Which AI Assistants Actually Matter to Irish Buyers
SEO

Which AI Assistants Actually Matter to Irish Buyers

ChatGPT drives most referrals here, but Gemini's share is five times lower than the European average. Why that changes prompt panels.

If you are building an AI visibility programme for an Irish business and you have to choose which assistants to monitor first, the available Irish data says ChatGPT and then a long way back to everything else. It also says Ireland does not behave like the rest of Europe, and that the most commonly cited numbers measure something other than what most people think they measure.

There is a bigger problem, and it is worth putting up front rather than burying. The Irish assistant data that exists is thin and it is old. Two time points, both from 2025. Anyone presenting it as a current picture of the Irish market in late 2026 is stretching it, including anyone quoting this article.

What the Referral Data Actually Shows

Share of AI chatbot referral traffic, Ireland
Two snapshots, four months apart, and a very concentrated market
March 2025
ChatGPT
87.82%
Perplexity
8.8%
Gemini
2.28%
June 2025
ChatGPT
83.9%
Copilot
7.75%
Perplexity
5.28%
StatCounter Global Stats for Ireland, March 2025 and June 2025, the June figures reported by TechCentral on 11 June 2025 (VERIFIED as referral measurement). Note what changed between snapshots: ChatGPT dipped roughly four points, Copilot appeared in second place where Perplexity had been, and Perplexity fell from 8.8% to 5.28%. In a market this concentrated, second and third place are volatile.

Two things stand out. ChatGPT holds something between 84% and 88% of Irish AI referral traffic across both snapshots, which is dominance on a scale you rarely see in any measured channel. And the ordering below it reshuffled completely in four months.

The practical read is that ChatGPT is not a choice, it is the baseline. Everything else is a judgement call about your specific audience, and the ranking of those alternatives is not stable enough to plan a year around.

Ireland Does Not Follow the European Pattern

Here is the finding that should stop anyone from importing a European benchmark. In the March 2025 Irish data, Gemini held 2.28% of AI referral traffic. The comparable European figure at the time was 11.16%.

Roughly five times lower in Ireland than across Europe. For the assistant built by the company that owns the search engine Irish users overwhelmingly use.

I do not have a verified explanation for that gap and I am not going to construct one. What it establishes is narrower and still useful: a European average is not a proxy for Ireland on this question. If your prompt panel weighting came from a continental report, it is calibrated to a distribution Ireland does not have.

This is the same pattern that appears in AI Overview coverage, incidentally. Ireland sits at 12.00% of tracked queries returning an AI Overview, against 19.10% in the United Kingdom and 20.50% in the United States, per Ahrefs' analysis of 108 million queries published in November 2025. Ireland ranked 44th of 50 countries. On two separate AI metrics, measured by two different organisations, Ireland comes in materially below its nearest comparators.

The Survey Data Says Something Different

Now the complication. CORE Research's AI Connect study, reported in February 2026, surveyed Irish consumers on assistant use and found ChatGPT at 25% weekly use and Meta AI at 24% weekly use.

Meta AI, effectively tied with ChatGPT for weekly usage. Meta AI, which does not appear at all in the Irish referral data.

Both findings can be true simultaneously, and understanding why is the most useful thing in this article. Meta AI lives inside WhatsApp, Instagram and Messenger, where Irish penetration is very high. People use it constantly. It rarely sends anyone to an external website, because the interaction completes inside the app.

So a referral-based instrument sees almost nothing. A usage survey sees a near-leader. Neither is wrong. They are answering different questions.

The same study put overall regular use at 18%, which sits oddly against the Central Statistics Office finding that 42% of Irish internet users had used a generative AI tool in the three months before being surveyed, rising to 65% among those aged 16 to 29. Again, not a contradiction. The CSO measured any use in three months. CORE measured self-described regular use. Different windows, different thresholds, different numbers.

Three Metrics That Keep Getting Merged

Three instruments, three questions, three answers
None of these is a substitute for either of the others
Referral share

Which assistants send clicks to websites?

Measures outbound traffic arriving from an assistant. Blind to any interaction that ends without a click, which is most of them, and blind entirely to assistants that keep users inside an app.

Survey-reported usage

Which assistants do people say they use?

Measures recalled behaviour against whatever threshold the questionnaire set. Sensitive to wording, window and sample. Catches in-app assistants that referral data misses.

Citation frequency

How often is a brand named inside an answer?

The only one of the three that measures commercial visibility in generative search. Requires a prompt panel and repeated runs. No public Irish dataset exists for it.

The third column is the one that matters for a GEO programme and the one with no Irish benchmark. Referral and survey data are useful for deciding which assistants to test. Neither tells you whether you are being cited, and using either as a citation proxy is the most common way this discipline gets oversold.

That third point deserves emphasis. If a supplier shows you assistant market share and implies it tells you something about your visibility, they are showing you a traffic statistic and calling it a citation statistic. Those are not related in any reliable way.

Why Referral Share Is the Wrong Basis for Panel Weighting

The natural instinct is to weight a monitoring panel by referral share: 85% of the effort on ChatGPT, a token amount on the rest. That would be a mistake, for four reasons.

Referral share measures clicks, and your exposure is not clicks. An assistant that names your competitor in an answer and sends nobody anywhere has still influenced a purchase decision. It contributes zero to referral share.

The tail is where enterprise buyers are. Copilot appeared at 7.75% in the June 2025 Irish data, which looks negligible. But Copilot is embedded in Microsoft 365 and Edge, which is the default desktop environment in a very large share of Irish enterprise, including much of the multinational cluster. If you sell business software into Dublin-based EMEA operations, a 7.75% referral share may represent a disproportionate share of your actual buyers.

Rank ordering below first place is unstable. Perplexity was second at 8.8% in March 2025 and third at 5.28% in June. Copilot was absent from the top three and then second. A panel weighted to a snapshot is weighted to noise.

In-app assistants are invisible to the instrument. Meta AI at 24% weekly usage does not register in referral data at all. If your audience is consumer and Irish, that is a substantial blind spot created by the choice of measurement, not by the market.

What Should Drive Panel Weighting Instead

Four inputs, in order of weight.

InputHow to get itWhy it beats market share
Where your own buyers say they lookedAdd a single question to your enquiry form or sales qualification script asking which tools they used before contacting youCategory-specific and yours. A B2B software buyer and a consumer service buyer do not use the same assistants, and no national average captures that difference.
Your customers' desktop environmentAsk, or infer from your existing account baseA Microsoft 365 estate makes Copilot a priority regardless of its national referral share. A Google Workspace estate shifts weight toward Gemini and AI Overviews.
Whether the assistant retrieves the web at all for your query typeTest it. Ask your real prompts and record whether a search was triggeredAssistants that answer from parameters rather than retrieval cannot cite you no matter what you publish. Knowing which is which redirects effort.
National referral and usage dataStatCounter for referrals, published surveys for usageUseful as a tiebreaker and a sanity check. Last, not first, and never as a proxy for citation.

A defensible starting configuration for most Irish businesses: ChatGPT and Google AI Overviews as non-negotiable, Copilot added for anything selling into enterprise, Gemini monitored despite its low Irish referral share because it is fed by the search engine Irish users actually use, Perplexity included where technical or research-heavy buyers matter, and Meta AI tested manually for consumer categories even though it will not show in any dashboard. That is a reasoned configuration, not a market-share allocation. The distinction is the whole point, and it shapes how we scope GEO programmes for Irish clients.

What Nobody Has Measured Here

Four gaps, stated plainly.

There is no Irish citation-share dataset for any assistant. Referral share exists. Usage surveys exist. Nobody has published how often Irish brands are named inside generated answers, by assistant, for commercial queries.

The Irish referral data is a year old. Both StatCounter snapshots are from 2025. Given that the ordering below first place changed completely in four months, extrapolating a 2025 distribution to late 2026 is not defensible. Anyone quoting these figures, including this article, should date them explicitly.

There is no Irish measurement of assistant use split by business versus consumer context, which is exactly the split that would determine panel weighting.

And there is no Irish data on how assistants behave for prompts written in Irish. The question is legitimate for public bodies with statutory obligations, and completely unstudied.

The honest summary is that Ireland has enough data to make a reasoned first choice about which assistants to monitor, and nowhere near enough to claim a market picture. Anyone selling you the second thing is selling you the first thing with confidence added. We work through the measurement side of this in our core GEO practice, and the organic counterpart in our SEO service for Ireland.

The framework for separating these instruments properly is set out in Cited or Silent, and the gated edition is free to download. Our own cross-market citation measurements are published in the AI Citation Rate Report 2026.

Written by Tessar Napitupulu, Founder and CEO of PT Arfadia Digital Indonesia, a member of the Forbes Agency Council, and author of Found Before They Search and Cited or Silent. Arfadia has documented its generative engine optimisation practice since 2023 and works from Jakarta, Bandung and Bali.




Frequently Asked Questions


Which AI assistant matters most in Ireland?

ChatGPT, by a very wide margin, on the referral data available. StatCounter recorded it at 87.82% of Irish AI chatbot referral traffic in March 2025 and 83.9% in June 2025. Nothing else came close in either snapshot. The practical consequence is that ChatGPT is a baseline rather than a choice, and the interesting decisions are about which of the remaining assistants your specific audience actually uses.


Why is Gemini so weak in Ireland?

It holds a much smaller share of Irish AI referral traffic than the European average. StatCounter put Gemini at 2.28% in Ireland in March 2025, against a comparable European figure of 11.16%, roughly five times lower. No verified explanation for the gap has been published and this article does not construct one. What it does establish is that a European average is not a usable proxy for Ireland on this question, so any prompt panel weighted from a continental report is calibrated to a distribution Ireland does not have.


Why does Meta AI appear in surveys but not in referral data?

Because the two instruments measure different things. CORE Research's AI Connect study reported in February 2026 found Meta AI at 24% weekly use among Irish consumers, effectively tied with ChatGPT at 25%. Meta AI operates inside WhatsApp, Instagram and Messenger, where Irish penetration is high, and the interaction usually completes inside the app without sending anyone to an external website. Referral measurement therefore sees almost nothing while a usage survey sees a near-leader. Both findings are accurate.


How current is the Irish assistant data?

Not very, and this is a real limitation. Both StatCounter snapshots are from 2025, in March and June. The ordering below first place changed completely between them: Perplexity went from second at 8.8% to third at 5.28%, and Copilot appeared in second place at 7.75% having been absent from the top three. If the ranking can reshuffle in four months, extrapolating a 2025 distribution into late 2026 is not defensible. Date these figures explicitly whenever you use them.


Should we weight our monitoring by assistant market share?

No. Referral share measures clicks, and your exposure in generative search is not clicks. An assistant that names a competitor in an answer and sends nobody anywhere has still influenced a purchase decision while contributing nothing to referral share. The tail also matters more than its share suggests: Copilot at 7.75% is embedded in Microsoft 365 and Edge, which is the default desktop environment across much of Irish enterprise including the multinational cluster. For a business software vendor, that small share may represent a disproportionate number of real buyers.


What should determine which assistants we monitor?

Four inputs, weighted in this order. First, where your own buyers say they looked, which you can capture with one added question on your enquiry form or qualification script. Second, your customers' desktop environment, since a Microsoft 365 estate makes Copilot a priority and a Google Workspace estate shifts weight toward Gemini and AI Overviews. Third, whether each assistant actually retrieves from the web for your query type, because an assistant answering from parameters cannot cite you regardless of what you publish. Fourth, and last, national referral and usage data as a tiebreaker.


Is there Irish data on how often brands are cited inside AI answers?

No. Referral share data exists and usage surveys exist, but no published dataset records how often Irish brands are named inside generated answers, by assistant, for commercial queries. That is the metric a GEO programme actually needs, and it has no Irish benchmark. It has to be produced on your own prompt panel, which is achievable but means the number will be category-specific rather than comparable to a national figure.


Why do Irish AI adoption figures vary so widely?

Because they measure different populations against different thresholds over different windows. The Central Statistics Office found 42% of Irish internet users had used a generative AI tool in the three months before being surveyed, rising to 65% among those aged 16 to 29, in a release published in December 2025. CORE Research put overall regular use at 18%. One measures any use in a three month window, the other measures self-described regular use. Neither is wrong and they should not be averaged or merged into a single headline figure.


Is Ireland behind on AI adoption compared with the UK and US?

On AI Overview exposure, measurably so. Ahrefs analysed 108 million AI Overview queries and found the feature on 12.00% of tracked Irish queries, against 19.10% in the United Kingdom and 20.50% in the United States, placing Ireland 44th of 50 countries. Combined with Gemini's low Irish referral share relative to Europe, two separate metrics measured by two different organisations both put Ireland below its nearest comparators. Whether that reflects lower adoption, different query composition or something else has not been established.

Sources & References:

  • StatCounter Global Stats, AI chatbot referral share for Ireland, March 2025: ChatGPT 87.82%, Perplexity 8.8%, Gemini 2.28%. Comparable European figure for Gemini at the same period 11.16%. Measures outbound referral traffic from assistants to websites, not citation frequency and not usage.
  • StatCounter Global Stats for Ireland, June 2025, reported by TechCentral on 11 June 2025: ChatGPT 83.9%, Copilot 7.75%, Perplexity 5.28%. Ordering below first place changed materially from the March snapshot.
  • CORE Research, AI Connect 2026 study, reported via adworld.ie on 27 February 2026. Overall regular use of AI among Irish consumers 18%; ChatGPT 25% weekly use; Meta AI 24% weekly use. Definition of regular use not published, and the study is a single commissioned survey.
  • Central Statistics Office, Household Digital Consumer Behaviour 2025, published 5 December 2025, survey conducted in the second quarter of 2025. 42% of internet users had used generative AI in any form in the three months before completing the survey, rising to 65% among those aged 16 to 29, falling to 15% among those aged 60 to 74 and 6% among those aged 75 and over. Generative AI module asked for the first time in 2025.
  • Ahrefs Brand Radar, "Which Countries Have the Most AI Overviews", published 4 November 2025. 108 million AI Overview queries across 50 countries. Ireland 12.00% coverage from 1.4 million tracked keywords, ranked 44th; United Kingdom 19.10%; United States 20.50%.
  • Irish citation-share data by assistant for commercial queries: UNAVAILABLE. No published dataset located.
  • Irish assistant usage split by business versus consumer context: UNAVAILABLE.
  • Assistant behaviour for prompts written in Irish: UNAVAILABLE.
  • Current-year Irish referral share for 2026: UNAVAILABLE at the time of writing. The most recent Irish figures located are from June 2025 and are presented here with that date attached.
0 Comments 0 Comments
0 Comments 0 Comments