I've been working on a side project for a few months. mostly focused on SEO, getting listed in directories, the usual stuff.
about 3 months ago I randomly started typing product category queries into ChatGPT and Perplexity to see if my product showed up. I hadn't optimized for it or anything, just curious.
it did show up on one platform. but not the others. a competitor with worse SEO was the top recommendation on Perplexity. next week the results completely changed. no pattern at all.
so I started doing this every Monday. same queries across ChatGPT, Gemini, Perplexity, and a few others. wrote down everything. took about an hour each time.
after 3 months I had enough data to realize something: being mentioned, being recommended, and being cited as a source are three very different things. and every platform has its own logic. there's zero correlation with Google rankings.
I ended up building a small tool to automate the checking because doing it by hand was killing me. still early, still figuring out if this is a real market or just my own weird problem.
honestly curious if anyone else here has noticed AI search actually sending them traffic they weren't tracking. or is this too niche to matter.
I tracked what AI chatbots recommend in my niche for 3 months. it made no sense
byu/SolutionBright297 inEntrepreneur
Posted by SolutionBright297
4 Comments
The distinction between being mentioned, recommended, and cited as a source is something most people haven’t thought through yet. Those are genuinely three different things with three different causes.
The zero correlation with Google rankings is the part that should make more people nervous. Everything built around traditional SEO assumes some carryover. There isn’t much.
Curious what you found about which platforms were most consistent week over week, even if the logic wasn’t obvious.
Honestly, I think this is becoming a real category. AI recommendation systems feel much more unstable and opaque than traditional SEO, and most founders still have no idea how or why their products get surfaced. The distinction you found between “mentioned,” “recommended,” and “cited” is especially interesting because those probably drive very different user behaviors.
Google Search is a Ranker, the AIs you’re talking about are LLMs. The former predicts the relevance(or whatever google deiced it should), the later predicts next word. GenAI is pretty much picking next word randomly out of top N probabilities. A slightly different word at the start, might change the whole direction of the output. Therefore GenAI by design cannot recommend the same thing over and over again.
But I’m sure AI companies now are working hard to come up with something, predictable, but that probably would be – another ranker. Hidden underneath and using LLM to ~~convince you~~ describe why these suggestions are relevant.
Anyway, that’s theory, would be curious to compare it with your further findings.
That sounds interesting! I would love to test your tool.