The AI Lab

We run the experiments, so you don't have to guess.

Every week we test how AI platforms read, rank and recommend — then publish exactly what moved the needle and what didn't. Real prompts, real data, no vibes.

42
Experiments run
3,800+
Prompts tested
6
Platforms tracked
Weekly
New findings
Featured experiment
CompleteApril 2026 · AEO
113
Prompts tested
0%
Brand appeared
6
Platforms
Experiment 001

What 113 AI prompts revealed about one brand's visibility crisis.

We took a mid-market SaaS brand with strong Google rankings and asked ChatGPT, Gemini, Perplexity, Claude, Grok and AI Overviews 113 real buyer questions. It surfaced in zero of them. Here's the teardown of why — and the 90-day plan that fixed it.

AEOCase teardown6 platforms
Read the finding →
2 experiments live now
CompleteExp 004 · GEO

Do third-party reviews move AI recommendations? We seeded 40.

We placed 40 authentic reviews across forums and review sites, then tracked how AI described the brand before and after.

+62%
Brand mentions in answers
Read finding
LiveExp 002 · Paid Media

Which content format gets cited most inside ChatGPT ads?

Testing comparison tables vs narrative vs bullet answers across live ChatGPT ad placements.

Live
Week 2 of 6
Follow along
LiveExp 006 · Agents

Can a support agent resolve tickets with no human? 30-day test.

A production support agent handles live tickets end to end; we track auto-resolution and escalations.

71%
Auto-resolved so far
Follow along
UpcomingExp 003 · AEO

We asked AI the same question 50 ways. Here’s what changed.

Mapping how phrasing, intent and specificity shift which brands an AI names in its answer.

Soon
Starts next week
Get notified
How we experiment

A method, not a hunch.

Every experiment follows the same loop, so the findings are repeatable — and so you can run them yourself.

01

Hypothesis

We start with a specific, testable claim about how AI reads or ranks — not a vague theme.

02

Controlled test

We isolate one variable and run identical prompts across every major AI platform.

03

Measure

We score visibility, citations and share of voice with the same rubric every time, so results compare.

04

Publish

Win or lose, we write it up — method, data and what we’d do next. You can replicate it.

Want the next finding before your competitors do?

Get every experiment in your inbox the day it ships. No spam, just data.

Subscribe