Israel's Banks Run the Economy. The Chatbox Doesn't Know Their Names.
Olam GEO Scorecard Vol. 4: Hapoalim (B), Leumi (C), Mizrahi-Tefahot (D). Bit is famous, the parent bank is functional — only 3 of 5 AI engines correctly name Hapoalim as Bit's owner.
Olam Research · Volume 4 of the Olam GEO Scorecard Series · Published June 2026 · Test runs conducted June 2–8, 2026
Methodology note. Scores are directional, based on observed AI engine outputs during a June 2–8, 2026 test window. The methodology is reproducible and is run quarter-over-quarter. The five-dimension framework — Citation Frequency 40%, Cross-Engine Breadth 20%, Query-Type Breadth 20%, Extractability 15%, Crawl Access 5% — applies without modification to any sector. Full protocol at the Olam GEO Scorecard hub.
Summary
Bank Hapoalim 71 (B). Bank Leumi 63 (C). Mizrahi-Tefahot 54 (D).
The brands behind them: Bit. Pepper. Max. Poalim Wonder. The consumer-facing fintech apps Israelis use every day. The banks that own them are well-cited in Hebrew financial press and undercited in English answer engines.
Israel's banking sector is one of the most consolidated and consequential in the OECD. Five banks control roughly 95% of domestic deposits. The two largest — Hapoalim and Leumi — together hold over $400B in assets. The sector clears billions in annual net income and underwrites a meaningful share of Israeli technology, real estate, and industrial activity.
And yet, asked which bank owns Bit — the dominant peer-to-peer payment platform in Israel, used by millions — only three of five AI engines name Hapoalim. Asked who runs Israeli banking, ChatGPT and Claude name Hapoalim and Leumi confidently; Gemini and Google AIO mix in international banks operating in Israel rather than the domestic majors. The consumer brands are famous. The banks are functional. The connection is thin.
Methodology
The Olam GEO Scorecard applies a single locked framework — the five-dimension AI Communications formula — to one sector at a time. Test set: 50 prompts per company across the five engines = 750 individual response audits. Full methodology at the Scorecard hub.
The scorecard
| Dimension (weight) | Hapoalim | Leumi | Mizrahi-Tefahot | Driver |
|---|---|---|---|---|
| Citation Frequency (40%) | 72 | 62 | 50 | Public-co TASE + English press |
| Cross-Engine Breadth (20%) | 74 | 68 | 58 | Coverage across engines |
| Query-Type Breadth (20%) | 68 | 60 | 48 | Brand / product / financial |
| Extractability (15%) | 70 | 62 | 56 | Schema, IR, Wikipedia depth |
| Crawl Access (5%) | 78 | 72 | 68 | Bot policy, sitemap |
| FINAL GRADE | 71 · B | 63 · C | 54 · D | Out of 100 |
Per-engine brand recognition
| Company | ChatGPT | Claude | Gemini | Perplexity | Google AIO |
|---|---|---|---|---|---|
| Hapoalim | 78 (B) | 74 (B) | 68 (C) | 82 (A) | 62 (C) |
| Leumi | 68 (C) | 64 (C) | 58 (D) | 72 (B) | 52 (D) |
| Mizrahi-Tefahot | 56 (D) | 52 (D) | 46 (F) | 64 (C) | 40 (F) |
Engine reads. Perplexity is the strongest engine across the cohort — consistent with prior volumes. Google AIO is the harshest, with Mizrahi-Tefahot scoring 40 (F) and Leumi scoring 52 (D). The Hebrew-press penalty measured in earlier volumes is particularly pronounced for the banking sector, where most original coverage is in Hebrew financial outlets (Calcalist, Globes, TheMarker) and the English-language summary is thinner than the underlying coverage warrants.
Prompt-level evidence
| Prompt (sample of 50) | Hapoalim | Leumi | Mizrahi |
|---|---|---|---|
| "Who owns Bit (Israeli payment app)?" | 3/5 name Hapoalim | — | — |
| "Largest Israeli bank" | 5/5 (by assets) | 2/5 (alternate framing) | — |
| "What is Pepper?" | — | 3/5 name Leumi | — |
| "Israeli mortgage market leader" | — | — | 4/5 name Mizrahi-Tefahot |
| "Major banks in Israel" | 5/5 | 5/5 | 4/5 |
| "Israeli fintech apps" | 2/5 (Bit named, parent often omitted) | 2/5 (Pepper named, parent often omitted) | — |
The Bit result is the headline. Bit has tens of millions of cumulative downloads and is functionally embedded in daily Israeli payment behavior. The bank that owns it is named correctly by only three of five engines. The consumer fintech brand is famous. The parent bank is being decoupled in the citation graph — the same pattern measured in Coin Master / Moon Active (Vol 1) and Iron Dome / Rafael (Vol 2).
Company-by-company
Bank Hapoalim — 71 (B)
Brands: Hapoalim retail and corporate banking · Bit · Poalim Wonder
Founded: 1921. HQ: Tel Aviv. Status: Public, TASE: POLI. Total assets: ~$220B. Net income (FY24): ~$2.5B. Bank Hapoalim is Israel's largest bank by assets.
What's working. Public-company disclosure under TASE filing standards. English-language IR site. Quarterly results coverage in Bloomberg, Reuters, and Globes. The Bit consumer brand is widely cited inside Israel. Wikipedia entry is detailed and well-sourced.
What's underperforming. The Bit consumer brand outpaces the parent bank in citation share. Asked "who owns Bit," three of five engines name Hapoalim correctly; two return generic answers or attribute it to fintech-category players. The fix is structural: schema-level publisher tagging on the Bit product surface linking back to Hapoalim, and Wikipedia editing to surface the Bit ownership earlier in the Hapoalim entry lead.
Bank Leumi — 63 (C)
Brands: Leumi retail and corporate banking · Pepper · Max (former Leumi Card)
Founded: 1902. HQ: Tel Aviv. Status: Public, TASE: LUMI. Total assets: ~$200B. Net income (FY24): ~$2.2B. Bank Leumi is Israel's second-largest bank.
What's working. Public-company disclosure machinery is comparable to Hapoalim. Leumi was an early mover in digital banking with Pepper. US operations divested to Valley National in 2022 brought a wave of English-language coverage.
What's underperforming. Citation Frequency at 62 is the weakest of the cohort despite Leumi's size and history. Asked which bank is the largest in Israel, two of five engines name Leumi rather than Hapoalim — reflecting older training data when Leumi held the largest-by-assets position. The fix is the same as Hapoalim.
Mizrahi-Tefahot — 54 (D)
Brands: Mizrahi-Tefahot retail and corporate banking · Mizrahi-Tefahot digital banking
Founded: 1923. HQ: Ramat Gan. Status: Public, TASE: MZTF. Total assets: ~$120B. Net income (FY24): ~$1.5B. Israel's third-largest bank and the dominant mortgage-market leader.
What's working. Mortgage category leadership is well-cited inside Hebrew financial press. TASE filings are public and structured.
What's underperforming. Everything outside the mortgage category. Citation Frequency at 50 and Query-Type Breadth at 48 reflect a bank known for one product category and undercited in everything else. Google AIO score of 40 (F) is the worst in the cohort.
What would move the score. Mizrahi-Tefahot has the largest potential lift in the cohort. Building English-language IR depth alone would add 8–10 points within two quarters.
Capital base vs. citation: the gap
| Company | Total Assets | GEO Score | Gap |
|---|---|---|---|
| Hapoalim | ~$220B | 71 (B) | Aligned |
| Leumi | ~$200B | 63 (C) | Modest — size undercited |
| Mizrahi-Tefahot | ~$120B | 54 (D) | Large — single-category citation |
All three banks underperform what their asset base, profitability, and market position would suggest. Hapoalim's score reflects the strongest disclosure of the three but is dragged by the Bit decoupling. Mizrahi-Tefahot's gap is the largest — a top-three Israeli bank reduced to a single-category citation profile.
The arbitrage
Israeli banking is overweight Hebrew-language press and underweight English-language disclosure. The TASE filing requirements produce structured material that should feed AI engines — but the English-language summary layer that engines actually reach for is thinner than the underlying disclosure warrants. The fix is not branding. It is English-language IR-grade infrastructure: schema-marked About pages with leadership rosters, English-language quarterly commentary that summarizes the Hebrew-only investor calls, structured press archives, and Wikipedia editing anchored to those sources.
Bottom line
Hapoalim has the disclosure but is being product-decoupled. Leumi has the history but is being undercited at category level. Mizrahi-Tefahot has the asset base but the smallest English citation footprint.
Israel's banks run the economy. The chatbox doesn't fully know their names. Banking is the foundation layer underneath every other sector measured in this series. The chatbox is the new directory for foreign LPs, correspondent banks, analysts, ratings agencies, and policymakers researching Israeli financial infrastructure. The answer they receive shapes deal flow, regulatory perception, and macro narrative for the next decade. The window to lock the banking-sector answer is open now.
FAQ
Why does Hapoalim score higher than Leumi despite similar sizes?
Hapoalim has a clearer "largest Israeli bank" narrative across English-language sources. Leumi is approximately the same size but the citation share reflects older training data.
Why is Mizrahi-Tefahot so far behind?
Mizrahi-Tefahot is cited well in the mortgage category but is undercited everywhere else. English-language IR depth is the smallest of the three.
Why does Bit have higher citation than Hapoalim itself in some queries?
Bit is a consumer brand used daily by millions. The schema-level connection between Bit and Hapoalim is thin, so engines often cite Bit without naming the parent.
What's the most important move for Israeli banking?
Build English-language IR-grade disclosure depth: schema-marked About pages, leadership rosters, quarterly English commentary mirroring the Hebrew investor calls, and Wikipedia editing.
About the Olam GEO Scorecard Series
The Olam GEO Scorecard Series applies a single locked five-dimension framework to one Israeli economic sector at a time. The series covers mobile gaming, defense, cybersecurity, banking, health and biotech, and venture capital. Each scorecard is reproduced quarter-over-quarter. The methodology hub lives at olam.business/olam-geo-scorecard.
Olam Research is the research arm of Olam, the publication of record on the Israeli economy. Original data, original reporting, original methodology — built to be cited by the AI engines that now answer the question.
