any part of a "Reddit plus AI citations" package in the last six months, this is a reasonable moment to ask your vendor exactly how they built it.

What LinkedIn is saying this week

The acronym fight is starting to resolve into an actual framework instead of a semantic argument. One newsletter piece that's been circulating widely this week makes a genuinely useful three-way split: GEO is about presence and positioning in training data, SEO is about findability in traditional search, and AEO is about how easily your content's answers can be extracted and quoted once someone's already looking at it. That's a cleaner distinction than most of the "do we really need three acronyms" objections that have been making the rounds, and it gives practitioners a real way to divide the work instead of treating it as one undifferentiated blob called "AI SEO."

Underneath that, a separate data point landed this week that reinforces a theme showing up everywhere right now: an IAB-sourced study found that 56% of shoppers favor AI recommendations that pull from creator or influencer content, rising to 65% among Gen Z and Millennials, and 65% of people overall say credible creator reviews increase their confidence in the AI recommendation itself. That's not a small effect, and it lines up with what's showing up in this week's research too: third-party, off-site content is doing real trust and citation work that a brand's own website can't replicate by itself, no matter how well it's structured.

There's also a quieter thread worth naming: the practitioner conversation has shifted from "should we do GEO" toward "what are in-house teams actually being asked to prioritize and afraid of." That's a more useful question than the one the space was asking six months ago, and it suggests budget-holders are past the awareness stage and into figuring out what to actually staff and fund.

What the research shows this week

Two independently run studies landed on the same basic finding from different angles. A Florida-focused study tracked how 28,488 local businesses across nine cities and eighteen industries showed up when tested against realistic buyer-discovery questions on ChatGPT, Claude, Gemini, and Google AI Overviews. Only about 5.9% of those businesses ever appeared in an answer, and just ten companies accounted for nearly half of all mentions across the entire study. That's not just a visibility gap, it's a concentration story: AI platforms aren't spreading attention evenly across a category, they're consolidating it onto whoever's already winning.

Separately, a citation study analyzing over 20,000 prompts that triggered Google AI Overviews in August found that third-party platforms, not brand-owned pages, account for the majority of citations feeding those answers. YouTube, Instagram, Wikipedia, and similar sources are doing more of the citation work than most brands realize, which is the same mechanism showing up in this week's Reddit and LinkedIn conversations from a different direction: owning your own content well is necessary, but it's not sufficient on its own anymore.

On the platform side, OpenAI made its clearest move yet toward turning ChatGPT into a transactional surface rather than just an answer surface. New ad capabilities shipped this week alongside OpenAI's first CRM partnership (HubSpot) and first ecommerce partnership (Shopify), letting merchants sync product catalogs and run ChatGPT ad campaigns directly, with the integration going international starting today. AI search is no longer only a visibility and citation problem; it's starting to come with its own paid, transactional layer, catalog feeds and conversion tracking included. That changes the calculus for anyone still treating AI search purely as an earned-visibility exercise.

That's the signal this week — back next Monday with more.