PROOF

Trusted by the operators building for the AI-search shift.

Three engagements. Real names, real titles. Below: what we did, and the shape of a working program in the measurements we run every week.

CIBC Petro America CIMT Colleges
CASE BLOCKS

Three programs. Three shapes.

CIBC Big Five Canadian bank
Challenge
Enterprise-scale category presence across AI answer engines, with the operating constraints of a regulated bank.
What AISE did
Built the AI Search program end-to-end — baseline, prescription, quarterly re-analysis.
Outcome
Program running. Specific readouts covered by NDA.
Petro America Energy / fuel retail
Challenge
Not being named on ChatGPT for category questions their buyers were asking.
What AISE did
Baseline with Sherlock AI, prescribed changes, verified week over week.
Outcome
Named as the top recommendation in category within 60 days.
“We had no idea we were invisible on ChatGPT. Within 60 days of working with AI Search Experts, we went from zero mentions to being the top recommended solution in our category.”
JT Mundi · CMO, Petro America
CIMT Colleges Post-secondary education
Challenge
Programme visibility on AI engines when prospective students ask for options.
What AISE did
Sherlock AI baseline across five engines, prescription, quarterly re-runs.
Outcome
Engagement in flight. Specific readouts publish when signed off.

We do not publish numbers a client has not cleared. If it is on this page, it is on the record.

WHAT WE MEASURE

Three numbers. Nothing else counts.

Impressions, traffic, rankings — none of that survives when the answer is generated on the spot by a model. So we stopped counting them. These three are what remain.

01
Share of voice

Of the five engines we test each week, how many name you on the questions your buyer actually asks. Reported as X of 5 per query, aggregated across the query set. This is the ceiling metric — everything else feeds it.

02
Citation rate

When an engine cites sources for its answer, how often is your site among them. Different engines cite differently — Perplexity always, ChatGPT sometimes, Claude rarely — so we normalise per engine before we compare.

03
Named responses

The raw count. On how many of this week's answers did the engine actually type your brand name. This is the number that moves first when a program starts working — days, not months.

WHAT A WORKING PROGRAM LOOKS LIKE

Share of voice, week zero to week eight.

The shape of a program that's working. Not one client — the pattern. Week one moves Named Responses. Week two through four moves Citation Rate. Share of Voice — the ceiling — moves last, once the compounding starts.

Week 0 — before
0 / 5
Week 4 — mid
2 / 5
Week 8 — target
4 / 5

Bar shape is illustrative of the pattern, not one client's numbers. Real client readouts are attached to signed case studies as they clear.

See your own three numbers. Free.

Your Sherlock AI Report is a real Sherlock AI run against your live domain. You see your share of voice, citation rate, and named responses across all five engines.

Get Your Free Audit
Free. Delivered in 3 business days.