Proof
What we have measured for named clients
Three businesses have agreed to be named. For each one, here is what we changed, what we measured, and what we refuse to claim. Every number below is taken from a dated record; none of it is a promise.
Shaffer Capital — a rebuild that waited
Shaffer Capital is a Montreal commercial financing firm. A new website was ready to replace the live one. Our comparison of the two builds said do not launch yet, and wrote the blockers out in plain words for the builders. They fixed them; each new draft was re-measured; the verdict cleared when nothing measurable stood in the way. The final build carried 18 pages and close to twice the readable content of the site it replaced. The full account.
- Measured: a permanent day-by-day record of visitors, page views and machine visits started on 31 August 2026; the first three days show 128 visitors and 266 page views, busiest day 1 September.
- AI benchmark, September 2026: named in 0 of 6 buyer questions across ChatGPT, Claude and Gemini. That is the baseline the next measurement is compared with.
- What we cannot claim: any change in AI answers. The site went live days ago; the assistants have not been asked again. The next dated measurement will say what moved, or that nothing did.
crossborderloans.ca — the traffic that was hiding in "Direct"
crossborderloans.ca serves cross-border mortgage borrowers. Its analytics counted visitors arriving from ChatGPT as "Direct", because the assistant sends people without a referrer. Reading the arrival address itself (the utm_source it carries) showed 11 ChatGPT arrivals in a fifteen-day window where the old method had counted 3. That is a correction to the measurement, not a gain in visibility, and we say so.
- Measured, 19 August to 2 September 2026: 555 visitors, 686 page views, busiest day 2 September; 192 fetches by live-answer crawlers (the machines that read a page to answer a question) in the same period.
- AI benchmark, September 2026: named in 0 of 6 buyer questions. Baseline recorded.
- What we cannot claim: that the 11 arrivals came from being recommended. An assistant can send a visitor for many reasons; the recorded answers are what will tell us.
nataf.ca — what a busy site looks like to the machines
nataf.ca is a Montreal mortgage brokerage. It is the most visited site we measure, and the one the live-answer crawlers read most.
- Measured, 19 August to 2 September 2026: 2,352 visitors, 4,936 page views, busiest day 28 August with 241 visitors; 285 live-answer crawler fetches in the period. The owner reads the same numbers we do, on a page that never purges its history.
- AI benchmark, September 2026: named in 0 of 5 buyer questions. Baseline recorded.
- Found by the fixed checklist: two pages hidden from search by a noindex tag, two duplicated titles, and a mobile load slower than we want. Ours to fix; they are on the list.
What a client says
“Daniel came to us one day and said he was helping his friend’s website, I casually asked him to help mine too. The lift I have had has been beyond my expectation and above all the prospects are well vetted and match what we do to a T.”
“I was very sceptical but I asked for the free assessment anyways. It opened my eyes, the rest is history. I can’t get over the results.”
And ourselves
Be Preferred is measured on the same clocks as its clients. Our own baseline, with its failing grade, is published here; the report card that grades us is the same one that grades them.
Published 2 September 2026 with the permission of each business named. Measurements are read from server logs and recorded assistant answers; assistant answers vary over time and by assistant, and no outcome above is presented as caused by our work.