Proof
We publish our numbers. Including the ones that didn't work.
Nobody in this category can verify anybody's claims. Case studies are unfalsifiable, testimonials are unattributable, and screenshots of dashboards prove nothing about causation. So instead of asking you to trust us, we publish the raw material: measured baselines, every intervention with its date, and results with confidence intervals attached, including the interventions that moved nothing at all.
Last updated 31 July 2026
Why
Three things are true about answer engine optimisation in 2026.
Buyer intent is near-universal, Conductor's survey of 250+ digital leaders found 97% reporting positive AEO impact and 94% planning to increase investment.
Execution is nearly absent, only 14% of marketers track AI search performance at all.
And the gap between those two numbers has been filled with unverifiable claims.
There is a documented case of a business paying $50,000 for no citations, no AI referrals and no AI Overview appearances. Google's own John Mueller has linked the acronym push directly to spam and scamming. He was describing a real pattern.
We can't fix that market. We can decline to participate in it.
What we publish
Every engagement we're permitted to publish: baseline share of voice with confidence intervals, every intervention logged with its date, results on a fixed cadence.
Interventions that produced no measurable change. This is the page that can't be faked and won't be copied.
Recurring public research measuring AI visibility across a defined universe of businesses. Full methodology, downloadable data, confidence intervals on every figure.
The rules we publish under
- Client consent first, always. Nothing appears here without written permission. Clients approve their entries before publication and can redact anything commercially sensitive.
- Dates on everything. An undated result is an anecdote.
- Confidence intervals on every figure. If we can't state a margin of error we don't publish the number.
- Correlation stated as correlation. We log interventions with dates so relationships can be assessed. We don't claim causation we can't demonstrate, and platform changes get flagged as a possible explanation rather than quietly absorbed into our results.
- Failures published on the same cadence as successes. Not in an appendix. Not quarterly when the wins are monthly.
- Corrections stay visible. If we get something wrong we amend it and leave the amendment on the record.
The standing offer
If you'd rather your results weren't public, that's completely fine and most clients choose that. But if you're willing to let us publish your baseline, interventions and outcomes, including the disappointing ones, we reduce your fee by 15% for the life of the engagement.
It isn't a discount with a marketing motive attached. Published, dated, honest evidence is the only credible currency in this category, and it's worth more to us than the margin.