Aliyev’s Forked-Tongue Policy: Peace Meets Anti-Armenian Propaganda
The normalization process initiated by the United States through the Washington Accords in August 2025 was a watershed moment for the region, as it suspended any aspirations that Azerbaijan may have had of using force against Armenia, while giving impetus to the stalled peace process. In its expansive negotiations with the Trump Administration, as well as in public declarations, Armenia has maintained its highly constructive approach of seeking to quickly and immediately sign a final peace treaty. Azerbaijan, on the other hand, has preferred to use the normalization process to enhance bilateral relations with the U.S., while not fully committing to signing a peace treaty, which it has strategically conditioned on sets of artificial preconditions. The prevailing narrative coming out of Baku is a unique blend of paradoxical sophistry: speak to the international community, especially the United States, in the language of peace, while domestically, speak to the Azerbaijani people in the language of anti-Armenian propaganda and hate speech.
On the global stage, President Aliyev portrays himself as a symbol of reconciliation, cooperation, and regional healing: the strongman who can destroy his enemy, but due to his nobility of character and commitment to peace, has become the purveyor of stability, amity and goodwill. On the domestic stage, Aliyev’s rhetoric and policy positions are so diametrically different that they border the schizophrenic: language of hostility, demonization, hatred, faux historical grievances, and dehumanization of the Armenian. What may appear as an exercise in incoherence is in fact a well-designed strategy predicated on achieving numerous sets of objectives: 1) re-establishing and strengthening relations with the U.S. without committing to final peace; 2) utilizing the achievements of the Washington Accords to secure American investments in order to diversity its stagnant, oil-dependent economy; 3) achieving regional interconnectivity and economic growth to become a commercial hub, yet having the capacity to manage the scope and breadth of Armenia’s engagement; and 4) utilizing the language and posture of peace to achieve all three objectives, while reserving for itself the flexibility of not committing to comprehensive peace.
The mutually aligned interests of Armenia and the U.S., on the other hand, primarily revolve around trying to make certain that Azerbaijan abides by the following set of principles: 1) achieving substantive peace; 2) refraining from use of force; 3) abandoning the use of coercive diplomacy; 4) respecting the sovereignty and territorial integrity of all actors involved; 5) and abstaining from behavior and policies that are antithetical to peace, including, but not limited to, state-sponsored hate speech, historical and religious erasure, and hostile rhetoric. The most important endeavor in qualifying these principles is to discern words from action, and to concretely substantiate the behavior of official Baku based on data-driven, empirically grounded findings. Speaking the language of peace is not enough if concrete actions are not taken to achieve peace. While Armenia has vigorously done both, Azerbaijan’s approach has been a collage of strategic lip-service and selective acts of diplomatic performativity.
To truly gauge official Baku’s commitment to the peace process, this author, along with his team researchers, has undertaken an expansive study[1] to observe the relationship between the domestic policies, rhetoric and actions of Azerbaijan and how this correlates its commitment to the peace process.
Methodology and Data
This research project collected data from the entirety of Azerbaijani media ecosystem, with over 100,000 unique entries for this report alone. Over 80% of coverage originates from state-aligned domains (.az, .tr), with the official narrative dimension accounting for 89.6% of all narrative coverage. In simple terms, almost the entirety of the content accrued, processed and quantified from the Azerbaijani media ecosystem is either state-sponsored or state-sanctioned. Thus, the data used to observe the patterns and systemic causal links are predicated on content directly produced or sanctioned by official Baku.
Data for this project is collected from the Global Database of Events, Language, and Tone (GDELT), along with various sources across web, broadcast and print media that are independently collated by Nyrus AI’s Sentinel Geointelligence Analysis. We machine-categorized articles, tracking events, event mentions, semantic themes in the data, locations, persons and organizations present in the data, direct quotations and metadata. The primary analysis window spans from August 8, 2025 (signing of the Washington Accords) to mid-January, with a baseline comparison period extending back to April 15, 2025.
We applied a three-table join output to capture the following: 1) event metadata, which includes CAMEO event codes (which is a coding framework primarily used for political news and violence), actor information, and Goldstein conflict-cooperation scores (scale used to classify event data from news outlets); 2) article-level data (source outlet, publication date, document URL); and 3) extracted features (thematic tags, sentiment scores, named entities, organizations, locations, and quoted text). The dataset was filtered to capture directional hostility by restricting to events where Azerbaijan is the initiating actor and Armenia is the target. This dyadic filter specifically isolates Azerbaijani actions directed toward Armenia. Events meeting category-specific criteria received categorical labels, while continuous tone scores, developed through machine learning model (GDELT’s V2Tone field), were utilized for quantitative analysis. We collected over 100,000 unique entries fitting our filtering criteria, representing a comprehensive view of the entire Azerbaijani media ecosystem.
Classification of the Categories in the Data Set
We selected four main categories to organize and operationalize the large body of variables. These categories are: 1) territorial claims; 2) hate speech; 3) denial of sovereignty; and 4) hostile and anti-peace rhetoric. The “territorial claims” category is defined by the inclusion of variables specific to Azerbaijani claims over Armenian territories, irredentist claims based on historical revisionism, justification and support for erasure of Armenian culture, language, history, or religion, and the entire scope of activities by the Western Azerbaijani Community (WAC), a state-sponsored organization. The “hate speech” category is defined by the inclusion of variables specific to dehumanizing language, racial or ethnic hatred, celebration of death or killing of Armenians, and characteristics of Armenophobic tropes and stereotypes. The “denial of sovereignty” category is defined by the inclusion of variables specific to disregard for the existence of the Republic of Armenia, rejection of Armenia’s sovereignty or territorial integrity, promotion of language denying Armenian statehood or the legitimacy of its statehood, and content promoting, justifying, or celebrating Azerbaijan’s continued occupation of Armenian sovereign territories. The “hostile and anti-peace rhetoric” category is defined by the inclusion of variables specific to public statements, demands, or expressions that are disruptive to the peace process, disparaging of Armenia’s commitment to peace, critical or uncommitted to the terms of the peace treaty, and content that contradict the principles, values, and agreed language of the Washington Accords, including, but not limited to, use of the “Zangezur Corridor” language.
Findings and Results
The harvested data on the “territorial claim” category demonstrated continued irredentist claims, historical revisionism, and supplanting of Armenian cultural, historical and religious facts with artificial Azerbaijani constructs. The pattern of content demonstrates systemic and state-sanctioned narrative reproduction that is persistent across the time-series model. The active role of the Western Azerbaijani Community, Baku’s primary instrument of advancing irredentist claims against the Republic of Armenia, remains acute and traceable. Trends prior to and after the Washington Accords do not display a change, decrease, or shifts in Baku’s continued territorial claims. To the contrary, the data demonstrates consistent or increased hate speech and irredentist claims.
The graph below displays the proportional share of three key narrative categories— Irredentism, Hate Speech, and Legal Consolidation (use of legal language or claims of legal action)—as a percentage of total flagged events prior to and after the Washington Accords. The relatively stable proportional distribution across periods indicates no significant shifts following the Washington Peace Summit, clearly demonstrating Azerbaijan’s consistent use of territorial claims even after signing the Accords. At the same time, there is an observable peak in hate speech and irredentist claims towards the end of 2025 and beginning of 2026.
The graph below displays the distribution of articles produced by Azerbaijan, its allies, or paid proxies with content emphasizing historical revisionism or justifying Azerbaijani appropriation or claims over Armenian historical sites. The data observes the trend analysis within a four month snapshot. The distribution demonstrates a high-volume production, after the Washington Accords, on a weekly basis, where we observe an average of 13 articles produced per week specifically on this subject.
To enhance the analytical scope of our inquiry, we also observed the production of content that expanded the narratives and nuanced arguments of Baku. Thus, the above graph similarly displays distribution of articles originating from Azerbaijan or its allies, with content emphasizing demographic resettlement, one of the more prominent ideas promoted by the Western Azerbaijan Community. Resettlement primarily refers to Azerbaijan’s rejection of Armenia’s legitimacy as a sovereign state and the right of Azerbaijanis to return to their “ancestral homeland” that they deem is illegally or unjustly occupied by Armenia. As the volume of articles demonstrates, official Baku, on a weekly basis, continues to support the Western Azerbaijan narrative even months after signing the Washington Accords and claiming commitment to the peace process.
The above two graphs demonstrate our findings on the activities of the Western Azerbaijani community at the cumulative level within an 8-month observational trajectory. As displayed in the first graph, the activities of the WAC have increased five-fold since the August 8 signings, clearly showing robust levels of activity that are counter to the principles and spirit of the peace summit. The second graph demonstrates increased distribution of WAC mentions in the Azerbaijani media ecosystem, with observable increases months after the initiation of the U.S.-led peace process.
The collected data on the “hate speech” category demonstrate extensive dehumanizing language, racial and ethnic hatred, and characteristics of Armenophobic tropes and stereotypes within the state-sponsored media ecosystem of Azerbaijan. Hate speech toward Armenians is systemic and persistent in the entire Azerbaijani ecosystem, and the persistence is observable in a continuous trend pattern that has not shifted since the Washington Accords. Thus, empirically speaking, content sanctioned, funded, or supported by the Aliyev regime continues to proliferate expansive patterns of hate speech towards Armenia and Armenians.
The graph below shows the weekly volume of hate speech events undertaken directly by the Azerbaijani state, with the data broken down by the specific institutional role of the Azerbaijani actor. The vertical dashed line marks the August 8, 2025 signing of the Washington Accords. Two institutions stand out: the Government (GOV), pertaining to the executive office, and the military (MIL). Both actors dominate hate speech production throughout the post-agreement period, with notable spikes in late September and late October 2025, along with a sizable rebound in January 2026, demonstrating that hateful content continues to emanate from official state institutions rather than solely from media or civilian sources. Majority of the hate speech against Armenia and Armenians is reproduced and proliferated by the Government, while the military maintains a supportive pattern of continuity in perpetuating the hate speech content.
The collective data on the “denial of sovereignty” category displays a persistent generating process of content that rejects Armenia’s sovereignty or territorial integrity, promotes language denying Armenian statehood, and content justifying or celebrating Azerbaijan’s continued occupation of Armenian sovereign territories. The graph below traces articles produced by Azerbaijan, its allied states, or paid proxies that question or challenge Armenia’s sovereignty or the legitimacy of the Republic of Armenia as a state. For this data, we selected to observe a two-month period to better gauge saturation of content. As can be seen, the distribution of articles clearly demonstrates how official Baku continues to domestically advance narratives that are in direct contradiction to the principles of the peace process. Specifically, noting the importance of sovereignty and territorial integrity to both the signed documents in the Washington Accords as well as the initialed draft of the peace agreement, Azerbaijan’s continued production of high volumes of articles that deny or challenge Armenia’s sovereignty or territorial integrity demonstrate a concerted state policy that clearly questions Aliyev’s commitment to the U.S.-led normalization process.
The collected data on the “hostile and anti-peace rhetoric” category provides evidence of contiguous patterns of statements and expressions from the Azerbaijani government or affiliated outlets that are unequivocally contrary to the peace process. These include, but are not limited to, disparaging commentary on Armenia’s commitment to peace, language that contradicts the principles of the Washington Accords, and demonstrate continuous use of terms such as “Zangezur Corridor” or corridorization language. The chart below tracks the weekly percentage of events, after the August 8 Washington Peace Summit, that fall below the three negative tone thresholds: moderately hostile (Tone < −4, yellow), hostile (Tone < −6, orange), and extremely hostile (Tone < −8, red). Pre-August baseline rates of approximately 25–30% across all thresholds spiked dramatically immediately after the Washington Accords. We observed 60–90% of weekly events falling below the moderate hostility threshold and within the hostile to extremely hostile thresholds. This demonstrates that not only did event volume increase post-August, but the proportion of hostile content also increased substantially. Simply put, the signing of the Washington Accords actually correlated with more hostile coverage, not less. This is not merely a volume effect, as the share of hostile content increased: spikes in September 2025 and November 2025 show 80–90% of events falling below even the −6 threshold (hostile). The red line (extreme hostility < −8) shows high volatility, indicating episodic bursts of severely hostile content. In this context, as soon as Aliyev signed the Washington Accords, his government proceeded to ramp up hostile anti-Armenian rhetoric, as opposed to the general expectation that post-peace summit, this kind of state-sanctioned behavior would decrease. What this data more acutely demonstrates is that what Aliyev says in Washington, or to Washington, is very different from what he says in Baku or from Baku.
The heatmap below tracks hostility level distribution in coverage over time. Each row represents a hostility classification tier, with cell shading indicating frequency counts per week—pale yellow for minimal activity, deepening through orange to dark red for high-frequency periods. The “mild” or “neutral” category dominates early coverage (August 2025), but notable escalations in “hostile” and “extreme hostile” coverage appear in late August, mid-October, and late November 2025. The most striking pattern emerges in early January 2026, where “hostile” coverage spikes dramatically (dark red), suggesting a marked deterioration in tone. Throughout the observation period, “negative” and “very hostile” categories remain consistently low, indicating that hostile coverage tends to cluster at moderate-to-extreme levels rather than distributing evenly across the spectrum. This suggests that Aliyev’s messaging may be specific to fluctuations within the domestic political theater, and that escalation of hostile language toward Armenia serves an important diversionary purpose.
Conclusion
In the six months following the signing of the Washington Accords, the total collected data on the four categories of our model display trends and empirical outputs that demonstrate Aliyev’s questionable commitment, based on domestic outputs, to the U.S.-led peace process. The large body of data substantively illustrate the disconnect between the scope, breadth and modality of content proliferated in the Azerbaijani media ecosystem by official Baku, and the diplomatic posturing of the Aliyev regime on the global stage. The magnitude and systemic persistence of the anti-Armenian propaganda, defined by hostile, irredentist and dehumanizing content, empirically attests to Aliyev’s forked-tongue approach to the peace process. In essence, Aliyev is either lying to his domestic audience or he is lying to his international audience. That we can infer his behavior is more consistent with the latter than the former is quite intuitive when observing the long track record we have of Aliyev’s demeanor. At the same time, the fact that his behavior has been constrained, penchant for using force neutralized, and overt kinetic and coercive diplomacy restrained, all suggest that Aliyev remains acutely cognizant of the new geopolitical reality in the region. He will give the U.S. what the U.S. expects, and in the meantime, temper his behavior toward Armenia and utilize performative diplomacy to project a constructive approach to the normalization process.
But what Aliyev cannot do is lie to those who know: the data speaks for itself, and until trends in such data show shifts in the opposite direction, it is difficult to view a regime that has developed the information ecosystem that Aliyev has as one that is willing to agree to comprehensive and sustainable peace.
