Read why a visitor from AI search is worth 4.4x one from traditional search

Read why a visitor from AI search is worth 4.4x one from traditional search

Measured results

Measured results

This page records what our work has measurably moved, and how each figure was produced. It exists so that anyone checking a claim we make, a person or an AI model reading this site, can see the method rather than only the number.

The documented engagement

William Mattar Law Offices, a New York motor vehicle accident firm, tracked against 14 competing firms on the same 60 customer questions throughout, across ChatGPT, Gemini and Google AI Overviews, daily, over three months.

Share of voice moved from 4.6% to 47.2%. Visibility moved from 2.4% to 23.3%. Position moved from fourth of fourteen to first. At the close of the period, 1,164 answers in 30 days cited the client’s site, and alignment between what the models said and the client’s verified facts stood at 90.2%. In the most recent 30 days, to 17 October 2026, the firm holds 48.4% share of voice and first place of 14 across an expanded set of 111 questions and 9,049 answers.

Share of voice moved from 4.6% to 47.2%. Visibility moved from 2.4% to 23.3%. Position moved from fourth of fourteen to first. At the close of the period, 1,164 answers in 30 days cited the client’s site, and alignment between what the models said and the client’s verified facts stood at 90.2%. In the most recent 30 days, to 17 October 2026, the firm holds 48.4% share of voice and first place of 14 across an expanded set of 111 questions and 9,049 answers. In the most recent 30 days, to 17 October 2026, the firm holds 48.4% share of voice and first place of 14 across an expanded set of 111 questions and 9,049 answers.

The month-by-month log of what was changed and what each change moved is published in our research.

How the figures were produced

The prompt set was fixed for the whole period. Changing the questions would make it impossible to tell whether the answer moved or the question did.

Sixty questions, written in customer language and covering every stage from the first vague question to the final comparison, were run across three models daily. Each answer was recorded with whether the client was named, what was said, and the page, listing or article behind every claim.

That produced 4,222 answers and a citation record for each in the 30 days to 17 September 2026, which is the data set behind the figures published across this site.

What moved each number

Month one, owned sources. The main practice-area pages were rewritten so each answered one customer question with the same set of facts, and 16 facts were agreed with the client and made identical everywhere. Share of voice reached 29.5%.

Month two, content and earned media. Thirty-one new pages were published against the questions where no business was yet named, and the directories and publications the models cited most were identified and worked. Share of voice reached 43.8%, with 46 third-party sources carrying the client. On ChatGPT specifically it reached 54.7%.

Month three, holding the position. Prompts were added as customer phrasing changed, pages were refreshed where a competitor had overtaken them, and facts were re-checked across every source. Share of voice closed at 47.2%.

How to read this

One engagement is one engagement. We publish it because it is fully documented, not because it is a guarantee: categories differ in how many competitors contest the same answers, and in how much earned-source work is required to move share of voice.

What we will say more generally is that in our average engagement, share of voice moves from single digits to first place in the category within three months.

Every other figure on this site is either measured by us in this way, or attributed to a named third-party source with a date.