Author: Kita Yohei Published: June 2, 2026
Through this series, I've written that "primary source information is important," that it's something you "find," and that it "starts from observation." But that raises the next question: "I observed something. Now what?" This article is about how to turn observation into content — my thoughts on making it work.
I Failed at First Too
I'll be honest: at first, I was just writing down what I noticed. "I tried this." "This was interesting." It was closer to a diary than anything else.
Nobody read it. Which makes sense — my experience wasn't information that meant anything to anyone else.
Noticing something and having information are different things. Observing and primary source information are also different. When I started being conscious of that distinction, things slowly began to change.
Observation Alone Doesn't Become Primary Source Information
"I tried it in ChatGPT." "I tried it in Gemini." — those are observations. But they can't be called primary source information.
What AI wants isn't "results" — it's "what did you learn from this?" Only when observed facts are combined with interpretation and hypothesis does information become something worth citing.
Another way to put it: the difference between a diary and an analysis.
Three Things I Keep in Mind
There are three things I'm conscious of when turning observation into primary source information.
① Compare
Information has more value with comparison than as a standalone observation. Not just "appeared in AI or didn't" — but "what changed between three months ago and now?" or "Competitor A appears but Competitor B doesn't." Adding a comparison axis raises the resolution for the reader.
② Include Numbers
It doesn't have to be large-scale. "Appeared 4 out of 5 times." "18 of 30 companies were cited." "Observed weekly for 3 months." Specificity matters more than scale. Numbers make it easier for AI to summarize and cite.
③ Verbalize the Insight
This is the most important. Think through the observed fact to "why does this happen?" Skip this step and it stays a diary. Verbalize a hypothesis and it becomes analysis.
The Structure of Information AI Wants to Cite
AI prefers information with the structure of "observation → result → interpretation → hypothesis."
Observation alone (diary) won't get cited. Results alone (a list of numbers) is weak. When there's observation, a result, interpretation, and a hypothesis derived from it — that's when AI can treat the information as "something worth citing."
→ Why Is Industry Research So Often Cited by AI?
Large-Scale Research Isn't Necessary
Let me confirm this again. 10 customers, 5 sales conversations, one month of observation — that's enough.
What matters is "irreplaceability." A 10,000-person survey that only says what everyone already knows won't get cited. 10 people, if it surfaces an observation nobody in the industry has articulated, has value.
→ Can Anyone Create Primary Source Information?
Observation Becomes Primary Source Information Only When Interpreted
Let me return to the story from article ④. I was observing weekly whether my X account appeared in Grok's guest mode search results. The fact that emerged was: "it appears when I search 'AI search optimization.' It doesn't appear when I search 'GEO.'"
As just a fact, that's a diary. An interesting anecdote, but nothing more.
Now add interpretation. "AI search optimization" is a phrase I've consistently used on X over a long period — it appears repeatedly across posts, quotes, and conversations. "GEO," on the other hand, is a term the whole industry uses more broadly. It's recognized more as a field than as something tied to a specific individual.
From this, a hypothesis emerges: AI may be judging expertise not by keyword, but by the strength of association between a concept and an individual. Not "has this person used this word," but "has this person repeatedly appeared in the context of this concept" — and that's how it recognizes someone as an expert.
This hypothesis can be tested or refuted. Try it with other keywords. Try it with other people. This is the "observation → result → interpretation → hypothesis" flow — the shape that functions as primary source information.
My Perspective
I think of primary source information not as "data" but as "observation + interpretation."
So "just observing" isn't enough. "What did I learn?" "Why does this happen?" "What hypothesis can I draw?" — writing to that point is what makes information worth citing.
What Genview wants to publish isn't the results of large-scale surveys — it's hypotheses born from observation. That's what I think Genview-style primary source information looks like.
→ What Is Original Research?