Experiments
Phase I · VIS-001 to VIS-005
Visibility and citation
Which changes to a page actually raise the chance that an AI answer cites it, and do they hold up outside a benchmark?
- A positive result
- A change that lifts citation rate against a matched, unchanged page on at least three AI engines, with a 95% confidence interval clear of zero.
- A negative result, also reported
- The lift appears on one engine only, fades within the observation window, or is no better than an ordinary content refresh.
VIS-001 · In progress
Replicating the GEO tactics on live pages: citations, quotations and statistics, outside the benchmark
VIS-002 · In progress
Direct-answer formatting: does leading with the answer change how often a page is cited?
VIS-003 · In progress
Schema markup and AI citation: does structured data change anything an answer engine shows?
VIS-004 · In progress
Freshness: does a substantive update raise citation rate, or does a new date do the same?
VIS-005 · In progress
Search rank as a predictor: how much of AI citation does ordinary ranking already explain?
Phase II · MEA-001 to MEA-005
Measurement and variance
How stable is an AI visibility score, and how many runs does a prompt need before the number means anything?
- A positive result
- A prompt-panel method whose score repeats within a declared margin across days, accounts and locations.
- A negative result, also reported
- Answer variance swamps the signal, so a week-to-week change in visibility cannot be told apart from noise.
MEA-001 · In progress
Run-to-run variance: one prompt, one engine, thirty runs
MEA-002 · In progress
Prompt wording: how far does a paraphrase move the brands named?
MEA-003 · In progress
Location and language: does the same question get different brands in Hyderabad and New York?
MEA-004 · In progress
Referral blind spots: how much AI-driven traffic does analytics record as direct?
MEA-005 · In progress
Do AI visibility trackers agree with each other on the same brand?
Phase III · TRU-001 to TRU-005
Representation and accuracy
When an AI describes a brand, how often is it right, and can the brand do anything to correct it?
- A positive result
- A documented content change that reduces false statements about a brand on matched prompts, and holds at re-test.
- A negative result, also reported
- Errors persist whatever the brand publishes, or a correction only moves the error somewhere else.
TRU-001 · In progress
Brand fact audit: scoring AI answers against verified product facts
TRU-002 · In progress
Citation support: does the cited page actually say what the answer claims?
TRU-003 · In progress
Correction latency: how long does a fixed fact take to reach AI answers?
TRU-004 · In progress
Comparison prompts: how does AI frame a challenger brand against the category leader?
TRU-005 · In progress
Whose voice wins: do reviews and forums outweigh a brand's own website?
Phase IV · AIM-001 to AIM-005
AI Marketing and business outcomes
Does AI visibility change what a business actually cares about, and does any AIO tactic beat plain, good SEO?
- A positive result
- An AIO programme that beats an SEO-only control on qualified leads, with the comparison declared before it starts.
- A negative result, also reported
- SEO fundamentals alone do just as well. The lab reports that and revises the hypothesis.
AIM-001 · In progress
Visitor quality: AI referrals against organic search on conversion
AIM-002 · In progress
Mention without a click: does being named lift branded search?
AIM-003 · In progress
AIO against SEO-only: a matched-page trial
AIM-004 · In progress
AI-assisted content: does it earn citations as often as content written by hand?
AIM-005 · In progress
Agents as an audience: can an AI shopping agent finish the buying path on a product site?
Let's connect
Want a brand or a site in one of these experiments?
I'm always happy to connect with people who are curious about marketing, building products, or exploring new ideas.