Question.Marketing

Proof

The client ledger

Every client engagement we have permission to publish, in full. Baseline measurement, every intervention with its date, and results reported with 95% confidence intervals. Updated on the first working day of each month.

Last updated 31 July 2026

· DOWNLOAD LINK, CSV NOT YET PUBLISHED Download the raw data (CSV)

How to read this

Share of voice is the proportion of sampled AI answers, for a defined prompt set, in which the brand is mentioned. It is not a ranking and it is not traffic.

The interval after each figure is a 95% Wilson confidence interval. A reading of 12.4% ±1.3 means the true value is very likely between 11.1% and 13.7%. If two readings' intervals overlap, we have not demonstrated a change.

Verdict states plainly whether movement is distinguishable from noise. "No change detected" appears often. That's the point.

Interventions are dated so you can assess the relationship yourself. We describe them as preceding a change, not causing one, unless we have grounds for the stronger claim.

Entry template

Duplicate this block per client. Every number below is EXAMPLE, UNCONFIRMED and must be replaced or deleted.

EXAMPLE, CLIENT NAME, EXAMPLE, SECTOR, EXAMPLE, MARKET

Engagement: EXAMPLE, TIER · Started: EXAMPLE, DATE · Prompt set: EXAMPLE, N PROMPTS prompts · Sampling: ~EXAMPLE, N RUNS runs per prompt per cycle

CycleDateChatGPTAI OverviewsAI ModeVerdict
BaselineEXAMPLE, 2026-08-04EXAMPLE, 4.1% ±1.1EXAMPLE, 9.8% ±1.6EXAMPLE, 6.2% ±1.3,
1EXAMPLE, 2026-09-01EXAMPLE, 4.6% ±1.2EXAMPLE, 11.9% ±1.7EXAMPLE, 6.0% ±1.3EXAMPLE, AI Overviews up, statistically distinguishable. ChatGPT and AI Mode: no change detected

Interventions this cycle

DateInterventionDetail
EXAMPLE, 2026-08-06Crawler access correctedGPTBot and ClaudeBot were disallowed in robots.txt
EXAMPLE, 2026-08-14Schema implementationOrganization, Service, FAQPage across 9 commercial pages

Notes: ANYTHING AFFECTING INTERPRETATION, NOT YET WRITTEN

Supplementary panel (quarterly): Gemini, Claude, Perplexity, Copilot, FIGURES WITH INTERVALS, NOT YET MEASURED

Current entries

Stitch-Ai.com is your first entry. It's mid-programme, so publish the baseline and completed cycles only, don't project.

Stitch-Ai.com, SECTOR, UNCONFIRMED, MARKET, UNCONFIRMED

Engagement: TIER, UNCONFIRMED · Started: START DATE, UNCONFIRMED · Status: N CYCLES COMPLETED, UNCONFIRMED cycles completed

Populate from actual measurement. Do not publish until you have a genuine baseline plus at least one completed cycle, a baseline alone gives a reader nothing to assess and invites them to wait rather than enquire.

Aggregate, across all published engagements

Publish once you have three or more clients with two or more cycles each. Before that the sample is too small to mean anything, and saying so is better than implying otherwise.

MeasureValue
Engagements publishedN, NOT YET AVAILABLE
Measurement cycles completedN, NOT YET AVAILABLE
Interventions loggedN, NOT YET AVAILABLE
Interventions preceding a statistically distinguishable improvementN, NOT YET AVAILABLE
Interventions with no change detectedN, NOT YET AVAILABLE
Interventions preceding a declineN, NOT YET AVAILABLE

That last row exists because it will happen, and a ledger that never records one isn't being kept honestly.