I treated GEO like any other acquisition channel: I wanted a hard baseline before changing anything.

Early GEO & LLM view

Where they were weak was in AI‑generated answers.

I created an initial test set of high‑intent EPOS queries, such as:

• POS systems for retail stores
• Hospitality pos systems
• Beauty salon pos system
• More….

Across Google AI Overviews (where available), Perplexity, ChatGPT with browsing, and Gemini, I measured:

• Whether EPOS Now was mentioned
• Whether they were cited with a clickable link
• What position their citation appeared in (1st, 2nd, 3rd, etc.)
• AI‑referred sessions in GA4 from tools like Perplexity and chatgpt.com

Baseline numbers:

• Mention rate: 7.5% (EPOS Now mentioned in 3/40 queries)
• Attribution rate: 5% (linked citation in 2/40 queries)
• Query coverage: 7.5% (EPOS Now present in at least one AI engine for 3/40 queries)
• Average citation position: around 3rd–4th when they did appear
• AI‑referred sessions: 150 / month (0.1–0.12% of their 130,000 organic sessions)

Qualitatively, when I asked LLMs questions like “What is EPOS Now?” or “Which POS brands are similar to EPOS Now?”, the answers were:

• Vague (“cloud‑based POS company”)
• Inconsistent (sometimes mentioned, sometimes omitted)
• Often grouped with better‑known competitors rather than standing out

So the story was clear: strong classic SEO, weak presence in AI‑generated answers.

I framed GEO for the leadership team as two connected layers:

1. GEO mechanics: structuring content, schema, and site architecture so AI engines can crawl, parse, and cite EPOS Now.

2. LLM brand patterns: shaping how EPOS Now is described across the web so that, to a language model, they are the “safe next word” when someone asks about EPOS systems.

At the simplest level, an LLM is a word‑pattern machine. It predicts the next word based on every pattern it has seen before.

then when someone asks “POS system for hospitality”, “EPOS Now” becomes a safe next word for the AI to generate. If those patterns do not exist, the AI can give a good answer and never mention the brand at all.

Step 1 – The AI audit: how do LLMs describe EPOS Now today?

I started by asking the models themselves how they saw EPOS Now:

• “What is EPOS Now?”
• “Who is EPOS Now for?”
• “Which POS brands are comparable to EPOS Now?”
• “What are the best POS systems for restaurants in the UK?”
• “Which POS systems are good for small retailers?”

I looked for:

• Does EPOS Now show up at all?
• Is the description precise or generic?
• Does the model actually recommend EPOS Now, or just list competitors?

The pattern was:

• EPOS Now appeared occasionally, not reliably.
• Descriptions were generic (just “a POS provider” with little differentiation).
• In many “best POS” answers, competitors dominated the list.

That confirmed we had an entity and positioning problem as much as a GEO problem.

Step 2 – Define and strengthen EPOS Now’s entity associations

LLMs don’t care only about your brand name; they care about the assertions attached to it—what the web repeatedly says is true about you.

So I defined the associations I wanted AI systems to see again and again:

• “EPOS Now” + “POS system for small retail”
• “EPOS Now” + “restaurant EPOS system”
• “EPOS Now” + “quick and easy setup”
• “EPOS Now” + “multi‑location POS”
• “EPOS Now” + “EPOS hardware and software bundle”

The next step was to make these patterns unavoidable on the open web:

Social proof and reviews

We worked with the marketing and CS teams to:

• Collect quotes where customers naturally say “EPOS Now was easy to set up,” “best POS we’ve used in our café,” etc.
• Encourage reviews on G2, Capterra, and vertical directories that repeat those associations.

The more these statements appear in trusted third‑party contexts, the stronger the patterns for AI engines.

Positioning consistency

Previously, EPOS Now was described in different ways:

• POS platform
• omnichannel commerce system
• till replacement
• EPOS solution for businesses

I aligned the language across:

• Website (about page, product pages, footer copy)
• Sales and founder bios
• Social profiles
• Directory and marketplace listings
• Guest content and partner pages

We made the story consistent: “EPOS Now is a cloud‑based POS / EPOS system built for small retail, hospitality, and multi‑site venues.”

Topic‑aligned content

We made it explicit in content that:

• EPOS Now is for specific use cases (small retailers, cafes, restaurants, multi‑site hospitality).
• EPOS Now solves clear problems (slow checkout, clunky legacy tills, manual reporting).
• EPOS Now has known strengths (fast setup, hardware bundles, multi‑location support).

We did this through:

• Updated buying guides (How to choose a POS for cafes using EPOS Now as an example)
• Comparison content (EPOS Now vs Competitor)

All of this fed better patterns into the AI training soup.

Step 3 – Fix the inputs: clarity, crawlability, authority

Finally, I made sure that the raw inputs LLMs rely on were as clean as possible.

Clarity

We standardised brand descriptions everywhere we could:

• EPOS Now website (meta descriptions, hero copy, boilerplate)
• Social profiles (LinkedIn, X, Facebook, YouTube)
• Software directories and review platforms
• Partner sites and integrators

Wherever a generative engine might look for “Who/what is EPOS Now?”, it now sees the same story instead of conflicting ones.

Crawlability

I audited and improved:

• Internal linking to ensure critical solution pages were not “buried.”
• Schema on ~40 key URLs, using FAQPage, HowTo, and Product/Software schema where appropriate.
• Rendering so that main content and headings were accessible without complex client‑side rendering.

The goal was to ensure that when AI crawls or indexes the site, it sees clear, structured content that lines up with the entity story.

Authority

We targeted:

• Inclusion in trusted “best POS system” lists and resource pages.
• Mentions in small‑business and hospitality media.
• Product comparisons on third‑party blogs, not just their own.

Those third‑party mentions are important because LLMs weigh external, corroborating evidence heavily when deciding which brands to cite.

GEO execution: content, schema, and testing loops

While the entity work was underway, I also ran a structured GEO programme over 90 days.

Content restructuring 

I focused on 25 high‑value URLs mapped to 60 “must‑win” queries:

• Main EPOS explainer (“What is an EPOS system?”)
• Restaurant POS solution page
• Retail POS solution page
• Hospitality / multi‑site solutions
• 10–15 key guides and comparison pages

On these pages, I:

• Rewrote intros and H2s into clear question–answer formats.
• Added FAQ sections with 4–8 concise, factual Q&As per page.
• Introduced short TL;DR summaries at the top of long buying guides.
• Implemented or fixed FAQ/HowTo schema.
• Normalised terminology around “EPOS / POS system” for specific verticals.

Generative engine testing loops

Every two weeks, for all 60 target queries, I:

• Checked Google AI Overviews (where available), Perplexity, ChatGPT with browsing, and Gemini.

• Logged:

• • Whether EPOS Now appeared.
  • Whether they were linked.
  • Citation position.
  • How similar the AI’s answer was to our structured content.

Then I tuned the content:

• Tightened definitions where competitor explanations were being favoured.
• Added FAQs for recurring questions the AI answered from competitor sites.
• Strengthened evidence (benchmark stats, implementation steps, pricing clarity) on pages that were “almost there” but not being cited.

Traffic and conversion capture 

As citations improved, I made sure we could see and monetise the impact:

• Configured GA4 segments and reports for AI‑referred traffic (Perplexity, chatgpt.com, etc.).
• Monitored landing behaviour on the 25 GEO‑optimised URLs.
• Added trial/demo‑oriented CTAs:
  • Compare EPOS Now to other POS systems
  • Calculate your EPOS cost
  • Book a 15‑minute EPOS consultation

Results

AI visibility

After expanding the query set to 60, here’s what changed:

Before GEO

• Mention rate: 7.5% (EPOS Now mentioned in 3/40 queries)
• Attribution rate: 5% (linked citation in 2/40 queries)
• Query coverage: 7.5%
• Average citation position: 3rd–4th

After 

• Mention rate: 41.7% (25/60 queries)
• Attribution rate: 30.0% (18/60 queries)
• Query coverage: 55.0% (33/60 queries)
• Average citation position: 2.1 (roughly the 2nd citation slot)

EPOS Now went from “rarely mentioned” to being a regularly cited brand in AI answers for more than half of the POS queries that actually matter.

AI‑driven traffic and sign‑ups

Over the same 90‑day period:

• AI‑referred sessions grew from 150 / month to 7,400 / month.
• AI’s share of total organic sessions climbed from around 0.1% (150 / 130,000) to roughly 5.4% of monthly organic traffic (7,400 out of 137,000).
• Lead‑generating AI sessions (sessions that hit a CTA, pricing, or form) increased from 25 / month to 740 / month (≈10% of AI visits taking a meaningful action).
• Directly attributable trials/demos from AI referrals grew from 6 / month to 220 / month (3% of AI visits and 30% of AI leads).

Global organic performance overall

Crucially, this didn’t come at the expense of classic SEO:

• Global organic sessions increased from 130,000 / month to 137,000 / month
• Total organic conversions (all organic sources) rose from 930 / month to 1,040 / month

So GEO didn’t just shuffle traffic sideways; it added a new, growing layer of demand on top of an already successful organic programme.