Author: Chris Long
Profile: https://www.linkedin.com/in/chris-long-marketing
Holy smokes SEOs, ChatGPT completely changed the way it searches. GPT 5.4 now uses "site:" search A TON in the fan-out queries:
After reading an article from Writesonic on how GPT 5.4 change ChatGPT search, I've been obsessed with looking under the hood of how fan-out queries have evolved. One the biggest points in their article was that Google is showing brands more.
Well they're doing that by utilizing "site:" search a lot more aggressively. It's not uncommon to see 4+ "site:" queries in a string of fan-out queries.
The initial query will narrow down a list of brands and look for comparison information on them. After multiple comparison queries, ChatGPT will start to perform "site:" searches for each brand. It then evaluates the on-site content and ends up citing that in the end output.
So I think there's a few implications from this:
1. On-site content that connects to what your users are prompting is dramatically more important. If your products is known for "AI features", you better have several different pages talking about that on your site.
2. Brands are getting cited directly in the end output. I wonder if this is OpenAI's solve for citing spammy third party listicles. They source the information they need and then cite the brand in the end result.
3. I still wouldn't sleep on off-site (reviews, third party mentions). I'm assuming that's a big part of the "evaluation process" from the fan-out queries.
PS If you want me to pull these for a prompt or two - shoot me a DM.
After reading an article from Writesonic on how GPT 5.4 change ChatGPT search, I've been obsessed with looking under the hood of how fan-out queries have evolved. One the biggest points in their article was that Google is showing brands more.
Well they're doing that by utilizing "site:" search a lot more aggressively. It's not uncommon to see 4+ "site:" queries in a string of fan-out queries.
The initial query will narrow down a list of brands and look for comparison information on them. After multiple comparison queries, ChatGPT will start to perform "site:" searches for each brand. It then evaluates the on-site content and ends up citing that in the end output.
So I think there's a few implications from this:
1. On-site content that connects to what your users are prompting is dramatically more important. If your products is known for "AI features", you better have several different pages talking about that on your site.
2. Brands are getting cited directly in the end output. I wonder if this is OpenAI's solve for citing spammy third party listicles. They source the information they need and then cite the brand in the end result.
3. I still wouldn't sleep on off-site (reviews, third party mentions). I'm assuming that's a big part of the "evaluation process" from the fan-out queries.
PS If you want me to pull these for a prompt or two - shoot me a DM.
Images:
