Organizations: Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates · University of Maryland, College Park, United States
Can AI reason about a war before its trajectory becomes historically obvious? Analyzing this capability is difficult because retrospective geopolitical prediction is heavily confounded by training-data leakage. We address this challenge through a temporally grounded case study of the early stages of the 2026 Middle East conflict, which unfolded after the training cutoff of current frontier models. We construct 11 critical temporal nodes, 42 node-specific verifiable questions, and 5 general exploratory questions, requiring models to reason only from information that would have been publicly available at each moment. This design substantially mitigates training-data leakage concerns, creating a setting well-suited for studying how models analyze an unfolding crisis under the fog of war, and provides, to our knowledge, the first temporally grounded analysis of LLM reasoning in an ongoing geopolitical conflict. Our analysis reveals three main findings. First, current state-of-the-art large language models often display a striking degree of strategic realism, reasoning beyond surface rhetoric toward deeper structural incentives. Second, this capability is uneven across domains: models are more reliable in economically and logistically structured settings than in politically ambiguous multi-actor environments. Finally, model narratives evolve over time, shifting from early expectations of rapid containment toward more systemic accounts of regional entrenchment and attritional de-escalation. Since the conflict remains ongoing at the time of writing, this work can serve as an archival snapshot of model reasoning during an unfolding geopolitical crisis, enabling future studies without the hindsight bias of retrospective analysis.
Figures & tables
Figure 1: Critical temporal nodes during the early stages of the 2026 Middle East conflict (top) and representative model analyses at selected nodes (bottom). Models exhibit recognizable patterns of strategic reasoning under the fog of war, and their analyses evolve as new information becomes available across the timeline.
Node
Date
Event
Theme
Theme Disruption
T0
Feb 27
Operation Epic Fury
I
Initial Outbreak
T1
Feb 28
Israeli-US Strikes
I
Initial Outbreak
T2
Feb 28
Iranian Strikes
I
Initial Outbreak
T3
Mar 1
Two Missiles towards British Bases on Cyprus
II
Threshold Crossings
T4
Mar 1
Oil Refiner and Oil Tanker Was Attacked
III
Economic Shockwaves
T5
Mar 2
Qatar Halts Energy Production
III
Economic Shockwaves
Table 1: The selected 11 critical temporal nodes with their corresponding themes.
#
General Exploratory Questions
1
What are the potential future actions by the United States and Israel?
2
What are the potential future actions by Iran?
3
What are the potential involvement or reactions from other major countries?
4
Will the conflict escalate into a global war?
5
What is the most probable pathway to de-escalation or resolution of the Iran-US conflict, and what is a realistic timeline?
Table 2: General Exploratory Questions.
Figure 2: Distribution of news articles across sources over time. The figure plots the number of news articles published by major media outlets at each temporal node.
Idx
Node-Specific Questions
T0: Feb 27, Operation Epic Fury
1
Is the United States likely to conduct direct military strikes on Iranian territory?
2
Could the United States prioritize economic sanctions or diplomatic pressure over military action?
3
Does the announcement of Operation Epic Fury signal the beginning of a broader military conflict?
T1: Feb 28, Israeli-US Strikes
1
Are the Israeli-U.S. strikes likely to escalate into a broader military campaign?
Table 3: The Node-Specific Questions for Theme I: Initial Outbreak & Military Spillover (T0, T1, T2).
Idx
Node-Specific Questions
T3: Mar 1, Two Missiles towards British Bases on Cyprus
1
Is the United Kingdom likely to become directly involved in the conflict?
2
Is NATO likely to become involved, expanding the conflict into the Mediterranean theater?
3
Could the conflict disrupt commercial shipping or maritime security in the Mediterranean?
T6: Mar 2, Natanz Nuclear Facility Damaged
1
Are the United States and Israel likely to continue targeting Iranian nuclear facilities?
Table 4: The Node-Specific Questions for Theme II: Threshold Crossings & Internationalization (T3, T6, T7, T8).
Idx
Node-Specific Questions
T4: Mar 1, Oil Refiner and Oil Tanker Was Attacked
1
Is Iran likely to continue targeting oil tankers in an attempt to disrupt traffic through the Strait of Hormuz?
2
Could international naval forces establish escort missions to protect commercial shipping?
3
Could American naval forces establish escort missions to protect commercial shipping?
4
Are energy facilities such as refineries, desalination plants, and oil terminals likely to become primary targets?
5
Could these attacks lead to significant volatility in global oil prices?
Table 5: The Node-Specific Questions for Theme III: Economic Shockwaves & Market Escalation (T4, T5).
Idx
Node-Specific Questions
T9: Mar 3, Mojtaba Khamenei Becomes Supreme Leader
1
Is the new leadership more likely to escalate military retaliation rather than pursue negotiations?
2
Are the United States and Israel likely to target the new leadership structure in further strikes?
3
Could the leadership transition trigger domestic unrest or protests in Iran?
T10: Mar 6, Iranian Apology to Neighboring Countries
1
Is Iran likely to reduce or halt attacks on neighboring Gulf states?
Table 6: The Node-Specific Questions for Theme IV: Political Signaling & Regime Dynamics (T9, T10).
We introduce Age of LLM, a turn-based 1v1 benchmark in which two LLMs face off on a 13x7 grid to destroy the enemy base. Three stressors are deliberate: fog of war, full diplomacy (messages, ceasefires, ultimatums; uranium kept secret), and a reliability dimension where every turn must follow a strict JSON schema and an illegal action is silently discarded. The engine is private and each match uses a fresh random map seed and opponent, mitigating the data contamination that affects public benchmarks. Models receive a (near) rule-only prompt with no build-order advice (two tactical seed phrases were present during data collection; see Section 2.7). We benchmark 15 reasoning models across 54 matches and 5,258 actions. Findings: (1) the nuclear rush dominates (78% on the rules-coherent v0.11+ sub-corpus; 85% corpus-wide) with a sole-launcher signature that is largely mechanical under secret-simultaneous launch rules, not a cognitive deterrence failure; (2) military conquest is rare but faster (12.3 vs 18.9 turns); (3) diplomacy is prolific yet almost never consummated; (4) ~58% of illegal actions are fog/state errors, making the illegal-action rate a measure of belief-tracking; (5) -- the least established, and the only one we label exploratory -- a weak link associates reliability with winning. The corpus is small, unbalanced and not side-swapped, so the ranking is a preliminary descriptive view, not a contribution. Beyond ranking, the turn-by-turn traces of actions and messages make the corpus a lens on how LLMs reason under adversarial uncertainty -- their belief-tracking, spontaneous deception, and per-model cognitive "personas" -- which we frame as a future research direction. We release the replay format, an isometric viewer and all replays; engine source on request.
Artificial Intelligence (AI) answer engines now field a growing share of the questions that analysts, scholars, and the public ask about issues of peace and conflict. Large Language Models (LLMs) are known to hallucinate under certain conditions, but do these errors have discernible patterns when they are asked about conflicts, and if so what can that teach us about the changing global conflict information environment? To answer, we first asked a battery of questions about 28 conflicts to five leading answer engines and scored their 5,460 answers against documented evidence. We found that the thinner the retrievable record around a given conflict, the more the engines invent, misattribute, and miscount. Thin records don't just encourage hallucination, but create structural exposure to mis- and disinformation, because they are the easiest records to warp through Generative Engine Optimization (GEO) to bias engine responses. Through an analysis of 1,048 websites that the AI LLMs pulled conflict facts from, we found that GEO source optimization is already happening, and while state-partisan digital capture remains incipient it is rapidly growing. We explain what these findings mean for scholarship with the rise of GEO information warfare, and for policy argue for a return to the deep local monitoring and translation-based research that AI tools cannot replicate, closing with a discussion of future research opportunities and challenges in this fast-moving space.
Jason Miklian
Centre for Global Sustainability, University of Oslo
Large language models (LLMs) are increasingly proposed as agents in strategic decision environments, yet their behavior in structured geopolitical simulations remains under-researched. We evaluate six popular state-of-the-art LLMs alongside results from human results across four real-world crisis simulation scenarios, requiring models to select predefined actions and justify their decisions across multiple rounds. We compare models to humans in action alignment, risk calibration through chosen actions' severity, and argumentative framing grounded in international relations theory. Results show that models approximate human decision patterns in base simulation rounds but diverge over time, displaying distinct behavioural profiles and strategy updates. LLM explanations for chosen actions across all models exhibit a strong normative-cooperative framing centered on stability, coordination, and risk mitigation, with limited adversarial reasoning.