“Baked into the system” being the most important takeaway here. It’s all going in this direction.
1) Glad to see GP in the fan outs for B2B SEO agency prompts
2) 95% of the agencies in fan outs are over 10 years old.
We have done zero AEO work (and that's going to change shortly) but we are baked in to the system.
TL;DR this is not a channel for brand new companies
Show more
Did everyone spamming Reddit for the past 2 years really not think there would eventually be repercussions? lol 🤦♀️
More personalization in Discover 👀
Today, we’re introducing new personalization features across Search, Discover, and Google News that give you even more control over your experience — and help publishers build stronger, direct connections with their readers 🧵
Show more
New Peec AI data from suggests ChatGPT 5.6 is reducing fan-out queries and citations for listicles, comparisons.
Are they cracking down on GEO-focused articles?
@tomekrudzki @peecAI
Show more
ChatGPT Search changed how it looks things up on 8 August.
It started using site: at scale. Promptwatch saw those domain-scoped searches jump from about 0.4% of background queries to about 17% in a day. Same week, the queries started saying “official” more often. @suganthan has been reading the traffic: one search that already names the brands, then a site: look-up on each brand’s own website for pricing and specs.
Reddit’s share of citations inside ChatGPT Search then fell from 3.83% to 0.52% between 14 and 17 August. The first slip was 8 August, the same day as the site: change. Google’s AI Overviews and AI Mode did not do this.
So this is not “Reddit stopped working.” People still read it. Google still cites it. We still do that work with clients. Most discovery is not ChatGPT Search. What changed is ChatGPT Search asking official sites first, and Reddit is not an official site.
OpenAI has not said why. Two reads both fit.
One is quality. For a year it was pulling the same gamed listicles everyone in SEO already hates. Judging a random page at ChatGPT’s volume is expensive. Searching site: or a .gov is a shortcut. @lilyraynyc write-up this week is that idea: site: as a blunt trust filter.
The other is cost and speed. A short list of official domains is cheaper and faster than wandering the open web. From the outside those two look the same.
There is a catch. If it guesses the wrong official domain, it searches a parked site. That has already been shown. This is not a clever authority system. It is a blunt one.
What to do this week is still the same test Suganthan published. Ask the category question your buyers ask, five times. Read the search queries ChatGPT wrote. The names that appear every time are the shortlist in its head.
If you are never on it, a perfectly structured page will not get you named in ChatGPT Search this week. The work is getting written about: reviews, roundups, comparison pages, the publications your category already reads, communities, people arguing about you in public. Same family of work as Reddit. Not a replacement for it, and not a reason to stop.
If you are on it every time, then the page work is what’s left. Make sure your own site survives the site: look-up: pricing, specs, the actual answer, in plain HTML.
Suganthan:
Lily Ray:
Reddit numbers:
site: change:
Show more
This is a great addition. I’ll be speaking there too! FYI for those in 🇨🇦🇨🇦🇨🇦
Excited to be speaking at SEO IRL 2026 in Toronto on October 6/7!
🎟️ 🔗
People who use AI to generate “devil’s advocate” comments on LinkedIn posts that end in a question for the OP
lol “everyone, quick! Go build a real brand!”
@lilyraynyc So now everyone goes and tries to make official brand sites and stops spamming Reddit..
Yeah if this change sticks it’s going to be pretty hard to optimize anything without good brand recognition
If you have high AI visibility and the trade pubs, influential people,and brands that influence your industry don't mention you, ever, you in danger girl.
New KPIs for GEO.
Take the trade pubs (search engine journal, search engine land, emarketer) and influential brands (ahrefs,semrush) showing up in fan outs, and use analytics + @alertmouse to see do those brands ever mention you, your trademarks or your brands.
Then take the personal brands (Lily Ray, Mike king, rand fishkin, etc) showing up in fan outs.
Use @officialpbuster to harvest their x posts, LinkedIn etc. do they ever mention your brand or your site? You can also use alertmouse to catch these too.
Consider the tactics that got you all that AI visibility won't survive scrutiny if the models continue to use brands in fan outs.
Show more
Watch all of this return Reddit back to what it was supposed to be in the first place
Reddit barely exists in Claude.
Reddit is down citations in ChatGPT.
Reddit has 5.7 billion total visits per month.
Reddit ranks in 1-3 in Google for 70M queries.
Reddit is still the front page of the internet.
Show more
Interesting how many spam updates Google has had this year and how little they seem to do to reduce spam 😅
Will see if this one is any different…
I can't share the URL, but I just found an example of ChatGPT using fan-out queries that contained site: for the *old domain* for a well-known brand that migrated its domain over 6 months ago. 👀
The migration was handled well from an SEO perspective, but the old domain name must be stored in ChatGPT's memory.
The worst part? When you use Google or Bing to do a site: search that is limited to the old version of the domain, the new domain's URLs don't show up in the search results.
So your shiny new URLs likely won't be part of the response unless they are discovered through another fan-out query or another part of the search 🫠
Show more
The scraping battles are heating up 👀👀
