-
I realised ChatGPT is quietly acting like a recommendation engine
Posted by on November 28, 2025 at 11:09 amI used to think of ChatGPT and similar models as “fancy autocomplete for text”. Then a weird pattern started showing up in my own projects.
New users replied to onboarding emails with things like:
“I found you through ChatGPT.”
“I asked an AI assistant what to use and your site came up.”They weren’t searching in the classic sense. They opened an AI, typed something like “best [tool/service] for [niche] in [country/city]”, and just trusted the answer. No ad clicks, no listicles, just whatever the model decided to name first.
From that moment I stopped looking at these models as just Q&A and started seeing them as a kind of recommendation engine. If that’s true, then there’s a new ranking problem: when does the model decide to mention you, when does it ignore you, and why does the answer change when you tweak the wording or the location?
That question pushed me into a rabbit hole and I ended up building a small project called aioscop. The idea is to track when assistants like ChatGPT or Gemini actually recommend a brand versus its competitors, how that changes by prompt and by GEO, and then use that as a starting point to adjust content so the model is more likely to include you.
I’m less interested in “how do I game this” and more in “what signals are these models picking up that make one brand show up and another disappear”. If users are treating AI answers as trusted recommendations, that feels like an important behaviour to understand.
Curious how people here think about that layer. Are we heading toward a world where “AI recommendation optimisation” becomes a thing, or is this just an emergent side effect we should mostly ignore?
Project if anyone’s curious aioscop com
-
1 Reply
-
I definitely was skeptical when using AI for anything personal related, whether I could trust that it even deleted my data when told to. But I feel that upon my own research that chatgpt uses search engines to skim for data. It doesn’t open the link source but just reads the description of the site, then creates a response so it is not as accurate.
-
It’s pretty helpful when your business and website uses keywords and a formula for queries (e.g. service + industry + location). Google Gemini skims through the top results to present them to inquirers
-
You are spot on that these AI models are acting like powerful recommendation engines now. Digging into what signals actually drive those mentions is seriously important. I got obsessed with this same problem and built MentionDesk for it. It lets you track and optimize your brand’s chance of popping up in AI generated answers. Totally agree this is the start of a whole new area in digital strategy.
Log in to reply.