I spent the last couple months checking how AI models actually describe SaaS products when people ask for recommendations. What I found was kind of wild.
Most products just dont exist in AI answers. Like at all.
You can rank page 1 on google for your main keyword and when someone asks ChatGPT "whats the best tool for X" you are nowhere. Your competitor who barely has a website gets mentioned instead because their docs happen to be structured in a way the model can understand.
every model is different
This is the part that threw me off. You cant just optimize for "AI" because each one works completely differently.
Perplexity searches the web live so if you publish something new it shows up in a couple days. Thats the fast one.
Gemini is somewhere in the middle. It mixes google search results with its own model so changes show up in about a week.
ChatGPT is slow. It runs on training data thats months old. What matters most for ChatGPT is what other people say about you. Reddit threads and review sites carry way more weight than your own blog.
Claude is the slowest to change. Really depends on training data.
what actually works
I tracked a product over 60 days to see what moves the needle. It started at 0.7 out of 10 on AI visibility. After 9 days of publishing specific types of content it hit 3.3. Then we stopped publishing and the score just flatlined. Seven weeks later still flat.
The content that worked:
- Comparison pages with actual tables not just "we are better"
- FAQ sections where the questions match how real people ask things
- Clear one sentence definitions of what the product actually does and who its for
- Documentation that explains capabilities specifically not vaguely
The content that did nothing:
- Marketing pages full of buzzwords
- Blog posts stuffed with keywords
- Generic "we are the best" claims
- Anything behind a login wall that AI cant even access
the reddit thing
This one surprised me the most. A single reddit thread where someone genuinely explained what a product does seemed to move AI recommendations more than five blog posts on the products own site.
AI models are basically looking for what real people say about you. Your marketing page says "revolutionary platform" and AI ignores it. Some random person on reddit saying "I switched to this because the API docs were actually good" and suddenly AI starts recommending you.
the simple version
If you want to check where you stand just go ask ChatGPT and Perplexity and Claude "whats the best [your category] tool" and see if you show up. Try it with 5 different buyer questions.
Most people who do this for the first time are surprised. And not in a good way.
The gap between google SEO and AI visibility is real and almost nobody is working on it yet. Feels like SEO in 2015 where the people who figure it out early get a massive head start.
Curious if anyone else has been looking at this or if you have found things that work.
I built bersyn.com to automate this. Free first scan if you want to check your product.
Top comments (1)
This is so true — I ranked well on Google but got zero mentions in AI answers. What actually worked for me was adding simple FAQs + real comparison tables… way more impact than blog SEO.