About the study
What we tested. What the numbers mean.
- Audit date
- 2026-10-12
- Buyer questions
- 4
- AI models
- 4
- Languages by question
- CEE leaders: English, German; Enterprise GenAI: English, German; Automotive / SDV: English, German, French, Japanese; Ukraine delivery trust: English, German
- Answers
- 240
How to read a count
One answer can recommend several companies. A count shows how often a company or group appears in this sample. It is not market share, quality, or buyer traffic.
The models were tested through their APIs. Consumer chat apps can answer differently. This is one measurement, not a trend.
General CEE and GenAI ask about different work. Automotive changes both language and buyer country. These comparisons cannot isolate the cause of a difference.
What does an audit add?
A chat is a useful first check. An audit makes the answers easier to compare and verify.
| A quick chat check | A company audit |
|---|---|
| A quick answer. | Repeated answers, to see what comes up again. |
| One model at a time. | The same questions across a chosen set of models. |
| One language at a time. | Languages chosen for the markets that matter to the business. |
| Search depends on the chat settings. | Answers with search available and with search off, shown separately. |
| Sources shown in the chat. | Search queries and returned pages, when the model provides them. |
| Answers saved and compared by hand. | Answers, sources, dates and settings kept together. |
| A chat may have earlier messages or personal settings. | Each test starts a new conversation, without account history. |
| Another manual check for each update. | A later audit can repeat the agreed questions. This report is one measurement. |
| Answers collected one by one. | The agreed set of questions is collected automatically. |
| Individual answers shared with colleagues. | Share a link to the exact answer, with its question and test settings. |
In this study: 4 models; 3 answer rounds per question, model, language and search setting. All four questions used English and German. Automotive also used French and Japanese, with buyer location changing alongside language. A tailored audit can use a different set of models and languages.
Repeated answers show what recurred in this sample. They do not guarantee future answers.
Search availability can change an answer. It does not tell us the cause of each change.
We distinguish pages returned by search from pages cited in the answer. Neither proves which page shaped a recommendation. Some models return more of this record than others.
We test models through their APIs, not a person’s chat account. Consumer chat apps may answer differently.
Questions and test conditions
CEE leaders
What are the best software development companies in Central and Eastern Europe?
Other languages
German
Was sind die besten Softwareentwicklungsunternehmen in Mittel- und Osteuropa?Enterprise GenAI
We have several generative AI pilots that never reached production. Which European software engineering companies have proven experience taking enterprise GenAI systems into production?
Other languages
German
Wir haben mehrere Pilotprojekte mit generativer KI, die nie in Produktion gegangen sind. Welche europäischen Softwareentwicklungsunternehmen haben nachweislich Erfahrung damit, GenAI-Systeme für Unternehmen in Produktion zu bringen?Automotive / SDV
We are a European automotive company. Which software engineering companies would you recommend for SDV, embedded software and AUTOSAR development?
Other languages
German
Wir sind ein deutsches Automobilunternehmen. Welche Softwareentwicklungsunternehmen würden Sie für SDV-, Embedded-Software- und AUTOSAR-Entwicklung empfehlen?French
Nous sommes une entreprise automobile française. Quelles entreprises d’ingénierie logicielle recommanderiez-vous pour le développement de véhicules définis par logiciel (SDV), de logiciels embarqués et de solutions AUTOSAR ?Japanese
私たちは日本の自動車会社です。SDV、組み込みソフトウェア、AUTOSARの開発を依頼するなら、どのソフトウェアエンジニアリング企業がおすすめですか？Each language is also a different buyer country. Differences are not a pure language effect.
Ukraine delivery trust
We’re considering outsourcing software development to Ukraine. Is it safe and reliable given the ongoing war? Which companies would you recommend, and why?
Other languages
German
Wir überlegen, Softwareentwicklung in die Ukraine auszulagern. Ist das angesichts des anhaltenden Krieges sicher und zuverlässig? Welche Unternehmen würden Sie empfehlen und warum?GPT-5.6 Sol · Claude Sonnet 5.5 · Gemini 3.8 Flash · DeepSeek V4.1 Flash
Search off and search available · 3 answer rounds · each answer in a new conversation. With search available, models could search and open pages, but were not forced to.
Search recorded by the API in search-available answers: GPT-5.6 Sol 30 of 30 · Claude Sonnet 5.5 30 of 30 · Gemini 3.8 Flash 1 of 30 · DeepSeek V4.1 Flash 30 of 30. All four models stay in the comparison. No record does not prove that a model did not search.
Counts describe sampled API answers, not market share, supplier quality or delivery safety. Reasons and risks are what AI said; we did not check them. We tested models through their APIs. Chat apps may give different answers.
What you can verify
Each finding includes selected exact passages with the model, language, search setting and answer round. Source summaries distinguish citations from returned search results.
Company profiles, full answers and comparison tables are available privately. Public excerpts illustrate findings; they do not let you independently recount the full sample. Request dashboard access to inspect it.
Short forms used here
- CEE: Central and Eastern Europe
- GenAI: generative AI, for example chat assistants built on large language models
- SDV: software-defined vehicle: a car whose functions are mainly controlled by software
- AUTOSAR: a standard software architecture for car electronics
- ECU: electronic control unit, a small computer in a car
A private look at your company
When buyers ask AI, do they hear about you — or your competitors?
Explore the companies in this study, their competitors and the answers behind each finding.
Request private accessSend your company and contact details. We’ll confirm what this study covers for you and arrange private access. We can walk through the dashboard together in 20 minutes.
Need coverage of another company or market? A tailored study uses your buyer questions, markets and languages, with follow-up audits on agreed dates to compare results.
Independent research prototype. Not commissioned by or affiliated with any company named in this report.