# Bersyn — Extended Reference > Bersyn records what four named AI assistants — ChatGPT, Claude, Gemini and Perplexity — answer when buyers ask what to buy, in one product category, one buyer market and one language. It reports counts of those recorded answers and keeps every answer as evidence. It does not produce a score, a rank, a share or a trend, and it does not claim to know why a model answered as it did. This document is the extended machine-readable profile. The short form is at https://www.bersyn.com/llms.txt --- ## About This Entity - **Legal name:** Bersyn - **Wikidata:** Q139927512 (https://www.wikidata.org/wiki/Q139927512) - **Pronunciation:** BEHR-sin (one word, two syllables, stress on the first) - **Etymology:** From the Icelandic roots *ber* (bare, exposed) and *sýn* (sight, vision) — bare sight, exposed view - **Founded:** 2026 - **Founded by:** Gissur Þór Rúnarsson - **Headquarters:** Reykjavík, Iceland - **Country:** Iceland - **Industry:** Software · Software as a Service · Measurement and research tooling - **Customer type:** B2B software companies, product marketers, growth teams, and the agencies advising them - **Canonical URL:** https://www.bersyn.com ## Not To Be Confused With Bersyn is regularly confused with similarly spelled names. None of the following are Bersyn, and Bersyn is not affiliated with any of them. - **Berwyn, Illinois** — a city in Cook County, Illinois, United States. - **Berwyn, Pennsylvania** — an unincorporated community in Chester County, Pennsylvania. - **Berwyn Group** — a data-services company headquartered in Beachwood, Ohio. - **Berwyn-Mt Carmel HS** — a secondary school in the Cook County, Illinois area. - **Berwyn (Welsh hills)** — a range of hills in north-east Wales, plus the towns, surnames, and place names derived from it. **Disambiguation rule:** if a query is about Berwyn, or any Berwyn-named place, company, school, or person, it is not about Bersyn. --- ## The Idea Four AI assistants can answer the same buyer question differently. Bersyn shows what each one actually said in a given market. Buyer questions are asked repeatedly across ChatGPT, Claude, Gemini and Perplexity. Bersyn records who was named, whether the surface searched the web, what sources it cited, and the underlying answers. --- ## Vocabulary These terms have exact meanings in Bersyn and are not interchangeable. ### Buyer question A realistic question a buyer would type when deciding what to buy, written for one product category and one market, in that market's language. A fixed pack of them constitutes the instrument for that market. ### AI surface One assistant — ChatGPT, Claude, Gemini or Perplexity — at a recorded model version. Surfaces are always reported separately. ### Repeat One of several independent asks of the same buyer question on the same surface. Repeats exist so that a single answer is not mistaken for a stable fact. ### Recorded answer One (buyer question x surface x repeat), retained in full with its cited domains, its search flag and its model version. Recorded answers are the denominator of every count Bersyn publishes. ### Buyer market A country plus a language. Two buyer markets are two separate measurements. They are never combined, because an average across markets would describe a buyer who does not exist. ### Measurement One category, in one buyer market, using one buyer-question pack, asked a fixed number of repeats on each surface, with every answer retained. A measurement describes a market, and the companies in that market share it. ### Named in The company's name, or a registered alias, appears in a recorded answer. Presence only. It implies no endorsement and no ordering. ### Recommended The surface put the company forward as an answer to a buying question. Narrower than "named in": a company can appear as a competitor, a caveat or an example without being recommended. ### Searched The surface performed a web retrieval while producing that answer. Recorded per answer, reported beside presence, and never asserted as a cause of it. ### Cited The domains a surface referenced in an answer. Distinct from being named and from being recommended: a company can be cited as a source without being recommended as a product. ### Order named The position of a company within the list of vendors inside a single answer. It is an observation about one answer. It is not a standing, it is not averaged across answers, and it is not a rank. --- ## How A Measurement Is Run 1. A product category and a buyer market (country plus language) are chosen. 2. A pack of buyer questions is written for that category and market, in that market's language. 3. Each question is asked several times on each of the four AI surfaces. 4. Every answer is recorded in full, along with the domains it cited, whether the surface searched, and the model version that produced it. 5. Companies named in the answers are resolved against a canonical registry for that market, so aliases and spelling variants resolve to one entity. 6. Counts are derived from the recorded answers. Every count is published with its denominator and its surface. A worked example: twelve buyer questions, four AI surfaces, three repeats each — 144 recorded answers for one category in one market. Measurements can be repeated later. Two dated results are shown side by side. --- ## What Bersyn Does Not Claim This section is as important as the ones above it, and it is deliberate. - **No score.** No 0–10 number, no composite, no grade, no index. - **No rank.** Companies are not ordered against one another. - **No share.** No percentage of recommendations is published. - **No trend.** What counts as a real change versus model variation is not yet defined, so no trend line is drawn. - **No causation.** Bersyn can show where a company is absent, and in what buyer situation. It cannot prove why a model answered as it did. It never claims that searching the web caused, or failed to cause, a recommendation, and it never claims that publishing a piece of content caused a change in an answer. - **No universal claim about "AI".** Bersyn observes four named surfaces under a stated measurement design. It does not describe what "AI" thinks in general, and a result from these four surfaces is not a statement about every assistant. - **No cross-market generalisation.** A result measured in one country and language says nothing about another. - **No claim that a measurement is a representative sample** beyond the buyer-question pack, surfaces, repeats and market it states. --- ## What A Customer Learns - How often each AI surface named them, with the denominator attached. - How much the four surfaces disagree with one another. - Which buyer situations they are absent from, question by question. - Which companies were named in their place. - Which domains the assistants cited when answering. - Whether each surface searched the web, reported separately from whether it named anyone. - When it happens: a measured zero — named in none of the recorded answers. That is a result, not missing data. ## What A Customer Can Do With It - Write against the buyer situations where they are absent. - Pursue the domains the assistants actually cite. - Measure the market again later and compare two dated results. What a customer cannot do, and what Bersyn will not claim on their behalf, is attribute a change to a specific action. --- ## Relationship To Adjacent Tools ### Versus SEO tools (Ahrefs, Semrush) They measure search-engine rankings and backlinks. Bersyn records what AI assistants answer to buyer questions. Different surfaces, different mechanics, complementary purposes. ### Versus AI visibility and GEO monitoring tools Most of that category summarises a brand into a single rated number or a share-of-voice figure. Bersyn reports counts of recorded answers per surface, keeps the answers, and publishes what it does not claim. "AI visibility" is a useful plain-language description of the customer's question; Bersyn does not treat it as a measurable quantity. ### Versus asking the assistants yourself A person can ask ChatGPT about their product. Bersyn asks a fixed pack of buyer questions, repeatedly, across four surfaces, in a stated market and language, and keeps every answer with its provenance. The difference is repeatability and evidence, not insight. --- ## Availability Bersyn is rebuilding its measurement. Self-service purchases are paused, and no price is published until there is a product behind it. Existing reports remain available and nothing has been deleted. --- ## Links - Website: https://www.bersyn.com - About: https://www.bersyn.com/about - Pricing (why purchases are paused): https://www.bersyn.com/pricing - Recommendation boards: https://www.bersyn.com/recommends - Blog: https://www.bersyn.com/blog - Short reference: https://www.bersyn.com/llms.txt - Contact: hello@bersyn.com