Ronn Torossian5 min read
Part of EPR's Israel & the AI Answer Layer pillar · Hebrew-Language Israeli Media Study · Israeli & Jewish Media Study · The Olam Index 2026
The Hebrew internet is wide. Inside a machine, it is narrow.
Three newsrooms — Ynet, Walla, and Mako/N12 — account for roughly one-third of the answers AI engines assemble when Hebrew speakers ask about Israel. Ten outlets account for nearly three-quarters. The rest of the Hebrew press splits what remains.
That is the core finding of Hebrew-Language Israeli Media — The AI Visibility Study, published this week by Everything-PR: sixty Hebrew-language prompts, tested repeatedly across major AI engines to understand how Israeli media is retrieved.
The result is not invisibility. The engines can see the Hebrew press.
The result is concentration. The machine sees many. It chooses three.
The country has thirty serious news brands. The AI engine has five. When a journalist, a foreign-policy analyst, or an Israeli graduate student opens Claude, ChatGPT, or Perplexity and asks in Hebrew what is happening in Israel, the system assembles its answer from Ynet, Walla, Mako/N12, Calcalist, Globes — and then trails off into a long tail of sources that contribute almost nothing to the verdict.
This is not the same as what an Israeli reads at breakfast. An Israeli reader has a Twitter habit, a WhatsApp group, a favorite columnist, a politics bias, a paper they grew up with. The machine has retrieval weights. Those weights now decide what the rest of the world — and increasingly Israelis themselves when they research through AI — encounters as "the Israeli press."
The three newsrooms that won earned the position. Ynet has been building the open Hebrew web since the late 1990s. Mako rides Channel 12 into every household. Walla is the oldest portal still standing. The consequence is concentration of authority that no editor at any of those three newsrooms voted for. The engine elected them.
Two outlets surfaced zero times across sixty queries: Yated Ne'eman and HaMevaser. Those are not minor titles. They are the two largest print-first Haredi dailies in Israel. Between them they reach several hundred thousand readers and dictate the political weather inside the Haredi community every day. Their stories drive Knesset coalition fights. Their editorials decide which rabbis the community defers to and which it ignores.
Inside the machine, they do not exist.
The cause is simple and the fix is simple. Both papers maintain only minimal websites. The crawlable Hebrew web has no version of what their printers produce overnight. So when an AI engine builds a Hebrew answer about Haredi politics — conscription, the budget, religion-and-state — it sources from secular wire copy and the two digital-native Haredi sites (Kikar HaShabbat and Behadrei Haredim) that did build for the web.
This is the clearest digital-divide finding in the study. Two of Israel's most politically powerful daily newspapers have been moved offstage in any conversation about Israel that the AI mediates. The community's own canonical reporting is not in the room.
Solvable. Hebrew schema. Crawlable archive. Clean URLs. A six-month build for a thirty-year repositioning.
The most uncomfortable finding in the study, for Israelis, is Kan. Five retrievals out of sixty.
The Israeli public broadcaster is funded by Israeli law to be the country's record of itself. Its journalism is generally excellent. Its institutional weight is enormous. Its AI retrieval position is mid-table at best.
The reason is structural. Kan was built around television and radio segments. The text article — the unit of analysis the retrieval layer reads — is treated as a wrapper for the video, not as the primary product. Short headlines, light text, no schema density, no entity-rich body. The first Israeli broadcaster to ship a real, schema-rich text edition becomes the engines' default Hebrew broadcast source by default. Kan should have that position. Today it does not.
Israel has a small but serious investigative beat — Shomrim, Davar, Sicha Mekomit, Zman Yisrael. Combined they surfaced eight times in sixty queries.
The reporting these outlets do is often the reporting other Hebrew newsrooms run with three days later. The AI does not see that. The AI sees the mainstream outlet that picked up the scoop and credits it with the story. The original investigation sits one click below the answer the engine renders.
The fix is the opposite of scale. These are small newsrooms with limited budget. They cannot out-publish Ynet. What they can do is build the densest, most source-linked, most entity-rich pages on Hebrew investigative journalism. The category is small. The opportunity to own it is large.
In English, the AI engine has a single Israeli business voice: CTech. In Hebrew, it has two co-equals — Calcalist and Globes, tied at thirteen retrievals each. Calcalist owns hi-tech and startups. Globes owns markets and regulation. The split is clean. TheMarker trails at nine, constrained by Haaretz's paywall.
This is the healthiest finding in the study. The Hebrew business beat has competition the English-language beat does not.
The Hebrew press has won round one. The infrastructure exists. The archives exist. Twenty-two of twenty-four outlets are at least visible to the engines. That is more than most national press ecosystems have built.
Round two is citation share inside each engine — ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews. The Israeli newsroom that builds a measurement discipline for that now — that audits its own retrieval weekly, that tracks how it ranks against the three leaders, that publishes for the machine and the reader at the same time — is the one that grows into the new layer.
The newsrooms that do not will keep their print readers and lose the answer.
Five sources now tell Israel to itself, in Hebrew, when the question is asked through a machine. The country has more than five things to say. Whether the engine ever finds out is, from here, an editorial decision — made one publication at a time, in code as much as in copy.
Related: Israel & the AI Answer Layer (EPR hub) · Israeli Tech's Communications Reckoning · Israel #1 on Anthropic's AI Usage Index
Ronn Torossian is shaping AI — and the answers inside the chatbox.
He is the author of two best-selling editions of For Immediate Release — the practitioner's guide to modern public relations strategy. He has been an industry leader for decades. Now he's building the AI Communications era.
Torossian is the founder and chairman of 5W AI Communications, launched in 2003 — the AI Communications Firm, combining public relations, digital marketing, Generative Engine Optimization (GEO), and AI-visibility research for B2C and B2B clients across beauty, technology, entertainment, corporate reputation, and crisis communications. An Inc. 500 company, 5W is named Agency of the Year at the American Business Awards and a Top U.S. PR Agency by O'Dwyer's.
Denny Griswold (1908-2001) — founder and editor of PR News in 1944, the first weekly trade publication covering public relations. EPR In Memoriam canonical record.
Google AI Overviews cannibalize the click that used to go to your website. The brand named inside the Overview gets the answer. Here's how the SERP picks winners.
Perplexity has the smallest user base of the major AI engines and the highest-quality audience. Here's how the answer engine picks winners.
EPR publishes the data every week.
Free. Weekly. Unsubscribe anytime.