Generative Engine Optimization

Get AI to recommend your store.
With data, not luck.

GEO is the technical and structural work that lets ChatGPT, Claude, Gemini and Perplexity read your catalogue, understand your brand and name it when someone asks what to buy. We audit it, we implement it ourselves, and we measure it every two weeks.

8

Signals audited

4

Engines monitored

15

Days between measurements

1–3

Months per package

What it is

What is GEO?

GEO (Generative Engine Optimization) is the set of technical and structural work that allows artificial intelligence engines to read, understand and recommend an online store when a user asks them what to buy.

When someone types "which trail running shoe brand do you recommend?" into ChatGPT or Perplexity, the model does not return a list of links: it returns an answer with brand names in it. Appearing in that answer is not about bidding more or publishing more articles. It is about your store being readable, verifiable and structured for a machine that decides who to name.

That is the work. It is not content and it is not links: it is data infrastructure.

Traditional SEO

Appearing in a list of results

GEO

Being the answer the model gives

Traditional SEO

Measured in clicks and rankings

GEO

Measured in mentions and citations by name

Traditional SEO

Optimises keywords

GEO

Structures facts the AI can verify

Traditional SEO

Competes for a slot on the page

GEO

Competes to be the trusted source

The method

The 8 signals an AI checks before recommending a brand.

Every LaPregonera audit is built on these eight signals. They are the conditions an AI assistant needs met before it can name a store with confidence. Each one is audited with evidence taken directly from the store, and re-measured every two weeks against the baseline.

01

AI read permission

If your site terms block AI crawlers, nothing else counts. It is the door: while it stays shut, the rest of the work does not exist for the model. It is the first failure we look for and the first one we fix.

02

Readable content

Content has to reach the crawler complete, not depend on the browser building it afterwards. Many stores load critical information in a way the AI never sees: for the model, that information does not exist.

03

Product data

Price, availability, SKU, brand and attributes declared in structured data, not just written in the copy or baked into an image. It is the difference between the model guessing and the model knowing.

04

Readable categories

A category must explicitly declare which products it contains, in a list the AI can walk through. If it does not, the model sees a loose page instead of a catalogue, and cannot answer "what does this store sell?".

05

Discoverable store

An architecture that lets an agent traverse the entire catalogue, with clean paths and no areas left out of reach. What cannot be traversed cannot be cited.

06

Brand identity

Legal entity, address, founding year, contact, logo and description, all structured. These are the "who" and "where" facts a model needs before it trusts a brand enough to recommend it to someone who is about to spend money.

07

Readable ratings

Average rating and real review count, structured and wired to the genuine source. A placeholder value inherited from the theme costs more than it gains: a fake figure, once detected, destroys the trust you are trying to build.

08

Catalogue without duplicates

Categories competing against each other for the same term dilute the signal and split authority. We consolidate into one authoritative category per term, preserving existing traffic with correct redirects.

Scope

We are deliberately narrow. We do the technical and structural work, which is where the real bottleneck sits, and we do it fast.

What we do

Included in the service

AI visibility audit

Where the engines name you today and where they do not, by query and by market, with reproducible evidence.

Structured data

Brand identity, catalogue, product and ratings turned into facts an AI can read.

Technical catalogue architecture

Category structure, duplicate consolidation, redirects and URLs.

Implementation and publishing

We apply and publish the changes ourselves, on a duplicated theme. The client touches nothing.

Fortnightly monitoring

Measurement of mentions and citations across ChatGPT, Claude, Gemini and Perplexity against the baseline.

What we do not do

Out of scope · we advise on it

Content creation

Product copy, About pages, FAQs or claims. We flag what is missing and why; if the client writes it, we structure it.

Digital PR and off-site

Wikipedia, press releases, review campaigns or link building. Not included.

Blog, YouTube and social

We do not produce editorial or audiovisual content as part of the GEO service.

Why we say it this plainly

Because the bottleneck for most stores is not a lack of content: it is that the AI cannot read the content they already have. Fixing that first multiplies everything else.

Packages

Two ways to work together, depending on whether you want the diagnosis or the diagnosis and the execution.

Package 01

Audit

1 month · one-off

Full audit of the 8 signals, with evidence.

Baseline measurement across the 4 engines.

Action plan prioritised by impact.

Results presentation session.

No implementation: the diagnosis is yours and whoever you choose executes it.

Package 02

Audit and implementation

3 months · monthly fee

Everything in the previous package.

Implementation and publishing of the changes by us.

Fortnightly report and review call.

Final dossier with every change, before and after, documented.

Complete visibility measurement history.

01

Discovery

Business, product, markets and problems.

02

Proposal

Package and scope are chosen.

03

Kick-off

Audit and plan presented.

04

Implementation

We apply and publish the changes.

05

Measurement

Every two weeks, same method.

06

Close

Final dossier and history.

Result

From 3 to 5 signals passing in the first implementation cycle.

D2C ecommerce · Recurring consumable product

The store was actively asking AI engines not to use its content. With that block in place, no other signal counted: the catalogue was invisible to any assistant. We opened read permission, made the store traversable for AI agents, and turned the categories into readable catalogues that declare the products they contain. Brand identity and ratings were structured and are in progress.

3→5

Signals passing

1

Implementation cycle

Frequently asked questions

GEO (Generative Engine Optimization) is the technical and structural work that allows artificial intelligence engines to read, understand and recommend an online store. Unlike SEO, which chases positions in a list of links, GEO aims for the model to name and cite your brand directly in its answer.
SEO optimises for appearing in a list of results and is measured in clicks and rankings. GEO optimises for being the answer the model generates and is measured in mentions and citations by name. SEO works on keywords; GEO works on structured facts the AI can verify: who you are, what you sell, at what price and with what real rating.
Technical changes are applied within weeks, but AI engines need to re-crawl and reprocess the store before mentions change. It is not a same-day effect. That is why we measure every two weeks against the baseline. In a three-month implementation, signal movement is visible from the first cycle.
It works particularly well for ecommerce and any business with a product or service catalogue where the user is looking to transact. The more specific the question someone asks an AI, the more value there is in being structured to answer it. We work with companies of all sizes.
ChatGPT, Claude, Gemini and Perplexity. We measure brand mentions and citations in each one using the same method and the same queries every two weeks, so the comparison against the baseline is direct.
No. We work on the existing store and on a duplicated theme, so nothing is published until it has been verified. The client does not have to touch code or migrate platform.
No. We do not write product copy, About pages, FAQs or claims, and we do not do digital PR, press releases or link building. We flag what content is missing and why; if the client writes it, we structure it so the AI reads it correctly.
In two layers. The first is the 8 technical signals, audited with direct evidence from the store and scored as passing, partial or pending. The second is real visibility: brand mentions and citations across the four engines, measured every two weeks with the same set of queries.
It is the most common and the most serious failure, and many stores have it switched on without knowing because it ships by default with the platform or theme. While the block is in place, no other visibility work counts: the model cannot read anything. It is the first thing we audit and the first thing we fix.
Because it is a different service with a different profile. The project is led by an artificial intelligence specialist, not a repurposed SEO team, and the scope is deliberately technical: we do not sell content or links. That is what makes one to three month engagements possible instead of annual retainers.

Is AI reading you,
or ignoring you?

The audit answers that question with evidence, signal by signal, in one month. If you want to know where your store stands before committing to anything, start there.