Someone trying to make sense of the war in Gaza, and not sure what to think, tends to do what everyone does: search “Israel Palestine explained” and watch whatever comes up first. People reach for those videos precisely because an explainer promises a neutral starting point, the history laid out and the argument left to the viewer, and the most-watched of them carry tens of millions of views each. So we read the transcripts of the 36 most-watched Israel-Gaza explainers on YouTube, a combined 127.5 million views, and measured how much they explain and how much they conclude. The answer: 19 of the 36 state a verdict on a contested question in their own voice, the apartheid label, who started the war, whether the land was taken, what Hamas is, the kind of conclusion an undecided viewer came to weigh rather than to be handed.
The verdicts run in both directions: of 50 that survived strict verification, 30 align with the Palestinian narrative and 19 with the Israeli one. On the surface grammar of violence the canon is close to balanced, naming Israel and Palestinian or Hamas actors about equally, while the words it chooses render Israel as a state and Palestinians as a displaced population well before any explicit argument is made. This report measures the language of the explainers, what words they use and what they assert, not the events themselves and not whether any framing is correct. The Hanover Institute does not adjudicate the contested claims in these videos; it measures how often, and in whose words, the videos do.
Key Findings
- 19 of 36 explainers (53%) assert at least one contested claim as established fact in their own narration; 50 such assertions survived strict verification out of 72 flagged, and 44 were captured as a validated verbatim quote (Hanover Institute study, 2026, first-party).
- The verdicts are bidirectional: 30 align with the Palestinian narrative and 19 with the Israeli one (1 neither). The most-adjudicated questions are who the land rightfully belongs to (10), what Hamas is (8), and whether the project is colonialism (7).
- The line between the two groups is grammatical. Wire and broadcast outlets attribute the same claims (“the UN says,” “human rights groups say,” “considered a terrorist organization by”); the verdicts cluster in advocacy channels on both sides and in several mainstream ones, including Vox, The Guardian, and CNN.
- Among 555 sentences depicting a violent act, Israel is named as the actor 206 times and a Palestinian or Hamas actor 198 times, with 119 (about 21%) naming no actor at all, a near-parity that diverges from the Institute’s headline study of the same conflict, where Israel was the named agent 79.8% of the time.
- The explainers’ own narration is essentially free of antisemitic tropes (checked across all 36), so the contested-claim verdicts are political conclusions about the conflict, not coded prejudice.
In the videos’ own words
The heart of the finding is verbatim. Each line below is a conclusion an explainer states in its own narration, captured word for word from the transcript and verified as an exact quote, not a paraphrase. They are grouped by the narrative each verdict aligns with, a descriptive label and not an Institute judgment of the claim.
Verdicts that align with the Palestinian narrative
| Channel | On | In the video’s own words |
|---|---|---|
| Al Jazeera English | the land | “the British promise that led to the destruction of Palestine and the creation of the state of Israel” |
| Vox | 1948 | “Zionist forces and militias began to forcibly expel hundreds of thousands of Palestinians from their land” |
| Things I Care About | colonialism | “Palestinians were facing a colonial project that had worldwide support and the outspoken goal of driving them from their land” |
| The Guardian | casualties | “Israel’s killed more than 64,000 Palestinians in Gaza, with the real toll likely much higher” |
| Dr. Roy Casagranda | the land | “they start telling the lie of a land without a people for a people without a land and it’s just not true” |
| Dhruv Rathee | colonialism | “their country is being colonized by the Israelis” |
Verdicts that align with the Israeli narrative
| Channel | On | In the video’s own words |
|---|---|---|
| Ben Shapiro | blame | “One side has accepted every single peace deal provided to it. And one side has said we will not accept any peace deal.” |
| Deeper Look Media | 1948 | “the Arabs living in Palestine were told to leave by the Arab forces against Israel” |
| TBN Israel | blame | “it was meticulously planned by the PLO, by the Palestinian Authority, to sabotage the peace with Israel” |
| explainitychannel | Hamas | “the terror organization Hamas carried out an attack on Israel” |
| History on Maps | the land | “the Roman Empire who gave the name Palestine to Judah intending to break the Jewish connection with the land of Israel” |
| Vox | Hamas | “Hamas, a violent extremist group dedicated to Israel’s destruction” |
A neutral explainer can narrate every one of these events without reaching the verdict. The contrast is visible in the same word: a wire outlet writes that Hamas is “considered a terrorist organization by the UK and US,” while an explainer that adjudicates writes that Hamas “is an actual terrorist group,” in its own voice. Vox appears in both tables, a reminder that this is not a property of fringe channels alone.
How many explainers state a verdict on a contested question?
Of the 36 explainers, 19 assert at least one contested claim as fact in their own narration rather than attributing it. The contested questions are the ones reasonable parties dispute and a research organization would not settle: whether Israel’s conduct is apartheid, ethnic cleansing, or genocide; whether Zionism is a colonial project; who started the conflict; whose land it rightfully is; whether the occupation is legitimate; and what Hamas is.
The verdicts cluster in advocacy channels on both sides, HasanAbi and Dr. Roy Casagranda toward the Palestinian narrative, Ben Shapiro and TBN Israel toward the Israeli one, but several mainstream outlets adjudicate too. By contrast the BBC, Channel 4 News, CrashCourse, HISTORY, ABC News, and RealLifeLore consistently attribute the same claims to a named party, which is what kept them out of the count.
Which direction do the verdicts run?
The 50 verified verdicts split 30 toward the Palestinian narrative and 19 toward the Israeli one, with 1 favoring neither. This is a measure of direction, not of correctness: the Institute records which narrative an asserted verdict aligns with and takes no position on the claim itself.
That both directions appear in volume is the point. The finding is not that the explainer canon leans one way; it is that more than half of these videos resolve questions they could have left open, so a viewer who treats an explainer as a neutral chronology is often absorbing a conclusion as well.
Who is named as the actor when the videos describe violence?
Across 555 sentences that depict a violent act, the explainers name Israel as the actor 206 times and a Palestinian or Hamas actor 198 times, name both in 32, and name no actor at all in 119, about 21%. The near-parity is itself the finding, and it diverges sharply from the same conflict’s news headlines.
In the Institute’s headline study, Israel was the named agent 79.8% of the time, because those headlines covered a sustained phase of Israeli operations in Gaza. Explainers narrate the full century, the 1948 and 1967 wars, Palestinian and Arab attacks, suicide bombings, the October 7 assault, and Israeli operations, so the agency spreads across both sides. The shared feature is the agentless share: about a fifth of violent acts in the explainers, like nearly a quarter of the headlines, are described with the actor removed from the sentence.
What words describe each side?
The descriptive vocabulary tracks the kind of actor each side is in the narration. Israel is rendered as a state with a military: forces, army, soldiers, and troops attach to it 39 times, against 18 for the Palestinian side. Palestinians are rendered through displacement and civilian status, with refugee attaching to them 13 times against 1 for Israel, while Hamas and Palestinian fighters draw militant and fighter (12 to 7). The word terrorist appears in both contexts at similar raw frequency, but reading the sentences shows it attaches overwhelmingly to Hamas and Palestinian attackers, including in sentences that name Israel as the party attacked.
This is a descriptive asymmetry, not a verdict on it. A state and a stateless population are genuinely different kinds of actor, and a faithful account would name a state’s army and a displaced population’s refugees. The measurement records the pattern; it does not claim the pattern is unwarranted.
Methodology and limitations
This report presents an original Hanover Institute study and ships its full method.
Provenance. A fixed set of neutral, explainer-intent queries (for example “israel palestine conflict explained” and “gaza explained”) was run against YouTube from a United States location. Results were de-duplicated and ranked by view count; the top 40 videos were kept, and the 36 with English captions were analyzed. Reach was the only selection criterion, so the corpus is reproducible.
Sample. 36 explainer transcripts, a combined 127.5 million views, the trusted explanatory layer a viewer is most likely to reach. The units of analysis are the transcript (for adjudication and vocabulary) and the violence-depicting sentence (for agency).
Enrichment. Three layers. Vocabulary and side mentions were counted deterministically with lemma-stable part-of-speech extraction. Agency and contested-claim adjudications were coded by a frontier-model pass reading each full transcript. Every flagged adjudication then went through a strict, default-reject verification that removed attributed, hedged, both-sided, and uncontested-factual statements (72 fell to 50), and the surviving verdict was matched to a verbatim quote validated as an exact substring of the transcript (44 of 50 located verbatim) and read by hand. The narration of all 36 was separately checked for antisemitic tropes and found essentially clear.
Limitations. Adjudication and agency are model-assisted judgments of language, probabilistic by nature, which is why every positive was independently verified, matched to a verbatim quote, and reported with its direction rather than a ruling. “In its own voice” counts narration and, in interview-driven formats, an on-screen speaker the video presents without counterpoint; the count is a measured estimate, not an exact census. Agency counts the surface grammar of who is named as the actor, which real-world base rates confound, so it is not a measure of bias or accuracy. The vocabulary counts are co-occurrence within sentences mentioning a side; attribution was confirmed by reading. The study covers 36 English-caption videos on one platform and a single subject. No private individual is named; the channels quoted are public publishers.
Conclusion
So does the explainer hand the undecided viewer the neutral chronology it promises, with the disputes left open? On the evidence of the transcripts, often it does not: of the 36 most-watched results, a combined 127.5 million views, 19 resolve at least one contested question in their own narration rather than attributing it, so more than half of the canon a viewer reaches for as a starting point arrives with a conclusion already reached.
That conclusion runs in both directions, and the portrait the numbers draw is sharper than “bias.” Of the 50 verdicts that survived strict verification, 30 align with the Palestinian narrative and 19 with the Israeli one, so the canon does not lean a single way; it resolves, on both sides, the very questions an undecided viewer came to weigh, who the land rightfully belongs to (10 verdicts), what Hamas is (8), whether the project is colonialism (7). On the surface grammar of violence it is close to balanced, naming Israel as the actor 206 times across 555 violent acts and a Palestinian or Hamas actor 198, far from the 79.8% Israeli-agent share of the same conflict’s headlines. Yet before any explicit argument, the words have already sorted the two sides: state-military nouns attach to Israel 39 times against 18 for the Palestinian side, while refugee attaches to Palestinians 13 times against 1 for Israel. The narration of all 36 is essentially free of antisemitic tropes, so what these videos carry is a political conclusion about the conflict, not coded prejudice.
A settled conclusion does not stay inside the video that states it. The claim that the land was taken, that the project is colonial, that one side started it, circulates well past any single explainer, in arguments on campus, in feeds, in the asides traded as common knowledge, and somewhere along that path runs the contested line between criticism of Israel and antisemitism. The transcripts settle one thing and reach no further. They fix how often the canon concludes (19 of 36 videos), in which direction (30 verdicts Palestinian, 19 Israeli), and in whose vocabulary (state-military nouns for Israel 39 times, refugee for Palestinians 13). What they cannot register is what happens after the video ends, whether a settled conclusion delivered as neutral chronology becomes one of the currents feeding what crosses the contested line. That is the question the counts open and stop short of, left standing for a society that watches these explainers to weigh on its own.
Frequently Asked Questions
Does this report say the explainers are biased?
No. It measures two separable things descriptively: whether a video asserts a contested claim in its own voice rather than attributing it, and which narrative an asserted verdict aligns with. It takes no position on whether any verdict is correct, and it notes that real-world base rates confound the agency counts, so they are not a bias measure.
Which questions do the explainers most often resolve?
Whose land it rightfully is (10 of 50 verified verdicts), what Hamas is (8), and whether the project is colonialism (7), followed by ethnic cleansing and who started the conflict (5 each). These are exactly the questions on which the historical and legal record is contested.
Why is the agency balanced here but lopsided in the headline study?
Because the corpora cover different spans. The headline study sampled a sustained phase of Israeli operations in Gaza, where Israel was the named agent 79.8% of the time. Explainers narrate the full century, including Arab and Palestinian attacks and the 1948 and 1967 wars, so the named agency splits roughly evenly.
Do the explainers themselves contain antisemitism?
Essentially no. The narration of all 36 was checked for antisemitic tropes and found clear; the contested-claim verdicts are political conclusions about the conflict, not coded prejudice. The boundary between criticism of Israel and antisemitism is measured in a separate report.
How were attributed statements kept out of the count?
A statement was counted only when the narration asserted it as the video’s own conclusion and could be matched to a verbatim quote. Statements attributed to a party (“the UN says,” “human rights groups say,” “considered a terrorist organization by”), hedged, or presented with the other side were excluded; 22 of the 72 first flagged were removed on that basis.
Sources
- Hanover Institute, 2026. Study of the language of the 36 most-watched Israel-Gaza explainer videos (127.5 million views): contested-claim adjudication with verbatim capture, agency, and per-side vocabulary. First-party study.
- YouTube, 2026. Public video search results and English captions for explainer-intent queries, United States locale, ranked by view count. Public platform data.