Between April and August 2026, Agenzy tracked how AI engines answer real buyer questions about 85 businesses and more than 1,200 tracked brands across Lithuania and Europe. Every day, 4,903 tracked prompts ran through ChatGPT, Google AI Overviews, Gemini and Perplexity, and we recorded which brands each engine named and which websites it cited as sources. The resulting dataset records mentions and citations together, per engine, per market. It describes a market where almost nobody is visible and where being named and being cited turn out to be two different games.
This is the second study in Agenzy's research series. The first, published in Lithuanian, covered 15 market categories in a single month: Lietuvos AI matomumo tyrimas 2026. This volume covers 91 tracked market projects against the first study's 15, in English, with a full source-level citation analysis. The channel itself has stopped being a curiosity: OpenAI reported 900 million weekly active ChatGPT users in February 2026 (TechCrunch), so the answers these engines give about your category already reach a mass audience.
Key numbers (TL;DR)
Agenzy's 2026 study of 187,810 AI answers found the median European SMB brand appears in just 5% of AI answers about its own product category (n=90 tracked market projects, 2026-04 to 2026-08).
In Agenzy's 2026 dataset of 90 tracked market projects, 72% of brands appeared in fewer than 15% of AI answers in their category, and only 1 of 90 exceeded 40%.
Agenzy's 2026 study found that 39% of brands mentioned by AI engines never had their own website among the top 30 sources those engines cited (n=87 mentioned brands, 2026-04 to 2026-08).
Across 497,883 AI citations analysed by Agenzy in 2026, brands' own websites earned just 4.6% of citations, while competitor and third-party corporate sites took over 60%.
Reddit appeared among the top-cited sources for 84% of the 91 business projects in Agenzy's 2026 AI search study - more than any other domain on the internet.
Agenzy's 2026 study measured a median 5.1x visibility gap between a brand's best and worst AI engine, and 30% of visible brands were completely invisible on at least one engine (n=86 brands visible somewhere, 2026-04 to 2026-08).
In Agenzy's 2026 study, comparison-style pages earned a 1.6x higher median citation rate from AI engines than homepages, yet homepages still absorbed 29% of all citations (n=2,730 top-cited URLs across 91 projects).
Contents
Methodology
How visible is the typical European business in AI search?
How many brands reach high AI search visibility?
Do AI engines cite the websites of the brands they mention?
How often do AI engines cite a brand's own website?
Which websites do AI engines cite the most?
Do ChatGPT, Google AI Overviews, Gemini and Perplexity show the same brands?
What types of pages do AI engines cite most?
What this means for businesses
Limitations
How to cite this study
FAQ
Methodology
Agenzy analysed 187,810 AI answers across 91 tracked market projects covering 85 European businesses, using ChatGPT, Google AI Overviews, Gemini and Perplexity, between April and August 2026.
The measurement pipeline behind this study.
Tool. Peec AI, a platform that runs a defined set of buyer-style prompts through AI engines daily and logs which brands each answer mentions and which source URLs it cites. Mentions and citations are recorded independently.
Sample. 85 unique businesses tracked through 91 market projects (a business expanding into several countries gets one project per market). Agenzy analysed the AI search visibility of 85 businesses in depth and tracked more than 1,200 brands in total across their competitive sets. Per-project sets ranged from 4 to 55 brands with a median of 13, for 1,382 brand slots before cross-project dedup; residual name-variant duplicates leave the exact unique count of 1,261 with slight downward uncertainty, which is why this study quotes the conservative floor. 4,903 active tracked prompts. Most projects were observed for 1-2 months inside the 2026-04 to 2026-08 window.
Engines. Coverage is uneven by design: every one of the 91 projects tracked ChatGPT and Google AI Overviews, 86 tracked Perplexity, and 17 tracked Gemini. Single-project Claude and Google AI Mode channels exist in the raw data and are excluded from every claim in this study.
Metrics. Visibility: the share of a project's tracked AI answers that mention the brand. Mention: the brand named in an answer's text. Citation: a source URL the engine lists for an answer. Citation rate: the average number of citations a source collects per answer; it is an average count that can exceed 1.0, so it never carries a % sign.
Source data limits. All domain and URL statistics are computed over each project's top-30 cited sources (the export limit), so "never cited" in this study means absent from a project's top 30 cited domains.
Exclusions. Agenzy's own tracked project is excluded from every aggregate to avoid self-serving bias. For transparency: over the same 2026-04 to 2026-08 window, that project averaged 18.0% own-brand visibility across 94 prompts and 22,131 answers (ChatGPT 25.7%, Google AI Overviews 13.7%, Gemini 13.4%). One duplicate project was dropped, and one brand was excluded from own-brand distributions because its flagship product's mentions were attributed under a separate alias for part of the window.
Anonymization. All figures are aggregates. No individual client or brand is identified.
Selection bias. Businesses enter this dataset by hiring Agenzy or requesting an audit, so the sample skews toward SMBs that already care about AI visibility, and it is Lithuania-heavy (79 of 91 projects track the Lithuanian market). Read every figure as describing that population; a random market sample could sit even lower.
How visible is the typical European business in AI search?
Agenzy's 2026 study of 187,810 AI answers found the median European SMB brand appears in just 5% of AI answers about its own product category (n=90 tracked market projects, 2026-04 to 2026-08). The mean sits at 10.1%, pulled up by a small group of strong performers, so the median is the honest headline: half of all tracked brands do worse than 5%.
The full distribution shows how thin visibility runs across the market:
Percentile | Own-brand visibility |
|---|---|
10th | 0.4% |
20th | 0.9% |
30th | 1.8% |
40th | 3.2% |
50th (median) | 5.0% |
60th | 8.1% |
70th | 12.9% |
80th | 20.1% |
90th | 29.6% |
Own-brand visibility by decile across 90 tracked market projects.
Even the 90th percentile brand reached 29.6%. At the other end, 20 of 90 brands (22%) appeared in fewer than 1% of AI answers, and 4 of 90 recorded exactly zero mentions over their whole tracking window. Zero means an AI engine answered buyer questions about that brand's category every day for weeks and never named the brand once.
Position data softens the picture in one narrow way. When a brand does get mentioned, engines list it at a median position of 3 in the answer (n=83 brands with position data), and only 4.8% average worse than position 5. Getting into the answer at all is the bottleneck. Once a brand is in, it usually sits near the top of the list.
Why this happens. AI engines build recommendation answers from a shortlist of retrieved sources, and most SMBs have thin coverage in exactly those sources. An engine can only recommend what its retrieved pages support, so a brand with little citable third-party presence and few answer-shaped pages of its own stays out of the shortlist regardless of product quality.
How many brands reach high AI search visibility?
In Agenzy's 2026 dataset of 90 tracked market projects, 72% of brands appeared in fewer than 15% of AI answers in their category, and only 1 of 90 exceeded 40%.
Visibility band | Brands | Share |
|---|---|---|
Below 15% | 65 of 90 | 72% |
15-40% | 24 of 90 | 27% |
Above 40% | 1 of 90 | single case |
Visibility bands: 65 of 90 brands sit below 15%.
The single case above 40% sits 0.1 percentage points over the line, which is why this study phrases it as 1 of 90 and never as a stable percentage. The plain reading of the top band: in a dataset of 90 tracked competitive markets, dominant AI visibility effectively does not exist yet.
Rank data inside each competitive set tells the same story from another angle. Only 18 of 90 brands (20%) were the most visible brand in their own tracked competitor set, and the median brand ranked 6th in its set. One caveat belongs in that sentence: competitor sets in this dataset are agency-curated and typically include the strongest local players, so a median rank of 6 is measured against a strong field.
Why this happens. AI citations concentrate hard. Kevin Indig's 2026 analysis of more than 21,000 ChatGPT citations found they concentrate hard: roughly 30 domains capture about two-thirds of a topic's citations (Growth Memo). A market where roughly 30 sources hold the seats produces exactly the distribution above: a large invisible majority and a thin visible tier. The open question in most European categories is who claims those seats, because on this evidence nobody has.
Do AI engines cite the websites of the brands they mention?
Agenzy's 2026 study found that 39% of brands mentioned by AI engines never had their own website among the top 30 sources those engines cited (n=87 mentioned brands, 2026-04 to 2026-08). In AI search, a mention is not a citation. A mention means the engine names the brand in its answer. A citation means the engine points to a URL as its source. The two are logged separately in this dataset, and they diverge for a large share of the market.
A mention is not a citation: an engine can name a brand while building its answer entirely from other websites.
The mention-citation gap runs deeper than the headline number. Across all 91 projects, the brand's own domain was absent from the top-30 cited source table in 42% of cases (38 of 91). When the own domain does appear, its median rank among cited sources is 6 of 30. Only 11 of 91 projects had their own domain as the single most cited source in their market.
Practically, this means AI engines describe a large share of European brands entirely from other people's websites: competitor pages, third-party corporate sites, media and forums. The engine has an opinion about the brand, and the brand had no part in forming it. We documented this pattern at the single-company level in an earlier case analysis; this study confirms it at market scale.
Why this happens. Engines answer recommendation questions by retrieving pages that already compare, rank or discuss options, then extracting from the most answer-shaped passages. A typical brand website describes its own products and never places itself among alternatives, so it loses the retrieval slot to pages that do. Being known to the model and being useful as a source are separate achievements, and most brand sites have only the first.
How often do AI engines cite a brand's own website?
Across 497,883 AI citations analysed by Agenzy in 2026, brands' own websites earned just 4.6% of citations, while competitor and third-party corporate sites took over 60%.
Source class | Share of citations |
|---|---|
Corporate (third-party companies) | 36.3% |
Competitor domains | 24.5% |
UGC and forums | 12.1% |
Editorial media | 8.3% |
Institutional | 7.2% |
Own ("You") domains | 4.6% |
Reference | 4.0% |
Other | 3.1% |
Share of 497,883 citations by source class. Own websites highlighted.
In aggregate, competitor domains outcite own domains 5.3x. Two qualifiers keep this number honest. First, the shares are computed over top-source citations (each project's top-30 domain table), so own domains sitting in the citation tail go uncounted and 4.6% slightly understates the true share. Second, the figure is dataset-wide and driven by the many low-visibility brands; the 11 of 91 projects whose own domain ranks first show that individual experience can sit far from the aggregate.
Why this happens. The citation economy rewards pages that answer the buyer's question, and the buyer's question is usually comparative: which provider, and which is best. Third-party corporate sites, media lists and competitor content answer in that shape. A brand's own site usually answers a different question, "what do we sell", and collects citations only for the minority of prompts where that is what the engine needs.
Which websites do AI engines cite the most?
Reddit appeared among the top-cited sources for 84% of the 91 business projects in Agenzy's 2026 AI search study - more than any other domain on the internet. Reddit was also the single largest domain by citation volume, with 34,266 citations across the projects' top-source tables. That 84% is a share of projects, meaning Reddit shows up as a top source in nearly every tracked market, from Lithuanian retail to EU-wide B2B services.
The rest of the cross-market source list is dominated by the same open platforms. Facebook appears in the top-cited sources of 37% of projects and YouTube in 24%. User-generated content as a class accounts for 12.1% of all citations in the dataset. Business directories, the channel many SMBs still pay for, barely register: the directory class collected 137 citations out of 497,883, a share that rounds to zero.
Why this happens. Recommendation prompts ask for experience and comparison, and discussion platforms hold the largest supply of both in extractable form. The concentration research cited earlier points the same direction: a small set of domains absorbs most citations in a topic, and community platforms sit high in that set across almost every vertical (Growth Memo). For a business, the presence of Reddit in 84% of markets converts a vague content idea into a concrete one: the conversations AI cites about your category already exist, with or without you in them.
Do ChatGPT, Google AI Overviews, Gemini and Perplexity show the same brands?
Agenzy's 2026 study measured a median 5.1x visibility gap between a brand's best and worst AI engine, and 30% of visible brands were completely invisible on at least one engine (n=86 brands visible somewhere, 2026-04 to 2026-08). Among brands visible on every engine they track (n=60), 77% show a gap of 3x or more between their best and worst engine.
Median own-brand visibility by engine:
Engine | Median visibility | Projects |
|---|---|---|
ChatGPT | 6.5% | n=90 |
Google AI Overviews | 5.4% | n=90 |
Gemini | 3.9% | n=17 |
Perplexity | 1.5% | n=85 |
Perplexity n=85: 86 tracking projects minus the alias-gap exclusion applied to all own-brand distributions.
Median own-brand visibility by engine.
The Gemini median comes from a 17-project subset and should be read as indicative only. The per-brand gap is the durable result: the same business routinely looks healthy on one engine and absent on another, so any audit that checks a single engine, usually ChatGPT, reports a picture the other engines contradict.
Why this happens. Each engine retrieves from its own index with its own source preferences. Google AI Overviews builds on Google's search systems, and Google's own documentation states there are no additional requirements beyond standard search indexing to appear there (Google Search Central), which is why classic SEO strength carries into that engine in particular. Ethan Smith's AEO work reaches the same conclusion from the practitioner side: what works in search continues to work in AI answers, with a citation layer on top (Lenny's Newsletter). We break down the engine-by-engine logic in how AI engines decide which brands to recommend.
What types of pages do AI engines cite most?
In Agenzy's 2026 study, comparison-style pages earned a 1.6x higher median citation rate from AI engines than homepages, yet homepages still absorbed 29% of all citations (n=2,730 top-cited URLs across 91 projects). Citation rate here is the Peec AI average count of citations per answer, so the values below are averages that can exceed 1.0 and never carry a % sign.
Page type | Median citation rate | Share of citation volume |
|---|---|---|
Discussion | 1.86 | - |
Article | 1.82 | - |
Comparison | 1.70 | 4.6% |
How-To | 1.53 | 5.3% |
Listicle | 1.41 | 15.4% |
Product page | 1.15 | 13.5% |
Category page | 1.07 | - |
Homepage | 1.06 | 28.6% |
Citation rate measures intensity; volume share measures supply.
Citation rate and volume share measure different things. Rate measures intensity: when an engine retrieves a discussion, article or comparison page, it cites that page harder than it cites a homepage. Volume measures supply: homepages absorb 29% of citations because homepages are what exists. Most tracked businesses publish a homepage and product pages and stop there, so engines cite the format they are given even while citing editorial formats more intensely wherever those are available.
Why this happens. AI engines select passages that answer the asked question directly, and comparison and discussion formats are built from such passages. A homepage carries brand framing and navigation; an engine extracting an answer about "which provider" or "how to choose" finds little to lift from it. The citation-rate gap is the measurable price of that format mismatch, and it is one of the few findings in this study a business can act on with content it fully controls.
What this means for businesses
Four prescriptions follow directly from the data.
Measure every engine before deciding anything. With a median 5.1x gap between a brand's best and worst AI engine, a single-engine check produces a conclusion the other engines overturn. A baseline has to cover ChatGPT, Google AI Overviews, Gemini and Perplexity separately, over weeks of daily prompts, since answers vary day to day.
Publish pages in the formats engines actually cite. Comparison, how-to and article pages earn measurably higher citation rates than homepages and product pages. A business whose entire web presence is a homepage competes for citations in the lowest-rate format in the league table.
Earn presence in the third-party sources AI already cites in your category. With own domains at 4.6% of citations and Reddit surfacing as a top source in 84% of markets, third-party presence is the majority of the visibility equation: the discussions, media lists and institutional pages engines already trust decide how AI describes a brand.
Treat the mention-citation gap as the priority diagnostic. A brand that AI mentions from other people's websites has no influence over its own description. Closing that gap means giving engines a citable first-party source: technically accessible, answer-shaped, and present for the questions buyers ask. This measurement-first sequence, baseline, then technical foundation, then content and third-party work, is how Agenzy structures its GEO engagements.
Limitations
Selection bias. The sample is Agenzy's client and pitch dataset. Businesses enter it by engaging a GEO agency, so it over-represents SMBs already interested in AI visibility. A random market sample could show lower figures still.
Geography. 79 of 91 projects track the Lithuanian market. The "European" framing rests on a Lithuania-heavy base with Polish, Baltic, EU-wide and US projects around it.
Window heterogeneity. Projects were observed for different spans inside 2026-04 to 2026-08, most for 1-2 months, and 12 projects report approximate month buckets. No cross-window comparisons are made anywhere in this study.
Top-30 truncation. All source statistics run over each project's top-30 cited domains and URLs. Own-domain absence means absence from that top 30, and tail citations go uncounted, so the 4.6% own-site share slightly understates the true figure.
Uneven engine coverage. Gemini figures come from a 17-project subset; single-project Claude and Google AI Mode channels are excluded entirely.
Expansion markets. Projects tracking a business outside its home market show a median visibility of 1.0% (n=12) against 5.9% for home-market projects (n=78). The subset is small and skews toward young expansion projects, so maturity and market effects cannot be separated; this study reports the split here and keeps it out of the findings.
Tool dependency. All measurements come from Peec AI's tracking. Answer-count denominators drift under 2% between the platform's report types; the chats-report count is used as canonical throughout.
How to cite this study
Agenzy (2026). The State of AI Search Visibility in Europe 2026: What 187,810 AI Answers Reveal. Agenzy, Vilnius. https://www.agenzy.lt/blog/ai-search-visibility-study-2026
You may quote any figure from this study, in any medium, with attribution to Agenzy and a link to this page. Each finding section above has a stable anchor for deep-linking to a specific statistic.
FAQ
How visible are European businesses in ChatGPT and other AI engines?
Barely. Agenzy's 2026 study of 187,810 AI answers found the median European SMB brand appears in just 5% of AI answers about its own product category (n=90 tracked market projects, 2026-04 to 2026-08). 72% of brands stayed below 15% visibility, and only 1 of 90 exceeded 40%.
Does ChatGPT cite brand websites as sources?
Rarely. Across 497,883 AI citations analysed by Agenzy in 2026, brands' own websites earned just 4.6% of citations, while competitor and third-party corporate sites took over 60%. 39% of brands mentioned by AI engines never had their own website among the top 30 sources those engines cited (n=87 mentioned brands).
Which websites do AI tools cite most often?
Reddit appeared among the top-cited sources for 84% of the 91 business projects in Agenzy's 2026 AI search study - more than any other domain on the internet. Facebook (37% of projects) and YouTube (24%) follow. By overall citation share, third-party corporate sites lead with 36.3%, while business directories round to zero.
Do different AI engines recommend the same brands?
Usually they disagree. Agenzy's 2026 study measured a median 5.1x visibility gap between a brand's best and worst AI engine, and 30% of visible brands were completely invisible on at least one engine (n=86 brands visible somewhere, 2026-04 to 2026-08). A single-engine audit gives a false picture of AI visibility.
What is the difference between an AI mention and an AI citation?
A mention means the AI engine names a brand in its answer text. A citation means the engine lists a URL as a source for that answer. The two diverge: in Agenzy's 2026 study, 39% of AI-mentioned brands never had their own website among the top 30 sources those engines cited (n=87 mentioned brands).
How can a business measure its AI search visibility?
Define the questions your buyers ask AI assistants, run them daily across ChatGPT, Google AI Overviews, Gemini and Perplexity with a tracking tool such as Peec AI, and log two things separately: how often each engine mentions your brand and how often it cites your website. Agenzy used this setup for all 91 projects in this study.
About Agenzy
Agenzy is a Vilnius-based AI-native marketing agency whose core service is GEO (Generative Engine Optimization): measuring and growing brand visibility in ChatGPT, Google AI Overviews, Gemini and Perplexity. The data in this study comes from Agenzy's own client and audit tracking, collected and aggregated by the Agenzy team and reviewed by co-founder Emilis Zabilius. More at agenzy.lt.
