A Ballou x Qwairy study on what it takes for a company to exist in generative AI’s answers
Introduction
Type ‘what’s the best accounting software for an SME?’ into ChatGPT, Gemini or Perplexity and you won’t get a list of links to sift through. You get a single answer, generated there and then. GEO (Generative Engine Optimisation), the practice of understanding, and wherever possible, influencing how a brand appears in those answers, much as SEO shapes where you rank in a list of links. GEO sits alongside SEO rather than replacing it, and its rules are still being written.
To understand these rules, we started with companies that have already succeeded against traditional measures of growth. A company that has reached unicorn status has found its buyers, grown rapidly and convinced investors. The question we set out to answer is whether that commercial success carries into what ChatGPT or Gemini say about a company when someone asks?
Ballou, a communications agency for innovative companies, and Qwairy, a GEO platform that measures and improves brand visibility in generative AI, set out to answer it. We ran this analysis across France, Germany, Spain and the UK: putting more than 800 buying-intent questions to six AI platforms, and tracking 73 European scale-ups and unicorns, each tested on its home market.
Our aim was to work out what makes some companies appear in AI answers, and to turn that into a model any company chasing visibility in these new search habits can use.
How we ran the study
- 73 European scale-ups and unicorns studied: 20 in France, 20 in Germany, 16 in Spain, 17 in the UK, each one asked only questions relevant to its own market.
- 876 questions asked between 12 August and 1 September 2026, in 7 collection waves, mostly phrased the way a real buyer would (‘what’s the best solution X for Y’), split between 716 questions targeting a specific company (without naming it), 71 direct company-to-company comparisons, and 89 broader questions about an entire sector.
- 6 AI platforms tested for every question: ChatGPT, Gemini, Google AI Mode, Google AI Overview, Copilot and Perplexity.
- 34,709 complete and usable answers collected.
Visibility varies hugely by market
Giving equal weight to all 73 companies in the panel, a measure we call the standardised visibility rate, a company was mentioned in 50.22% of the answers to questions about it.That average covers four very different national realities:
These four panels aren’t a ranking of each market’s economic health. They test different companies, sectors and questions from one country to the next, so a lower score for Spain or the UK doesn’t mean their scale-ups perform worse. It simply means they were mentioned less often on this particular set of questions.
#1: visibility is built across a mix of sources
Across the whole study, we looked at what kind of source an AI draws on when it names one of these companies.
To be clear, ‘commercial’ doesn’t mean ‘e-commerce site’. It’s a broad category including the company’s own website, a product page, or even a competitor’s site cited as an example, not only the unicorn in question.
The most consistent lesson across all four markets is that no single type of source wins on its own. The most-cited companies appear across several registers at once, including their own site, specialist and mainstream press, and – where relevant – a public institution or trade body. Relying solely on press coverage, or solely on your own website, leaves a void for competitors to step into and fill.
Where LinkedIn, YouTube and Wikipedia fit in
Within social networks and reference sites, three names recur often:
YouTube accounts for between 42.5% and 62.1% of all citations classed as ‘social’, depending on the market,, and LinkedIn between 18.4% and 24.6%. Together they account for most of this category. France has the strongest presence of both YouTube and Wikipedia across the four markets, and the UK has the strongest presence of LinkedIn. Wikipedia’s low frequency doesn’t make it useless, however. Rather, It’s just rarely cited as a direct source within an answer.
#2: rankings and comparisons are the format that wins
Among all the articles whose format could be identified with confidence, one type of content stands out: the ranking, comparison or shortlist, the kind of article that reads ‘Top 10 accounting tools’ or ‘Best business banks’. Averaged across the four markets, it accounts for 51.1% of articles with an identified format. Funding announcements, market analyses and single-company articles all trail some way behind.
Editorial format, share of articles with an identified format, by market
| Format | Global | France | Germany | Spain | UK |
|---|---|---|---|---|---|
| Ranking / comparison / selection | 51.1% | 45.1% | 58.0% | 56.7% | 48.4% |
| Funding announcement | 11.1% | 8.9% | 8.5% | 11.2% | 14.6% |
| Market analysis | 7.0% | 10.5% | 4.5% | 5.2% | 6.7% |
| Single-company article | 7.0% | 10.0% | 6.5% | 3.8% | 6.7% |
| Guide / explainer | 6.5% | 7.6% | 5.0% | 8.6% | 5.5% |
| Product / service launch | 5.8% | 7.5% | 6.4% | 3.6% | 5.1% |
| Study / report | 3.8% | 4.2% | 3.9% | 2.6% | 4.0% |
| Company / product page | 3.8% | 3.0% | 4.1% | 4.6% | 3.8% |
| Personal profile | 2.2% | 1.5% | 2.4% | 1.7% | 3.1% |
| Award / recognition | 0.6% | 0.6% | 0.3% | 0.4% | 1.0% |
| Academic paper | 0.6% | 0.6% | 0.2% | 0.6% | 0.9% |
| Opinion piece | 0.6% | 0.5% | 0.3% | 1.2% | 0.4% |
Base: all references in the qualification file with an identified editorial format
Ballou x Qwairy AI Visibility Study | Editorial format, share of articles with an identified format, by market
Qwairy x ballou
Three national patterns emerge from this. Germany and Spain are the two markets most concentrated on rankings and comparisons (58.0% and 56.7%). France, is the market where market analysis (10.5%) and single-company articles (10.0%) carry the most weight, meaning longer and more substantial content than the average. The UK has the highest share of funding announcements (14.6%), likely reflecting a particularly busy VC news cycle over the study period.
The most-cited media, across all markets
Across all questions targeting a specific company, five titles account for the most citations across the study Reuters (668 references), Handelsblatt (591), EU-Startups (582), Sifted (557) and Emprendedores (536). The overall ranking sits on top of very different media landscapes from one market to the next.
None of these four national rankings shares a single title with any other, which shows how local media visibility remains. A PR strategy built for one market doesn’t automatically transfer to the other three.
So, in practice rankings alone aren’t enough. A place in a strong sector comparison is the most useful position to hold, but lasting visibility also needs a regular presence across other formats: a well-placed funding announcement, a market analysis piece that cites the company as a reference point, or a dedicated article that cements its credibility. A solid AI reputation comes from the combination of these formats over time, not any one of them.
#3: which AI you ask matters as much as the company itself
The six platforms behave very differently. Averaged across the study, Google AI Mode is the platform most likely to cite a company (58.7% of answers), ahead of ChatGPT (56.5%), Google AI Overview (54.3%), Gemini (53.5%) and Copilot (48.5%), with Perplexity the most cautious across all four markets (37.4%).
Mention rate by Al platform and market
| Platform | Global | France | Germany | Spain | UK |
|---|---|---|---|---|---|
| Google AI Mode | 58.7% | 76.3% | 67.0% | 40.5% | 43.1% |
| ChatGPT | 56.5% | 69.8% | 63.8% | 39.4% | 48.2% |
| Google AI Overview | 54.3% | 65.4% | 63.9% | 39.4% | 42.2% |
| Gemini | 53.5% | 66.3% | 64.3% | 38.7% | 39.3% |
| Copilot | 48.5% | 62.9% | 53.5% | 34.3% | 39.0% |
| Perplexity | 37.4% | 48.0% | 41.0% | 25.4% | 31.5% |
Share of questions in which the platform names the company, among questions targeting it directly
Ballou x Qwairy AI Visibility Study | Mention rate by AI platform and market
The gap between the best- and worst-performing platform for the same company matters more than any single row in that table. Averaged across the 73 companies, that gap reaches 30.1 points, and 48 of the 73 companies show a gap of at least 25 points. Germany has the widest average gap of the four markets at 34.5 points, and Spain the tightest at 25.0 points. Some companies show a very wide gap: Black Forest Labs (Germany) spans from 91.2% of mentions on Google AI Mode to 20.0% on Perplexity; Ankorstore (France) from 90.3% to 28.8%; Payflow (Spain) from 73.1% to 18.5%; 9fin (UK) from 60.4% on Google AI Overview to 0% on Copilot.
Some AIs also prefer commercial sites, others favour the press
The citation rate isn’t the whole story. The six AIs also differ in the type of source they favour when building an answer.
Gemini leans most heavily on commercial sites across all four markets,(61.4% globally). Perplexity, draws most heavily on media sources in Germany, Spain and the UK, and on Copilot in France. Google AI Mode and Google AI Overview carry a higher share of social citations than the other platforms (7.5% and 8.1%), which partly explains why they also cite panel companies most often: they draw on a wider pool of sources. Citing a lot of media doesn’t mean citing a lot of companies. Perplexity draws on relatively more media sources than the other AIs while remaining the most cautious in terms of citation rate, across all four markets.
#4: when a company is cited, it’s usually cited alone
When an AI’s answer cites one of the panel companies, it rarely cites several at once. Across the study, 76.9% of answers that cite a target company cite only one from the panel in a structured way, rising to 95.2% in the UK and 96.2% in Spain. Being cited therefore often means being the single company put forward for that question, rather than one name among many. This is not proof of being ‘recommended’ over rivals, and the data does not support that claim, but it does mean that being cited, or missed, carries more weight than it might appear.
#5: some companies are simply invisible on certain platforms
Across the panel, 8 of the 73 companies had at least one platform that returned zero mentions across every question about them, which is 14 company-platform combinations out of 438. A zero does not mean a company can never be cited on that platform, only that it was not on this set of questions in this window. It is a reminder that visibility can be genuinely uneven between platforms, not only ‘slightly weaker’ on some.
On stability, repeating the questions across seven waves over the three-week collection period moved the overall mention rate only between 50.45% and 52.34%, a standard deviation of 0.56 points, with slightly more movement in Germany and the UK than in France and Spain. This says nothing about visibility after the study ended, but within the window the results were not the product of one unusually good or bad day.
Market by market
One overall figure on sector-level questions, the generic ones not aimed at a specific company: across the study, at least one company from the relevant national panel appears in 69.9% of these answers. That rate reaches 85.3% in France and 75.1% in Spain, and falls to 65.3% in Germany and 53.4% in the UK, the lowest generic sector coverage of the four markets. This matches the pattern in the company-level rates, where the UK and Spain are less covered under this protocol whether the question targets one company or a whole sector.
United Kingdom: the widest gap between sectors
The UK panel of 17 companies, including Fresha, Nothing, MUBI, Tide and Isomorphic Labs, shows a standardised visibility rate of 38.4%. The UK study also covers the widest range of sectors of any market, six against four or five elsewhere, and it holds the largest gap between the best- and worst-covered sector of any single market in the study.
Frontier AI (85.0%) and Fintech (40%) sit at opposite ends of the UK market and of the whole study, the widest sector gap recorded within a single market. The UK is also the only market where the leading platform is not Google AI Mode. ChatGPT cites UK companies most often here (48.2%), a local quirk. Two sources stand out more than elsewhere: LinkedIn at 1.0% of citations and forums such as Reddit at 1.4%, the highest shares recorded for both anywhere in the study.
France: where public institutions carry almost as much weight as the press
The 20 French scale-ups studied, including Brevo, Verkor, Pennylane, Sorare and Harmattan AI, show a standardised visibility rate of 64.0% and the highest share of institutional sources of any market, at 9.6% of all citations (commercial sources 55.5%, media 13.3%). Blogs and video content feature more strongly here than elsewhere, with YouTube at 2.3% of all French citations and Wikipedia at 0.7%, the highest shares for both across the four markets.
The five French sectors show very different profiles:
Fintech reaches full coverage, with every question surfacing at least one French unicorn, driven by the trade body France FinTech, cited more often than any individual media outlet in the sector. Deeptech Industriel is the least-covered French sector (62.8%), and one where public voices such as entreprises.gouv.fr, the SATT network and Bpifrance carry almost as much weight as the specialist press.
Germany: highly uneven coverage across sectors
The German panel of 20 companies, including Flink, Isar Aerospace, DeepL, Helsing and Black Forest Labs, shows a standardised visibility rate of 58.2%, the second-highest of the four markets. German media coverage is also the most concentrated on a single format, with 62.1% of media citations that have an identified format being rankings or comparisons, the highest share of any market studied.
Defence Tech is Germany’s best-covered sector (95%), but German coverage is uneven, running from 95% on Defence Tech down to 39% on Health Tech. French Fintech (100%) and Spanish Quantum Tech (96%) sit ahead of Defence Tech across the whole study. Defence Tech is also the only German sector where media citations (43%) clearly outweigh commercial sources, driven by titles such as deutsche-startups.de and Bloomberg. A few other source types are worth noting for Germany: blogs account for 3.8% of citations, just ahead of social media at 3.3%, with YouTube at 2.1% and LinkedIn at 0.6%.
Spain: Quantum Tech, the only sector where institutional rivals media
The Spanish panel of 16 companies, including Factorial, Fever, Cabify, TravelPerk and Payflow, shows a standardised visibility rate of 35.5%. Spain also has the highest share of commercial sources of any market (58.4%). Media is the second most-cited source category (10.8%), behind commercial, and the share of citations from blogs is the highest of the four markets at 5.1%.
Quantum Tech is the best-covered Spanish sector (96.1%) and the only sector across all markets where institutional sources (23%) come close to media sources (28%), helped in particular by Fundación Bankinter and the innovation agency CDTI. Spain’s most-cited media are distinctive too. Alongside specialist titles such as Emprendedores and El Referente, two of the country’s biggest national dailies, El País and Cinco Días, feature among the five most-cited titles. This pattern does not appear in the other three markets, where the specialist press leads the national general-interest press by some distance.
Methodology, in more detail
One important note for reading every figure in this report: this study is descriptive. It measures what generative AI said, how often, and from which sources, on this specific set of questions and this three-week window. It doesn’t measure audience reach, it doesn’t prove that any given communications action caused a change in visibility, and it doesn’t declare a ‘winning’ company. Every finding in this report should be read as ‘here’s what we observed’, not ‘here’s what works’.
The study was carried out by Qwairy in partnership with Ballou between 12 August and 1 September 2026. Seventy-three European scale-ups and unicorns were selected across four markets (20 in France, 20 in Germany, 16 in Spain, 17 in the UK), each one asked exclusively about questions relevant to its own market. The core question set mirrored genuine questions about finding and choosing solutions (‘what’s the best solution X for Y’) rather than simple factual questions, supplemented by a smaller set of direct comparison questions and broader sector-level questions.
Each question was put to the same six generative AI platforms (ChatGPT, Gemini, Google AI Mode, Google AI Overview, Copilot and Perplexity), in seven passes staggered between 12 August and 1 September 2026 (12, 13, 17, 19 and 24 August, then 27 and 28 August, then 31 August and 1 September), with no question or platform tested more often than any other.. We obtained 34,709 complete answers, and every figure in this report is based on these. A company was counted as ‘cited’ when it appeared explicitly in a complete answer to a question about it. The visibility rate shown for each market is a standardised measure that gives equal weight to each platform within a question, to each question within a company, and to each company within a market, so that no heavily-tested company and no over-represented platform skews the national average.
The 15,316 domains and 51,709 articles cited across all the answers were reviewed and classified by Ballou, first by publisher type (media, commercial, public institution, and seven other categories) and separately by editorial format, applying a cautious approach throughout: a category was only assigned where there was sufficient certainty, leaving a meaningful share of sources unclassified rather than guessed at. Any percentage in this report describing ‘sources with an identified format’ or ‘classifiable domains’ describes this classified subset, not the whole corpus, and should be read accordingly.
This study’s findings are descriptive from start to finish: they measure the frequency and nature of the citations observed, under this protocol and within this time window, without establishing any cause-and-effect link between a given communications action and the visibility measured.
About Ballou
Ballou is a communications agency that designs and delivers strategies in service of its clients’ business goals. We work hand in hand with leadership teams to embed communications at the heart of their commercial strategy. With offices in the UK, France, Spain and Germany, the agency supports major innovative companies and venture capital firms through every key stage of their growth: mergers and acquisitions, IPOs, high-growth phases, branding and positioning, and major crisis management. Find out more at www.balloupr.com. Follow us on LinkedIn.
About Qwairy
Qwairy is the go-to GEO platform for managing brand visibility in generative AI (ChatGPT, Gemini, Perplexity, Copilot and others). Beyond measurement, Qwairy turns this data into prioritised recommendations to improve a brand’s presence in AI answers. Founded by SEO specialists including a Skyscanner alumnus, the platform already supports major brands such as Nissan and Match Group, as well as agencies specialising in digital marketing. Find out more at qwairy.co.
Want to know how your company shows up in AI answers? Let’s talk.