Where does your brand stand when AI recommends?
When someone asks ChatGPT, Gemini, Claude or Perplexity what to choose, they don't get ten links: they get three or four names. Today there's no guaranteed spot for your brand: it can be recommended, fall behind competitors, or not show up at all.
Presence
Does it appear?
Position
Where does it rank?
Competition
How does it compare?
A new competitive space is emerging
Brands are already being recommended and compared inside AI —whether they measure it or not. That spot isn't assigned by budget: it's assigned based on the information the model finds about your brand.
The most common mistake is thinking that spot is out of your reach. It isn't. The thing is, almost no brand is measuring it — and whoever measures first defines the terrain.
A monthly index of how AI recommends your brand
It's not another tool to do GEO. It tells you where and how your brand is positioned in AI, versus your competitors, and where the opportunities are to gain relevance. Independent measurement, with the same discipline as a price index.
Four things, every month
Every month we run a battery of queries against the leading models, with web search enabled, in your market's real language. Out of that comes:
Share of Recommendation
How often each brand appears and in what position within the shortlist the model builds, month over month and against competitors.
Attribute map
Which attributes each model assigns to you —and which it assigns to your competitors.
Source map
Which sites it cites as authorities. That's where you can actually intervene to move the needle.
Inaccuracies
What incorrect information is circulating about your products today —and gets fixed in weeks.
This is what your position in AI looks like
Not a text report: a comparable number and a series that moves. An example of the kind of read we deliver each month.
Illustrative data · not a real brand
Repeating the same question three times returned the same first brand only 38% of the time.
The models aren't deterministic. A single measurement is a dice roll formatted as a PowerPoint.
That's why it's an index, not a snapshot: we measure systematically and continuously to build a comparable historical series month over month. The value is in the series, not the single run.
Measurement with lab-instrument discipline
You approve what gets asked before we measure. We bring the measurement, the map and the order of priorities.
Query battery design
900 unique queries per market, in real local language. You approve them before we measure.
Fixed profile
Configuration frozen all year, 4 providers with web search. March comparable to September.
Measurement
Each query 3 times against 4 models. Auditable raw data + a manual sample on the real interfaces.
Map & plan
Not a list of findings: a task list ordered by expected return.
Verification
Each quarter we measure again and check whether what was implemented moved the needle.
It's not another tool to do GEO
The global market calls it AI visibility / GEO. We're opening our own subcategory: algorithmic recommendation audit.
| AI Recommendation Index | What already exists |
|---|---|
| Delivers the comparable historical series | An ad-hoc study: a single snapshot, indefensible |
| The work done + a signed prioritized plan | A SaaS sells you access to a dashboard you operate yourself |
| We measure, we don't execute: judge, not player | A GEO agency executes and grades its own work |
| We measure what the machine answers | Social listening measures what people say |
From an anecdote to a KPI you can defend
You stop flying blind
In a channel where the purchase is already being decided. The cost of not looking grows with adoption; it doesn't hold steady.
You know where to intervene
The source map turns scattered spend into targeted spend. It doesn't ask for new budget: it redirects the existing one.
You fix what's false
AI states incorrect things about your products. It's attributable loss and it's fixable in weeks.
You find open space
There are situations where the model recommends no one. Whoever fills the gap first wins the default recommendation.
You build the series
The day this channel gets monetized, whoever has twelve months of history will be the only one who knows what that spot is worth. You can't build the series backwards.
You get a number
Share of Recommendation is a KPI: you measure it, compare it, set a target and put it on a dashboard.
Brands that don't control where the decision happens
CPG & retail
Categories organized around consumption occasions, dense competition and retailer private labels fighting for the spot.
"What should I get for…?"Financial services
Banking, cards, insurance: pure high consideration, explicit comparison, a decision-maker with their own budget.
"Which card is best for me?"Travel & hospitality
The best fit: a decision-maker with a budget and a return the client itself can compute, via intermediary commissions.
"Where should I stay in…?"Also applies to OTC health & pharma, automotive, education, telcos and electronics.
What people usually ask
What is the AI Recommendation Index?
A monthly, independent index that measures how AI models (ChatGPT, Gemini, Claude and Perplexity) recommend your brand versus competitors, which sources they draw that recommendation from, and what to fix first. It's not a dashboard: it's a comparable historical series plus a prioritized intervention plan.
How is it different from an AI visibility or GEO tool?
A SaaS tool sells you access to a dashboard you operate yourself. We deliver the work done: we design the query battery in your market's real language and situations, we measure, we rebuild the source map and we end in a prioritized plan. And we don't execute: we're the independent measurement layer that shows where to work and then audits whether it worked. Judge, not player. If you want to see how Epical operates, it's in the system.
Why is a single measurement not enough?
Because the models aren't deterministic: repeating the same question three times returned the same first brand only 38% of the time. A single measurement is statistically indefensible. The value is in the series: we measure systematically and continuously —20,000+ queries per measurement, in real local language, repeated against the four leading providers— so one month is comparable to the next.
What kind of brands is it for?
High-consideration categories, where the decision can be delegated to an AI assistant, where the brand doesn't control the point of decision, and with presence in two or more markets. It fits especially in CPG and retail, consumer financial services, travel and hospitality, and also OTC health, automotive, education, telcos and electronics.
Does the index attribute direct sales?
No. The service measures the system's behavior —what AI recommends and in what position— with a fixed basket of queries under constant conditions, like a price index. That discipline is what makes the numbers comparable month over month. We also run a manual control sample on the real interfaces and report the gap. It's the same rigor we apply across our work at Epical.
If AI recommends brands, brands need to start measuring it.
Let's see what AI is recommending in your category
We start with a full measurement: where your brand shows up today, against whom and from which sources. Leave your details and we'll set up the first read.
- Independent measurement — judge, not player.
- Designed in your market's real language and situations.
- Epical · social intelligence for C-level in LATAM.