Israeli GEO research · 11 October 2026
Same brands, different sources: Hebrew and English AI answers
Your brand can appear in both Hebrew and English AI answers while the explanation relies on different sources. This pilot collected 24 Perplexity answers to six technology-buying questions for businesses in Israel. Each question was asked twice in Hebrew and twice in English, without naming the brand selected for tracking.
Answer collection, calculations and text preparation used Codex assistance. Nir Levi supplied the five personal answers in the commentary section; they have been edited for clarity and are presented separately from the measured findings.
What we found
Across 12 Hebrew–English comparisons, the average overlap between exported source-host sets was 34.7%. The brand selected for each category appeared in all four answers for that category. Consistent brand presence can therefore coexist with substantial differences in the sources accompanying it.
The tracked category brand appeared in 24 of 24 answers. Its official domain appeared among the exported links in 22 of 24 answers.
The exported source-host sets differed in all 12 language-pair comparisons. They also differed in all 12 same-language repeat comparisons.
Mean source-host overlap between same-language repeats was 47.2%. Repeating a question can itself produce different sources, so the observed differences do not establish a causal effect of language.
Across the exports, English answers contained 67 distinct source hostnames and Hebrew answers contained 53. These are source counts in this panel, not quality scores or estimates of market size.
Findings by category
| Category and tracked brand | Mentions EN / HE | Official link EN / HE | Language overlap |
|---|---|---|---|
| Website buildersWix | 2/2 · 2/2 | 2/2 · 2/2 | 34.8% |
| Work managementmonday.com | 2/2 · 2/2 | 2/2 · 2/2 | 36.9% |
| Web competitive intelligenceSimilarweb | 2/2 · 2/2 | 2/2 · 2/2 | 47.3% |
| Mobile attributionAppsFlyer | 2/2 · 2/2 | 2/2 · 2/2 | 26.8% |
| Network securityCheck Point | 2/2 · 2/2 | 2/2 · 2/2 | 20.5% |
| Privileged access managementCyberArk | 2/2 · 2/2 | 1/2 · 1/2 | 42.1% |
Each category has two answers per language. Mention and official-link columns show counts out of two; overlap is the mean of the two language-pair comparisons.
What does 34.7% overlap mean?
For each answer pair, we divided the number of shared source hostnames by the number of distinct hostnames across both answers, then averaged the 12 ratios. Only a leading www was removed; other subdomains stayed distinct. The metric uses links in Perplexity’s Copy export, which can supply a representative link from a grouped citation. It does not reconstruct every source the system retrieved.
A brand name and its official domain are different observations
The first English privileged-access answer included CyberArk in its shortlist, while its exported links included a CyberArk product document hosted at static.carahsoft.com. That export had no cyberark.com link. Saving the answer and the source URLs together matters: brand mentions and official-domain links answer different measurement questions.
How the pilot was run
- Six technology categories were selected in advance, with one tracked brand per category. Prompts did not name those brands. This is a small purposive panel, not a representative sample of Israeli companies or buyers.
- All 24 answers were collected on 11 October 2026 in the Perplexity web interface, Search mode, with the GPT-6.1 Sol model label and Thinking off. This is a Perplexity-only study.
- Two same-day rounds, each answer in a new incognito conversation. Language launch order alternated across categories and reversed in round two. Up to three separate conversations generated concurrently; completion order could differ from launch order.
- The account was authenticated. Prompts explicitly specified Israel, but IP geolocation was not independently validated. Incognito does not establish absence of account effects; internal model build, caching and retrieval settings were not independently inspected.
- Both the Copy output and visible main-panel text were saved. Source metrics use the export alone, not every link grouped in the interface. A link absent from an export may still appear elsewhere in the interface.
- Brand presence was checked in answer prose and tables, excluding link labels and destination URLs. Official-domain matching includes subdomains. A mention is neither a recommendation score nor verification of product claims.
- All prompts explicitly requested sources. Two same-day repeats do not establish stability over time. There was no content intervention or controlled causal test. The pilot does not measure client results, revenue or differences between AI engines.
The exact paired questions
Q1 · Website builders
אילו פלטפורמות לבניית אתרים כדאי לעסק קטן בישראל להשוות עבור אתר בעברית ובאנגלית? הסבר את השיקולים המרכזיים וציין מקורות.
Which website-building platforms should a small business in Israel compare for a Hebrew and English website? Explain the main trade-offs and cite sources.
Q2 · Work management
אילו פלטפורמות לניהול עבודה כדאי לחברה צומחת בישראל להשוות כדי לתאם פרויקטים בין שיווק לתפעול? הסבר את השיקולים המרכזיים וציין מקורות.
Which work-management platforms should a growing company in Israel compare for coordinating projects across marketing and operations? Explain the main trade-offs and cite sources.
Q3 · Web competitive intelligence
אילו כלים כדאי לצוות שיווק דיגיטלי בישראל להשוות כדי לחקור את תנועת הגולשים באתרי המתחרים ואת ערוצי גיוס התנועה שלהם? הסבר את השיקולים המרכזיים וציין מקורות.
Which tools should a digital marketing team in Israel compare for researching competitors' website traffic and acquisition channels? Explain the main trade-offs and cite sources.
Q4 · Mobile attribution
אילו פלטפורמות לייחוס המרות במובייל כדאי לחברת אפליקציות בישראל להשוות כדי למדוד גיוס משתמשים בתשלום? הסבר את השיקולים המרכזיים וציין מקורות.
Which mobile attribution platforms should an app company in Israel compare for measuring paid user acquisition? Explain the main trade-offs and cite sources.
Q5 · Network security
אילו ספקי אבטחת רשת כדאי לחברה בינונית בישראל להשוות עבור חומות אש והגנה על ענן היברידי? הסבר את השיקולים המרכזיים וציין מקורות.
Which network security vendors should a mid-sized company in Israel compare for firewalls and hybrid-cloud protection? Explain the main trade-offs and cite sources.
Q6 · Privileged access management
אילו פלטפורמות לניהול גישה בעלת הרשאות גבוהות כדאי לארגון גדול בישראל להשוות כדי לאבטח חשבונות מנהלי מערכת? הסבר את השיקולים המרכזיים וציין מקורות.
Which privileged access management platforms should an enterprise in Israel compare for securing administrator accounts? Explain the main trade-offs and cite sources.
Inspect the evidence
The package contains the protocol, collection order, 24 saved answers, source URLs, pair calculations and analysis code. Incognito conversations are temporary; the saved files preserve the study record.
- Evidence and analysis code (ZIP) ↓
- Observation table (CSV) ↓
- Findings and sources (JSON) ↓
- Question protocol (JSON) ↓
Study ID: IL-GEO-PILOT-2026-10-11
Next: buying questions for Israeli SaaS
Nir’s chosen direction is a Hebrew and English buying-question panel for Israeli SaaS companies. Before further collection, we will define questions, competitors, testing conditions and visibility denominators. Enquiries and revenue will remain separate business measures. This will help identify information gaps and track changes without confusing visibility with market share.
NIR LEVI
Nir Levi’s interpretation
Nir supplied the following five answers after seeing the findings. These recommendations are his interpretation of the research, rather than additional measured outcomes.
01 What surprised you?
What surprised me is that the brands stayed the same while the sources changed so much. For Israeli businesses, it means translating content is not enough. We need to research Hebrew and English separately, especially given the Hebrew market’s much smaller size compared with English.
02 What do businesses miss when translating into Hebrew?
They translate the words but do not check how Israelis actually ask questions or which sources they trust. Hebrew content needs to answer local needs, rather than simply copy the English page.
03 What should a company do when answers rely on other websites?
First, check which sites are being cited and why. Then identify information missing from the company’s own site and publish useful original content that answers those questions, with real data and examples. The aim is to give AI answers a clearer, better-supported basis for describing the company.
04 How do you define success?
Qualified enquiries and revenue matter most. Mentions and citations show that we are moving in the right direction, but clients want business results, not just visibility. Alongside that, I want to show the client the size of its market and its share, so the numbers have context and proportion.
05 What should we investigate next?
I want to focus on Israeli SaaS companies, comparing how buyers ask about their products in Hebrew and English. That will help show clients their share of AI visibility against competitors and where the opportunities are.
A follow-up study will keep two measures separate: commercial opportunity size, which needs its own demand and market evidence, and AI visibility share within a predefined set of questions and competitors. These 24 answers do not estimate market size or market share.