New product
Introducing the AI Recommendation Index

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?

Part of Grupo OxeanIntelligence company
22 senior specialists+ proprietary AI
Regional Spanish · PT-BRYour market's real language
Fortune 500 clientsLATAM · MX & BR
Brands that trust Epical
Why now

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.

There's just one question: are you on the shortlist?
What it is

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.

20,000+
queries per measurement, in your market's real language
4
AI models monitored with the same setup all year
12
months of a comparable historical series, not twelve one-off reports
What it measures

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:

01

Share of Recommendation

How often each brand appears and in what position within the shortlist the model builds, month over month and against competitors.

02

Attribute map

Which attributes each model assigns to you —and which it assigns to your competitors.

03

Source map

Which sites it cites as authorities. That's where you can actually intervene to move the needle.

04

Inaccuracies

What incorrect information is circulating about your products today —and gets fixed in weeks.

The deliverable

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.

Share of RecommendationCategory · Mar 2026
Competitor A 34% Your brand 27% Competitor B 22% Competitor C 17%
How much AI recommends each brand and in what position. Your brand, second: with room to gain.
Historical seriesShare of Recommendation · 12 months
Your brand Leading competitor 403020100 31% 27% JFMAMJJASOND
The same measurement, month over month. The value is in the trend — not in a one-off snapshot.

Illustrative data · not a real brand

38%

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.

How it works

Measurement with lab-instrument discipline

You approve what gets asked before we measure. We bring the measurement, the map and the order of priorities.

01

Query battery design

900 unique queries per market, in real local language. You approve them before we measure.

02

Fixed profile

Configuration frozen all year, 4 providers with web search. March comparable to September.

03

Measurement

Each query 3 times against 4 models. Auditable raw data + a manual sample on the real interfaces.

04

Map & plan

Not a list of findings: a task list ordered by expected return.

05

Verification

Each quarter we measure again and check whether what was implemented moved the needle.

What makes it different

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 seriesAn ad-hoc study: a single snapshot, indefensible
The work done + a signed prioritized planA SaaS sells you access to a dashboard you operate yourself
We measure, we don't execute: judge, not playerA GEO agency executes and grades its own work
We measure what the machine answersSocial listening measures what people say
We're judge, not player. That's why the number is worth something. Having the measurer be separate from the executor is what makes the number useful to evaluate the work.
Why it matters

From an anecdote to a KPI you can defend

01

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.

02

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.

03

You fix what's false

AI states incorrect things about your products. It's attributable loss and it's fixable in weeks.

04

You find open space

There are situations where the model recommends no one. Whoever fills the gap first wins the default recommendation.

05

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.

06

You get a number

Share of Recommendation is a KPI: you measure it, compare it, set a target and put it on a dashboard.

Who it's for

Brands that don't control where the decision happens

High-consideration categories, where the decision can be delegated to an assistant, where the brand doesn't control the point of decision, with presence in two or more markets.

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.

FAQ

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.
Tomás Criado, CEO & Founder of Epical
Tomás Criado CEO & Founder · Epical
Featured in Forbes La Nación Infobae
Next step

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.
Request your diagnosis Goes straight to the Epical team. We reply within 24 business hours.