Engagement log: three months of source changes and what each one moved
Written by
Odin Lund
Category
Agents
Date
Share this article
Baseline: June 2026
One client in the legal category, tracked on 60 customer prompts across ChatGPT, Gemini and Google AI Overviews against 14 competitors. Starting position: share of voice 4.6%, visibility 2.4%, rank four of fourteen. The audit found three problems. Facts differed between the website, its directory listings and its review profiles (founding date, office locations and case types were each stated at least two ways). The content library was large but answered every question partially and none completely. There was no structured data on the pages the models retrieved most.
Month one, July: owned pages and facts
Changes made: the main practice-area pages rewritten so each answered one customer question with the same set of facts; 16 facts agreed with the client and applied to every owned page; schema added across 355 pages. Measured effect at month end: share of voice 29.5%, visibility 13.1%. The client’s site became the most retrieved domain in the category, from fourth. This is the largest single move in the log, and it came entirely from sources the client already controlled.
Month two, August: content and earned media
Changes made: 31 new pages published against the prompts where no business was yet named, highest intent first; the directories and publications the models cited most were identified and worked, starting with the ones that carried competitors but not the client. Measured effect: share of voice 43.8%, visibility 21.5%, 46 third-party sources carrying the business. On ChatGPT specifically, share of voice reached 54.7%, the first model where the client led outright.
Month three, September: holding the position, and sources
Changes made: prompts added as customer phrasing changed; pages refreshed where a competitor’s page had overtaken them; facts re-checked across every source. Measured effect on 17 September: share of voice 47.2%, visibility 23.3%, rank one of fourteen, 1,164 answers citing the site in 30 days, alignment 90.2%. The gains in month three were smaller than in month one, which is expected: the first pass removes the largest inconsistencies, later passes defend the position. The next measure is commercial: AI referrals are now being matched to the firm’s intake so that attribution runs to signed matters.
Sources
SourceNode tracking, aisearchmonitor.com, daily runs June to September 2026, 60 prompts, 14 firms.
Client-verified facts and page inventory, William Mattar Law Offices, published with permission.
Share Blog
