12/06/2026
Michał Suski presentation
How AI SEO is bullsh*tting You?
Myth 1: API monitoring shows AI ranking
ChatGPT
API call ≠ scraped UI
Scrape shows two times more brands compared to API.
Web search is triggered in API 77%, in real UI 100%
Sources provided from API 75%, in real UI 100%
Real UI answer is ~700 words, API ~400 words
Perplexity
Upside down. Shows more brands in API Real UI and API sources overlap only 8%
Summary
API scraping is easier and cheaper so most of AI visibility tracking tools use this method.
You must use real AI scraping to get more accurate data
Myth 2: Monitor thousands of prompts
Longer prompts / detailed prompts / topic drift
78% return the same brands, sources and answers
You should monitor only those prompts that have different search intent
Myth 3: We need to cover all fanouts
Most fanouts appear once. Track only those that show every day.
Myth 4: Global citation ranking (Reddit, Quora, Instagram and etc)
Every niche has its own core sources.
Don't chase Reddit, find relevant to you by running many queries.
Myth 4: Don't promote Yourself in listicles
Self-promotion listicles still works very good in Google AI overview, AI mode, Perplexity, but less good in ChatGPT. In any case self promotion still works in all of LLMs.
Do it. Do it frequently. In Your own website. In blog post. If You don't someone else will do it.
What works?
1. Semantics - cover facts and you will get a chance to become core source
2. Easy retrieval format
3. Models value different factors
4. Position in listicle
5. Blogs
6. Sentiment
7. Ranking in traditional search engines
8. Source rank in Google The End
Don't chase shiny objects, gather data and don't be fooled
Presentation link in comment.