Slice: Generative Engine Optimization for AI Visibility | Qoulomb Case Study
GENERATIVE ENGINE OPTIMIZATION FOR SLICE

Optimising for the answer, not the ranking.

More buyers now ask an AI before they ask a search engine. Slice wanted to shape what those systems say about the brand: how often it is mentioned, how it is recognised as an entity, and whether the sentiment attached to it is one a language model will repeat. So we went off-site, and built authority where the models actually read.

Client
Slice
Industry
Fintech
Focus
Generative Engine Optimization
Window
6 months
1,000+
Reddit comments placed across six months
500+
Quora answers published across the same window
GEO
Off-site authority, not traditional ranking, as the core lever
LLMs
Stronger mentions and entity recognition across AI answers
The Client

Slice wanted the AI systems on side.

The brief was not about rankings. It was about strengthening generative engine optimization: improving brand sentiment, increasing AI mentions and influencing the answers that language models generate about the brand, without leaning primarily on traditional SEO to do it.

Industry
Fintech
Objective
Strengthen GEO and influence AI-generated answers about the brand
What Success Looked Like
More AI mentions, better entity recognition, improved sentiment in AI responses
The Deliberate Choice
Off-site authority building rather than primarily on-site SEO
Primary Surfaces
Reddit, Quora, digital and paid PR, guest posts, brand mentions
The Underlying Logic
Language models reward trusted, widely referenced brands, so authority had to be built where they read
The Challenge

You cannot rank your way into an AI answer.

Traditional SEO wins a position on a results page. It does very little to change what ChatGPT or Gemini says when a user asks for a recommendation. That answer is shaped by how widely and how favourably a brand is referenced across the wider web, which is a different problem needing a different approach.

Answers, Not Links

When buyers ask an AI system for options, they receive a synthesised answer rather than a list of links. Being on page one does not guarantee being in that answer, or being described well within it.

AI
the surface that mattered here

Entity Recognition

For a model to recommend a brand confidently, it has to recognise that brand as a distinct, well defined entity. Weak or inconsistent recognition across the web means weak, hedged mentions in the answers.

Entity
recognition needed strengthening

Sentiment Carries Over

Language models absorb the tone of what is written about a brand. If the prevailing sentiment across forums and articles is mixed, that hesitation surfaces in what the AI tells a prospective customer.

Trust
the currency of an AI recommendation

An Off-Site Problem

Almost none of this could be solved on the brand's own website. The work had to happen across the platforms and publications that language models draw on when they form a view.

Off-site
where the real work sat
The Strategy

Four phases.
Build authority where models read.

The primary focus was off-site GEO and authority building. Rather than chasing rankings, we worked to improve how AI systems understand, mention and recommend the Slice brand, across the exact surfaces those systems learn from.

Phase 01

Community Presence

Reddit and Quora
  • Large scale Reddit optimization, over 1,000 comments in six months
  • Quora optimization, over 500 answers in the same window
  • Presence built in the discussions language models draw on most heavily
Phase 02

Digital and Paid PR

Publications
  • Digital PR campaigns to expand favourable coverage
  • Paid PR across digital publications for reach and credibility
  • Coverage placed where it strengthens the brand's reference footprint
Phase 03

Authority and Mentions

Backlinks and reach
  • Guest post acquisition on relevant, credible sites
  • High authority backlinks to reinforce trust signals
  • Brand mention expansion across the wider web
Phase 04

AI Entity and Sentiment

GEO core
  • AI entity optimization so models recognise Slice as a distinct, well defined entity
  • Brand sentiment management to shape the tone models absorb
  • Content optimised specifically for citation by AI systems
The Results

The work, and what it moved.

Execution Scale

Off-site authority building across six months
REDDIT 1,000+ comments placed in six months QUORA 500+ answers published in six months
Execution Scale
What the six month programme actually involved
ActivityExecution
Reddit1,000+ comments in 6 months
Quora500+ answers in 6 months
PRPaid PR and digital publications
AuthorityGuest posts and high authority backlinks
FocusBrand sentiment and AI answers
These are execution volumes rather than outcome metrics. Because the goal was influence over AI-generated answers, the impact shows up in how systems describe the brand rather than in a single dashboard figure.

What the Campaign Moved

The impact was qualitative, on how AI systems treat the brand
Stronger AI visibility More present across generative engines Increased mentions across LLMs Referenced more often in answers Better entity recognition Recognised as a distinct, defined brand Improved brand trust More favourable tone in AI responses
Instead of chasing rankings alone, the campaign focused on improving how AI systems understand and recommend Slice. The result was stronger AI visibility, more mentions across language models, better entity recognition and improved brand trust in AI-generated responses.
Why It Worked

Five decisions that shaped the answers.

Generative engine optimization is still new enough that most brands are not doing it at all. This worked because it treated the AI answer as the actual product and built toward it deliberately.

01

Off-Site by Design

The decision to lead with off-site authority rather than on-site SEO matched the problem. AI answers are shaped by the wider web, so that is where the effort went.

→ Effort concentrated where models actually learn
02

Presence Where Models Read

Reddit and Quora are among the most heavily referenced sources for language models. Over 1,000 comments and 500 answers put the brand into those conversations at scale.

→ 1,500+ community touchpoints in six months
03

Entity Before Opinion

AI entity optimization ensured models recognise Slice as a distinct, well defined entity. A model cannot recommend confidently what it cannot identify clearly.

→ Cleaner, more confident brand mentions
04

Sentiment Managed, Not Left to Chance

Because models absorb the tone of what is written, brand sentiment was actively managed rather than assumed. Trusted brands get recommended, hedged ones get qualified.

→ More favourable tone in AI responses
05

Content Built to Be Cited

Material was optimised specifically for citation by AI systems, so the brand appears as a source rather than an afterthought when a model composes an answer.

→ Increased mentions across LLMs
Behind the Work

A different scoreboard for a different game.

The hardest part of GEO is that the win does not show up as a rank you can screenshot. It shows up in what a language model says when nobody from the brand is in the room. That demands a different kind of patience and a different kind of measurement.

How it worked
The premise

Answers Over Rankings

The goal was set from the start: influence AI-generated answers rather than climb a results page. That reframing decided everything that followed.

Community

Into the Conversations

Large scale Reddit and Quora work, over 1,500 contributions in six months, placed the brand inside the discussions that language models weight most heavily.

Reach

PR and Authority

Digital and paid PR, guest posts and high authority backlinks widened the brand's reference footprint across credible, well read sources.

Recognition

Entity and Sentiment

AI entity optimization and active sentiment management shaped both whether the models recognise the brand and how warmly they describe it.

The result

A Brand the AI Recommends

Stronger AI visibility, more mentions across LLMs, sharper entity recognition and improved trust in the answers those systems generate about Slice.

Reddit
1,000+
Comments across six months
Quora
500+
Answers across six months
Reach
PR
Paid and digital publication coverage
Signals
Trust
Brand sentiment actively managed for AI
The outcome
Cited
Recognised, mentioned and recommended more readily across generative engines
1,500+

Combined Reddit and Quora contributions in six months, placed where language models learn what to say about a brand.

What does the AI say about your brand?

Ask ChatGPT or Gemini to recommend a provider in your category. If your brand is missing, hedged, or described in someone else’s words, that is a generative engine optimization gap, and it is only going to matter more.