At Catalyst 26, Partnership Leaders’ Partnership & Ecosystem Conference, Ansira ran a roundtable called “Winning AEO: Partnering for Visibility in an AI-First Search Landscape.” Whether you refer to it as AEO, GEO, or simply AI visibility (my preference), everyone made it clear within the first ten minutes of the session that they all had shown up hoping someone else in the room had already cracked this. I don’t want to bury the lede; nobody had.
I don’t think that’s a knock on anyone at the table. A story broke in August 2026 that explained exactly why nobody’s cracked it yet.
Reddit and Yelp swapped places
Reddit has been the go-to answer for “how do I get cited by AI search” basically all year. Then in mid-August, ChatGPT quietly changed how it pulls results from the web. It started sending narrower, targeted sub-queries at specific domains instead of crawling broadly.
A sprawling Reddit thread that used to be a fine answer to a broad question suddenly didn’t match those tighter queries, and it dropped out of the running before ranking even entered the picture. Reddit’s share of ChatGPT citations fell from about 3.8% to 0.5% inside a single week.
Google’s AI Overviews barely moved over the same stretch. Same underlying question, two platforms, and two different outcomes. It’s worth sitting with that for a second if you’ve been treating “AEO/GEO strategy” as a single thing rather than several.
Yelp, meanwhile, went the other direction. Yelp signed a licensing deal with OpenAI in July, so its reviews and business listings now feed ChatGPT’s local answers directly, with attribution and a built-in way to request a quote.
Link those two stories up, and it’s clear that platforms are pulling back from the open web and leaning on sources they’ve struck deals with or otherwise decided to trust. If you’ve been running AI search visibility out of the content team with zero involvement from whoever owns your data partnerships, this is probably the time to change that.
The analyst-quote trick
One person at the table mentioned their team has started writing content built around named, credentialed voices, analysts, and outside experts, quoted directly, on the theory that models weigh an expert opinion more heavily than an anonymous brand claim saying the same thing. They didn’t have specific numbers on it at the table, but they knew for sure that it was improving AI search visibility.
I’ll be honest, I don’t always have the proof either, but it costs almost nothing to test. The fact that a partnership team came up with it before there was a single case study to justify it says that ideas are coming from people in rooms like this one, comparing notes, well ahead of anything the platforms have published.
Why I think the content team cuts were premature
There’s a third piece that came up that I keep chewing on. As of August 2, 2026, the EU Artificial Intelligence Act’s Article 50 transparency rules kicked in, requiring generative AI systems to make their outputs machine-detectable. Anthropic already embeds an invisible watermark in Claude-generated text. Google does something similar with SynthID. This hands search engines, LLMs deciding what to cite, and social platforms a built-in way to spot AI writing and quietly discount it, which Pinterest is reportedly already doing in its own feed.
The AI research labs also want this. Researchers have documented, in a peer-reviewed 2024 Nature paper and a fair amount of work since, that models trained repeatedly on their own generated output degrade over successive generations, a real phenomenon called model collapse. So, labs want AI-written content flagged and filtered out of what they train on next.
Put a pin in both; my hypothesis is that plenty of companies gutted their content teams over the past two years to run on AI generation instead. Forrester expects a 15% reduction in agency headcount in 2026, driven in part by automation and AI. Gartner’s 2026 CMO survey found that labor’s share of the marketing budget rose this year, from 21.9% to 24.5%, even as AI spend climbed too. Fewer people who are more expensive are just a smaller and pricier team, not one that is getting replaced. Gartner recommends rebalancing AI investment toward people rather than tooling alone.
I think a chunk of the decision to replace people gets walked back because the fast, cheap, largely AI-written content those leaner teams produce is exactly what watermarking and citation-filtering are being built to catch. If answer engines start treating unverified AI content the way spam filters treat spam, the companies that kept real writers on staff will have a specific person to credit for their content.
I could be wrong about the timing. But if I’m right, it won’t suddenly show up in next quarter’s traffic report. You’ll see it in whether your content still gets cited at all once the filtering gets good. That’s the bet I’d rather make a year early than a year late, and it’s the argument I’d make to anyone currently justifying a smaller writing team solely on AI throughput.
