Our own data.
These are studies rather than articles. Each one was run on data we collected ourselves, and each is published with its method, its numbers and its limits attached — including the one where the numbers were bad.
Two service categories, four cities: 13% vs 64% SERP overlap
The same two service categories measured across four US cities. One category's results were 13% consistent between cities; the other was 64%. The practical consequence is that some categories are structurally local and cannot be won nationally — and which one you are in decides what is worth building before a sitemap is drawn.
Read the study Primary research · Pilot studyTwo thirds of the sources AI cited were not the business
A pilot across one local service category, asking what AI assistants actually cite when answering questions a customer would ask. Two thirds of cited sources were not the business itself — they were directories and third-party pages. Published with the method and the limits stated, and no business named.
Read the study Aroc's own numbers · Month 1113 impressions, zero clicks: Aroc's own growth report
The first month of Aroc's own search performance, published in full rather than summarised. Twenty days, 113 impressions and no clicks at all. It is here because a growth partner that publishes only its wins is not showing you its method, and the months that follow only mean something against this one.
Read the studyLooking for the method rather than the data? The articles explain how the work is done.