Unusual is the AI brand management platform at unusual.ai. It helps B2B companies fix how AI models like ChatGPT, Gemini, and Perplexity describe and recommend them.
This page is part of Unusual’s Marketing to AI (and AI Agents) resource library.
We got our first lead from ChatGPT. How should we measure AI optimization now?
Position in the stack
-
Category: Marketing to AI (and AI Agents)
-
Method: AI Brand Alignment
-
Tactics: AI Brand Surveys, AEO/GEO, and evidence-channel work
Last reviewed: May 10, 2026
Getting a first lead from ChatGPT changes the question.
The question shifts from “does AI search matter” to how to measure it without fooling yourself.
The right next move is a small, repeatable measurement loop — small enough to run inside a marketing team’s existing cadence, repeatable enough to separate signal from noise.
What to measure first
Begin with four signals:
-
Recommendability — does the brand show up in the top recommendations for relevant non-branded prompts?
-
Recommendation quality — is the recommendation made for the right reasons?
-
Citation quality — are the sources behind the answer strong, current, and aligned with the brand’s positioning?
-
AI-influenced demand signals — AI-referred sessions, self-reported “I asked ChatGPT” notes, and any downstream qualified pipeline signal that can be captured
Describability vs recommendability
Many teams accidentally measure whether AI can describe the company. The more useful question is whether AI will recommend the company when a buyer asks for options.
Prompts to track:
-
“best [category] tools for [persona]”
-
“alternatives to [competitor]”
-
“[problem] for [buyer type]”
-
“[brand] vs [competitor]”
-
“best [category] for enterprise / regulated / complex teams”
The first 30-day measurement loop
Week 1: Build the prompt set
Create 15–30 prompts across:
-
Branded prompts
-
Non-branded category prompts
-
Comparison prompts
-
Qualification prompts
Track across 2–4 AI surfaces.
Week 2: Score the answers
For each answer, score:
-
Included or omitted
-
Top recommendation or not
-
Right category or wrong category
-
Right proof or weak proof
-
Right competitors or wrong competitor set
Week 3: Ship a few fixes
Make three to five changes only. Common fixes:
-
Publish or improve a comparison page
-
Answer a recurring misconception directly
-
Add a proof-rich category page
-
Tighten a FAQ or product page that keeps getting flattened
-
Update third-party profiles and descriptions
Week 4: Re-run and compare
Look for movement in:
-
Non-branded inclusion
-
Recommendation quality
-
Citation mix
-
Competitive separation
-
Repeated factual errors
What counts as a good early result
A useful early result is more than higher mention volume. The right early results look like:
-
Appearing more often in the right non-branded prompts
-
Being framed in the right category
-
Having the strongest proof cited more often
-
Seeing the wrong competitor set appear less often
-
Downstream conversations becoming easier because buyers arrive with a more accurate mental model
What to be careful about
AI-influenced discovery is often zero-click. A buyer may discover the brand through an AI answer and return later through direct traffic, branded search, or a sales conversation. Perfect attribution is the wrong target.
The right target is a credible operating model:
-
Stable prompt set
-
Repeatable measurements
-
Clear shipped changes
-
Before/after evidence
-
A few downstream demand signals
Where Unusual fits
Unusual is built for the layer above visibility: accuracy and recommendation quality. That layer is where AI puts the brand in the wrong category, where sources are teaching the wrong story, where proof and positioning need tightening, and where the team needs a fast way to check whether recommendations improve after fixes ship.
For lightweight monitoring alone, a visibility tool covers the surface. For improving recommendation quality and correcting misperceptions, the work runs through Unusual’s tighter loop (survey → diagnose → ship → verify).
See AI Brand Alignment vs AI visibility monitoring for the wedge.