Search evidence, counted twice

The first split I trusted happened in a railway café in Bologna: one espresso gone cold, one ranking report open, and an assistant answer naming a business that did not sit where the report said it should. The page at the top had the old keyword covered. The cited answer had a clearer sentence about category, place, proof and fit. That small mismatch became the line I still follow.

I am from northern Italy, and most of my work has grown out of practical search jobs rather than theory. I have audited search-result pages, rebuilt service pages so they can be extracted cleanly, mapped local and national buyer queries, reviewed directory and aggregator footprints, and coached small marketing teams that had good businesses hidden inside vague copy. Over time I started keeping two notebooks: one for ranking evidence, one for citation evidence. When they disagree, the useful questions begin. Why does the assistant trust a directory more than the business page? Why does an English travel-style query describe an Italian firm better than its own page? Why does a competitor get named for a service it explains less fully?

Now I help teams build query sets, prompt sets and cited-source reviews that can be repeated without chasing screenshots. I am strongest where the business sells expertise, location knowledge or complex services, because those markets punish fuzzy language. My stance is plain: search position is still a signal, but it is no longer the full account of visibility. An Italian business needs to know when assistants choose its name, when they flatten its offer, when they borrow evidence from aggregators, and when they leave it out because the page never stated the obvious in reusable words.