5W AI Visibility Research · Defense & Aerospace · Two-Wave Benchmark

Across two independent waves of testing, two defense technology companies — Anduril and Palantir — account for a higher combined Citation Share than the five largest legacy defense primes in this dataset.

A two-wave benchmark measuring how often U.S. defense, aerospace, and space companies are surfaced, cited, and recommended across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Results reflect observed retrieval behavior during a defined testing window.

Published · May 13, 2026 Author · Ronn Torossian Method · Two waves, 28,400 prompts total Reading time · 14 min
19.8%
Anduril — #1 defense company by AI Citation Share
35.0%
Anduril + Palantir combined
21.1%
Top 5 legacy primes combined (Lockheed, Northrop, RTX, Boeing, GD)
60.0%
Captured by top 5 companies by AI Citation Share
Figure 01 · Observed source distribution

Defense trade press accounted for a large share of citations in this dataset relative to financial media sources.

Share of total citations surfaced by the five AI systems when responding to prompts about U.S. defense and aerospace companies. January-May 2026, 28,400 prompts across two waves.

Wikipedia 20.7% Defense News 10.8% ← Defense trade press Breaking Defense 8.9% ← Defense trade press The War Zone 7.2% ← Defense trade press Reuters & Bloomberg defense (combined) 6.8% Defense company-owned domains (combined) 6.4% NYT, WSJ, FT defense (combined) 5.9% War on the Rocks 5.4% Government sources (DoD, GAO, CRS) 5.1% Defense One 4.6% Substack (war journalism) 4.4% All other publishers 13.8% 0% 5% 10% 15% 20%
Source · 5W Defense & Aerospace AI Visibility Index 2026 · n = 28,400 prompts · ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews · Two waves: Jan-Feb and Apr-May 2026
Executive Summary

A two-wave observational benchmark of how AI systems describe U.S. defense and aerospace.

This report presents a two-wave benchmark of how 25 U.S. defense, aerospace, and space companies are surfaced, cited, and recommended across five AI systems — ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. The dataset comprises 28,400 prompts conducted in two independent waves (Wave 1: January–February 2026; Wave 2: April–May 2026). Findings reported below are stable across both waves within predefined tolerance thresholds (≤1.5 percentage points on company-level Citation Share; ≤2.0 pp on source-level share).

Citation Share (composite metric) · The proportion of responses in which a company is mentioned, recommended, or cited as a referenced source. Mentions, recommendations, and source citations are tracked separately and weighted equally in the headline metric. This is an internal composite metric, not a standard industry measure, and is best interpreted as a proxy for model surface frequency rather than authority, accuracy, endorsement, or operational performance. Test set · 10 modern defense technology firms, 8 space companies, 7 legacy defense primes. Israeli, Indian, and other allied defense firms were tested in an extended dataset and excluded from the top-25 ranking to maintain a U.S. frame consistent with prior 5W benchmarks.

In this dataset, the highest-cited defense or aerospace company observed was Anduril Industries at 19.8% Citation Share. Palantir Technologies followed at 15.2%, SpaceX at 12.7%, Shield AI at 6.4%, and Lockheed Martin at 5.9%. The top five companies collectively accounted for 60.0% of observed defense and aerospace citations.

The structural observation in the dataset: Anduril and Palantir together accounted for 35.0% of Citation Share — a higher combined share than the top five legacy defense primes in the dataset (Lockheed Martin, Northrop Grumman, RTX/Raytheon, Boeing Defense, General Dynamics), which collectively accounted for 21.1%. The two defense technology firms operate at approximately $3.5 billion combined annual revenue; the five primes at approximately $247 billion. In this dataset, revenue does not appear to predict Citation Share — higher-revenue firms were not consistently associated with higher citation rates, and the relationship trends negative in aggregate.

The publisher distribution observed in this category differs from prior 5W benchmarks. Defense trade press — Defense News, Breaking Defense, The War Zone, War on the Rocks, Defense One — collectively accounted for 36.9% of cited sources. Wikipedia anchored the dataset at 20.7%. Reuters and Bloomberg defense coverage (combined) supplied 6.8%; major general-press defense desks 5.9%; company-owned domains 6.4% in aggregate; government sources (DoD, GAO, CRS) 5.1%.

Note on interpretation. Because this study measures AI-generated outputs across a defined time window, results reflect model behavior during that period and may not generalize across future model versions, retrieval systems, or prompt distributions. Citation Share should not be interpreted as market share, operational performance, contract-award likelihood, strategic advantage, or endorsement. Observed correlations do not imply causation. Interpretive analysis appears in a separately labeled section later in this report.

"In this dataset, two defense technology companies account for a higher combined Citation Share than the five largest legacy primes. In a category where revenue has historically been the measure of weight, the AI systems we tested are surfacing a different pattern. The observation is reproducible; the interpretation is open."

Ronn Torossian · Founder and Chairman, 5W

The Findings

Ten observations from 32,200 prompts across two waves.

All figures below reflect averages across Wave 1 (Jan–Feb 2026) and Wave 2 (Apr–May 2026) within ≤1.5 percentage-point cross-wave tolerance. Findings are presented as observed patterns in this dataset. Interpretation and possible operational implications appear in a separately labeled section later in this report.

01

Two defense technology companies collectively account for a higher Citation Share than the five largest legacy primes in the dataset

Anduril (19.8%) and Palantir (15.2%) collectively accounted for 35.0% of observed Citation Share. Lockheed Martin (5.9%), Northrop Grumman (4.8%), RTX/Raytheon (4.2%), Boeing Defense (3.6%), and General Dynamics (2.6%) collectively accounted for 21.1%. The two defense technology companies operate at approximately $3.5 billion combined annual revenue; the five primes at approximately $247 billion. In this dataset, higher revenue was not consistently associated with higher observed Citation Share, and the relationship trends negative across the broader sample.

02

SpaceX accounted for the highest cross-category Citation Share in the test set

SpaceX accounted for 12.7% of overall Citation Share — third in the dataset — surfacing across three distinct query categories: pure space queries (where it accounted for approximately 60% share within that subset), defense queries (associated with coverage of Starshield and DoD contract activity), and adjacent technology queries. No other company in the test set exhibited comparable cross-category presence in this dataset. SpaceX's observed Citation Share is distributed across multiple query types rather than concentrated in one.

03

The observed source distribution differs from prior 5W benchmarks in related categories

Defense trade press — Defense News, Breaking Defense, The War Zone, War on the Rocks, Defense One — collectively accounted for 36.9% of cited sources in this dataset. In prior 5W benchmarks (banking, venture capital, credit cards), top citation sources were Wikipedia and mainstream financial or technology press. In this dataset, specialized defense trade outlets accounted for the largest share, with Wikipedia (20.7%) anchoring the distribution rather than dominating it. The source distribution reflects observed retrieval behavior during the testing window.

Figure 02 · Revenue and Citation Share — observed distribution

In this dataset, higher revenue was not consistently associated with higher Citation Share.

Each marker plots a defense or aerospace company's approximate annual revenue against its observed Citation Share in this dataset. Markers above the dashed reference line are associated with higher-than-revenue-predicted citation rates; markers below are associated with lower. Revenue is shown on a logarithmic-ish scale. The reference line is included for visual orientation only and is not a regression.

0% 5% 10% 15% 20% $0.1B $1B $5B $25B $75B Annual Revenue (approximate, log-ish scale) AI Citation Share ↑ Above the reference line ↓ Below the reference line Anduril 19.8% · ~$1B revenue Palantir 15.2% · ~$2.5B defense rev. SpaceX 12.7% · ~$10B revenue Shield AI 6.4% Saronic 3.8% Skydio 3.4% Lockheed Martin 5.9% · ~$67B revenue Northrop Grumman 4.8% · ~$40B rev. RTX 4.2% Boeing Defense 3.6% GD 2.6% L3Harris 2.1% BAE 1.4% Modern defense tech Space (cross-category) Legacy defense primes
Source · 5W Defense & Aerospace AI Visibility Index 2026 · Revenue figures per FY2025 public filings and 5W estimates for private companies. Palantir revenue reflects defense-segment estimate only. SpaceX plotted by total revenue including commercial.
04

Sustained founder-named content output is associated with higher observed Citation Share in several cases

Anduril Industries is the company in the test set whose owned domain and founder-named content (Palmer Luckey writing on Substack and X, Trae Stephens publishing essays, Brian Schimpf podcast appearances) collectively appears in retrieved AI responses at frequencies comparable to defense trade press. A similar pattern was previously observed for Andreessen Horowitz in the 5W Venture Capital AI Visibility Index. No legacy prime in the test set exhibited a comparable level of founder-named content output during the testing window. Whether founder content output is causal or correlated with higher Citation Share cannot be determined from this dataset alone.

05

References to prior reputational events appeared in retrieved responses across both waves

References to the Boeing 737 MAX episode appeared in 31% of Boeing Defense responses despite originating in commercial aviation. References to F-35 cost-overrun discussion appeared in 28% of Lockheed Martin responses. References to Palantir contractor and contract-related discussions appeared in 19% of Palantir responses. The references were stable across both waves and may reflect persistent embedding of reputational events in retrieval surfaces such as Wikipedia and high-citation press archives.

06

Named-founder content output is associated with company-level Citation Share variation in this dataset

Named-founder analysis of the test set surfaced Palmer Luckey (Anduril) at the highest individual share, followed by Alex Karp (Palantir), Elon Musk (SpaceX), Peter Thiel (Palantir co-founder, appearing independently), and Brandon Tseng (Shield AI). Companies whose founders maintained public writing or long-form podcast activity during the testing window were associated with higher observed Citation Share than companies whose founders did not. The association does not establish causality; founder activity may be a marker of broader organizational content cadence.

07

Substack and personal-platform content appeared as cited sources at higher rates than in prior 5W benchmarks

Substack-hosted war journalism and national-security analysis appeared in 4.4% of retrieved responses — larger than Defense One's contribution and approximately equal to Bloomberg's defense coverage. Substack has not appeared as a top citation source in prior 5W benchmarks. The observation may reflect a structural change in where defense journalism is currently being published, but other explanations (specific authors, specific topics, retrieval-index timing) cannot be ruled out by this dataset alone.

08

European defense technology companies registered low observed Citation Share in U.S. consumer AI systems

Helsing (Germany), BAE Systems (UK), and Rheinmetall (Germany) collectively accounted for less than 3.5% Citation Share in the test set. Helsing specifically registered at 1.5% during a period of substantial fundraising and Ukraine-deployment activity. Geographic variation in retrieval surface is consistent with patterns 5W has previously observed in other categories (e.g., Chinese AI labs in the AI Companies Index) where companies of substantial commercial importance can register low share in U.S. consumer AI systems while appearing in technical, trade, or government-source retrieval.

Figure 03 · Engine-level Citation Share — observed distribution

Citation Share for the same companies varied across the five AI systems in this dataset.

Observed Citation Share by AI system for the top 10 defense and aerospace companies in this dataset. Darker cells indicate higher observed shares. ChatGPT's distribution was associated with higher Palantir share than other systems — a pattern consistent with retrieval architectures that weight Wikipedia and longer-established entity references more heavily. Gemini and Perplexity were associated with higher Anduril share — consistent with retrieval architectures that weight recency, owned-domain authority, and recent social signals more heavily. Claude's distribution was the most uniform across the top three. Google AI Overviews exhibited the highest within-engine variance across re-runs. These patterns reflect observed model behavior during the testing window; mechanisms producing the differences are not directly observable.

ChatGPT Claude Gemini Perplexity Google AIO Anduril 17.2% 19.4% 22.1% 23.8% 16.5% Palantir 17.8% 14.6% 13.2% 13.7% 16.7% SpaceX 14.1% 12.4% 12.8% 12.1% 12.3% Shield AI 5.6% 7.2% 6.8% 6.4% 5.9% Lockheed Martin 7.2% 5.8% 5.4% 4.7% 6.4% Northrop Grumman 5.8% 4.6% 4.4% 4.1% 5.1% RTX (Raytheon) 5.1% 4.1% 3.9% 3.6% 4.4% Saronic Technologies 3.1% 3.8% 4.2% 4.6% 3.4% Boeing Defense 4.4% 3.4% 3.2% 3.1% 3.8% Skydio 2.9% 3.6% 3.4% 3.8% 3.1% Citation Share Low High →
Source · 5W Defense & Aerospace AI Visibility Index 2026 · All defense and aerospace queries · n = 5,680 prompts per engine across two waves
09

Topic-specific prompts about AI and autonomous systems were associated with higher observed Anduril share

For prompts specifically about autonomous weapons systems, military AI, machine-learning-enabled targeting, or AI-warfare ethics, Anduril's observed Citation Share rose to 28.1% (versus 19.8% overall). Shield AI's rose to 11.4% (versus 6.4% overall). Palantir's rose to 21.3% within AI-warfare ethics queries specifically. Topic-specialization effects of this magnitude were the largest observed in the dataset, suggesting that companies publishing on autonomous systems and AI-related topics may surface at higher rates within topic-specific prompts than within general category prompts. The mechanism producing the topic effect cannot be directly observed.

10

Observed rankings varied across the five AI systems in interpretable patterns

ChatGPT's observed distribution was associated with higher Palantir share (17.8% vs Anduril 17.2%) — a pattern consistent with Wikipedia-weighted retrieval and longer-established entity reference density. Gemini and Perplexity were associated with higher Anduril share (22.1% and 23.8% respectively) — a pattern consistent with retrieval architectures that weight recency, owned-domain authority, and recent social-signal indicators more heavily. Claude's distribution was the most uniform across the top three companies in the dataset. Google AI Overviews exhibited the highest within-engine variance across re-runs. Wikipedia source density was high across all five systems; the variation across systems lay primarily in which other sources appeared alongside Wikipedia. Mechanisms producing these patterns are not directly observable from the dataset.

The Index

Defense & Aerospace AI Visibility Index 2026.

Defense & Aerospace — Top 25
Rank Company Citation Share
01Anduril Industries19.8%
02Palantir Technologies15.2%
03SpaceX12.7%
04Shield AI6.4%
05Lockheed Martin5.9%
06Northrop Grumman4.8%
07RTX (Raytheon)4.2%
08Saronic Technologies3.8%
09Boeing Defense, Space & Security3.6%
10Skydio3.4%
11General Dynamics2.6%
12Planet Labs2.4%
13L3Harris Technologies2.1%
14Rocket Lab1.8%
15Helsing1.5%
16BAE Systems1.4%
17Epirus1.2%
18BlackSky1.0%
19Mach Industries0.9%
20Rebellion Defense0.7%
21Vannevar Labs0.6%
22Capella Space0.5%
23Astranis0.4%
24Stoke Space0.3%
25Relativity Space0.3%
Revenue and Citation Share — Observed Distribution
Segment ~Combined Revenue Citation Share
Anduril + Palantir (defense tech)~$3.5B35.0%
Top 5 legacy primes (LMT + NOC + RTX + BA + GD)~$247B21.1%
Defense tech leaders / Primes ratio1 / 701.66x

Revenue figures reflect FY2025 public filings (primes), most recent disclosed valuations and 5W revenue estimates (Anduril, Palantir defense segment). Combined revenue calculation: Anduril ~$1B + Palantir defense segment ~$2.5B = $3.5B. Top 5 primes: Lockheed Martin $67B + Northrop Grumman $40B + RTX $70B + Boeing Defense $25B + General Dynamics $45B = $247B. The defense tech leaders capture 1.66x the Citation Share with approximately 1/70 the revenue.

Top Cited Sources — All Defense Queries
Rank Source Share of Citations
01Wikipedia20.7%
02Defense News10.8%
03Breaking Defense8.9%
04The War Zone7.2%
05Reuters & Bloomberg defense (combined)6.8%
06Defense company-owned domains (combined)6.4%
07NYT, WSJ, FT defense desks (combined)5.9%
08War on the Rocks5.4%
09Government sources (DoD, GAO, CRS)5.1%
10Defense One4.6%
11Substack (war journalism)4.4%
12All other publishers13.8%

"In this dataset, defense trade press — Defense News, Breaking Defense, The War Zone, War on the Rocks, Defense One — collectively accounted for approximately 37% of cited sources about U.S. defense and aerospace. The pattern is the inverse of what we observed in financial-services categories, where mainstream financial press dominated. Whether this reflects the structural state of defense publishing or a transient retrieval pattern remains an open question."

Ronn Torossian · Founder and Chairman, 5W

Methodology

How the Index was built.

The Defense & Aerospace AI Visibility Index 2026 analyzed 28,400 prompts across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews, run in two independent waves to test for stability across retrieval drift, model updates, and active news cycles in this category.

Two-wave structure

Wave 2 used the same prompt set as Wave 1 with no modification. Only findings stable across both waves within reporting tolerance (≤1.5 percentage-point delta on company-level Citation Share; ≤2.0 pp on source-level share) are published here. Findings unstable across waves were excluded from the Index.

Prompt design

Queries simulated real founder, investor, journalist, policymaker, and procurement-officer research behavior. Prompts included branded company queries ("What does Anduril Industries do?"), non-branded category queries ("Best defense tech startup in 2026"), comparison queries ("Anduril vs Lockheed Martin"), capability queries ("Autonomous military drone makers", "AI-enabled targeting systems"), procurement-oriented queries ("DoD prime contractors 2026"), and ethics-and-policy queries ("AI in warfare ethics"). Prompts were distributed evenly across the five engines so each engine received the same prompt mix per category.

Seven defense and aerospace categories measured

  1. Modern defense tech (Anduril, Palantir, Shield AI, Saronic, Skydio, others)
  2. Legacy defense primes (Lockheed Martin, Northrop Grumman, RTX, Boeing Defense, General Dynamics, L3Harris)
  3. Space and launch (SpaceX, Rocket Lab, Stoke Space, Relativity Space)
  4. Earth observation and satellite intelligence (Planet Labs, BlackSky, Capella Space, Astranis)
  5. Autonomous systems and AI-in-warfare
  6. European defense tech (Helsing, BAE Systems — extended dataset)
  7. Procurement and DoD contracting

Sampling

Each prompt was issued three times per engine within a wave, with responses sampled at varied time-of-day windows to reduce within-engine retrieval drift. Reported Citation Share values represent the average of all retrieved responses for a prompt across both waves.

What "Citation Share" means operationally

Three distinct response signals were tracked per retrieved answer:

For the headline metric reported as Citation Share, all three signals were weighted equally per retrieved response. Source-level analyses (the publisher rankings in Figure 01) used source citations only. Equal weighting of three structurally different signals is an analytical choice made by 5W for the franchise; alternative weighting schemes (recommendation-weighted, citation-weighted, mention-weighted) would produce different rankings. The headline ranking is best understood as a composite signal of model surface frequency, not as a single underlying construct.

Citation Share is an internal composite metric. It is not a standard industry measure. It should not be interpreted as market share, operational performance, contract-award likelihood, AI-system endorsement, accuracy of representation, financial value, or strategic advantage. It is a proxy for how often a company surfaces in retrieved AI responses during a defined testing window, under a defined prompt distribution.

Cross-engine normalization

Each engine was weighted equally in aggregate figures, regardless of differences in response length or default-citation frequency per engine. This was a deliberate choice; weighting by raw citation volume would have over-weighted Perplexity and Google AI Overviews, both of which return more citations per response than ChatGPT, Claude, or Gemini by default. Where per-engine results diverge, those differences are reported separately (Figure 03).

SpaceX cross-category note

SpaceX's score reflects all SpaceX-related citations across defense, space, and adjacent technology queries. SpaceX's defense-specific Citation Share (Starshield, DoD contracts only) was approximately 7.4%; SpaceX's space-specific Citation Share was approximately 62.3%; the blended figure reported in the Top 25 reflects the company's total presence across the full test set. The cross-category nature of SpaceX is distinct among the 25 companies measured.

Allied defense exclusion

Israeli defense companies (Rafael, Elbit, IAI), Indian defense companies, and other allied defense firms were tested in an extended dataset but excluded from the top-25 ranking to maintain a U.S. frame consistent with prior 5W benchmarks. European defense technology firms (Helsing, BAE Systems) are included in the top-25 ranking due to substantial U.S. commercial presence and inclusion in U.S.-readership defense trade press.

What this benchmark does not measure

This benchmark does not measure model accuracy, factual correctness of retrieved responses, sentiment of citations, or quality of underlying source content. A company appearing frequently in retrieved responses may be cited favorably or critically; both forms of mention count equally for the headline metric. The benchmark also does not separately attribute citation patterns to underlying mechanisms (training-data composition, retrieval-index weighting, ranking heuristics); these mechanisms are increasingly entangled in production AI systems and not externally observable. Engine-level differences reported in Figure 03 reflect observed differences in retrieval output, not directly observed differences in underlying architecture.

Limitations

Because this study measures AI-generated outputs across a defined time window, results reflect model behavior during that period and may not generalize across future model versions, retrieval-index updates, or different prompt distributions. AI systems, training data, retrieval indexes, and ranking mechanisms evolve continuously; results may shift outside the testing window. Active news cycles during the testing window (Ukraine, Israel/Gaza, Pacific tensions) may have influenced cross-wave volatility for select firms. The benchmark is best understood as a structured snapshot of observed model behavior across two waves, not as a continuous live measurement and not as a predictive instrument. The full prompt set, per-engine response logs, and category-level datasets are available on request for replication.

Disclosure. 5W has commercial relationships across the defense and aerospace industry and operates as a senior advisor to Curium.io, the team that coined Generative Engine Optimization. Companies cited in this benchmark were selected based on category relevance and observable presence in the test set; inclusion does not imply a commercial relationship. 5W does not accept compensation for ranking placement and did not consult with any company in the top-25 ranking regarding placement, methodology, or interpretation prior to publication.

Interpretation · Possible Operational Implications

Five possible operational implications of the observed patterns.

This section is interpretive. The items below are not findings — they are possible operational implications, offered as informed commentary on the patterns reported earlier in this report. Each implication is presented as a hypothesis suggested by the data, not as a recommendation or prescription. Causal relationships between any operational practice and observed Citation Share cannot be established by this dataset alone.

Founder-named content output may be a relevant input to Citation Share

Anduril's observed Citation Share is associated with a sustained pattern of founder-named content during the testing window — including Palmer Luckey's writing on personal platforms, Trae Stephens's published essays, and Brian Schimpf's long-form podcast appearances. The data does not establish causation; founder content output may be a marker of broader organizational content cadence rather than an independent driver. The hypothesis worth testing: companies investing in sustained named-founder content output may observe corresponding changes in Citation Share over time. The hypothesis is testable through controlled re-measurement.

Coverage in defense trade press may be a relevant input given the observed publisher distribution

Defense News, Breaking Defense, The War Zone, War on the Rocks, and Defense One collectively accounted for 36.9% of cited sources in this dataset. The observation is consistent with a hypothesis that earned coverage in specialized defense trade press may be associated with higher Citation Share, given the relative weight these publishers appear to carry in observed AI retrieval. Companies whose earned-media activity is concentrated in general-press defense desks may be under-represented in retrieval relative to companies whose activity is concentrated in trade press. The relationship has not been tested causally.

Wikipedia entries (company-level and founder-level) may be a relevant retrieval surface to audit

Wikipedia accounted for 20.7% of cited sources in this dataset — the largest single source. Companies whose Wikipedia entries are sparse, outdated, or omit relevant context for the company or its founders may be under-represented in observed retrieval relative to companies with more developed Wikipedia entries. The hypothesis is not directly tested by this benchmark but is consistent with the source distribution observed.

Reputational events appearing in retrieval may persist beyond their news-cycle window

References to the Boeing 737 MAX episode, F-35 cost-overrun discussion, and Palantir contract-related discussions all appeared in retrieved responses at rates between 19% and 31%. These references were stable across both waves. The pattern is consistent with a hypothesis that once reputational events are embedded in high-citation retrieval surfaces (Wikipedia, press archives), short-term news-cycle decay may not remove them from retrieval. The hypothesis suggests that crisis-response strategies designed primarily to ride out news cycles may be incomplete relative to strategies that also produce sustained alternative-narrative content over a longer horizon.

Major procurement-cycle events may interact with pre-existing retrieval state

Major DoD contract awards and procurement decisions generate intense press coverage that subsequently appears in retrieval surfaces. The hypothesis worth testing: companies entering a major procurement cycle with already-developed Citation Share may be associated with different retrieval outcomes than companies entering with limited prior Citation Share. The hypothesis is consistent with patterns observed in this dataset but is not directly tested.

FAQ

Defense & Aerospace AI Visibility Index — Q&A.

What is the Defense & Aerospace AI Visibility Index?

The Defense & Aerospace AI Visibility Index 2026 is the first public two-wave benchmark measuring how often U.S. defense tech, space, and legacy defense companies are surfaced, cited, and recommended inside ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. It was produced by 5W, the AI Communications Firm.

What is Citation Share?

Citation Share is the share of AI responses in which a company is named, recommended, or cited as a referenced source. Mentions, recommendations, and source citations are tracked separately and weighted equally unless otherwise noted.

Which defense company has the highest observed Citation Share?

Anduril Industries accounted for 19.8% of observed Citation Share across the five AI systems — the highest of any defense or aerospace company in the test set. Palantir Technologies followed at 15.2%, SpaceX at 12.7%, Shield AI at 6.4%, and Lockheed Martin at 5.9%.

Do legacy defense primes have the highest Citation Share in AI systems?

In this dataset, no. The five largest legacy defense primes — Lockheed Martin, Northrop Grumman, RTX (Raytheon), Boeing Defense, and General Dynamics — collectively accounted for 21.1% of Citation Share. Anduril and Palantir together accounted for 35.0% — a higher combined share than the top five primes — despite operating at a fraction of the combined revenue. Citation Share in this benchmark is a composite proxy for model surface frequency, not a measure of market share, contract activity, or operational performance.

What sources do AI systems cite about defense and aerospace?

In this dataset, Wikipedia (20.7%), Defense News (10.8%), Breaking Defense (8.9%), and The War Zone (7.2%) accounted for the largest share of observed defense and aerospace citations. War on the Rocks, Defense One, Reuters, Bloomberg, NYT, and Substack-based war journalism followed. Company-owned domains accounted for 6.4% in aggregate.

What is associated with Anduril's high observed Citation Share?

Anduril Industries' observed Citation Share is associated with a sustained pattern of named-founder content output during the testing window (Palmer Luckey writing on personal platforms, Trae Stephens publishing essays, Brian Schimpf appearing on long-form podcasts), and with consistent coverage in defense trade press. A similar pattern of named-founder content output was previously associated with Andreessen Horowitz in the 5W Venture Capital AI Visibility Index. Causation cannot be established by this dataset alone; the association is consistent with a hypothesis worth further testing.

How does SpaceX appear in defense queries?

SpaceX accounted for 12.7% of overall Citation Share in this dataset — third in the test set — through cross-category presence. SpaceX accounted for approximately 60% share within space-specific queries, appeared in defense queries (associated with Starshield and DoD contract coverage), and appeared in adjacent technology queries. The company's cross-category citation pattern is distinct among the 25 companies measured.

Did references to prior reputational events appear in retrieved responses?

References to certain prior reputational events appeared in retrieved responses across both waves. References to the Boeing 737 MAX episode appeared in approximately 31% of Boeing Defense responses despite originating in commercial aviation. References to F-35 cost-overrun discussion appeared in approximately 28% of Lockheed Martin responses. References to Palantir contractor and contract-related discussions appeared in approximately 19% of Palantir responses. The pattern is consistent with persistent embedding of reputational events in high-citation retrieval surfaces such as Wikipedia and press archives.

Which AI system produced the most uniform observed distribution across the top defense companies?

In this dataset, Claude produced the most uniform observed distribution across the top three defense companies. ChatGPT was associated with higher Palantir share — a pattern consistent with Wikipedia-weighted retrieval. Gemini and Perplexity were associated with higher Anduril share — a pattern consistent with retrieval architectures that weight recency and recent social signals more heavily. Google AI Overviews exhibited the highest within-engine variance across re-runs. These patterns reflect observed model behavior during the testing window; mechanisms producing the differences are not directly observable.

What is GEO?

GEO — Generative Engine Optimization — is the practice of building brand authority and content infrastructure that AI systems surface, cite, and recommend. It is the discipline replacing SEO in an AI-mediated discovery layer.

How many defense and aerospace companies were tested?

The Index tested 25 leading U.S. defense and aerospace companies — 10 modern defense tech firms, 8 space companies, and 7 legacy defense primes — across 28,400 prompts in two independent waves between January and May 2026.

Work with 5W

5W measures Citation Share, audits retrieval surfaces, and builds the content infrastructure that AI systems cite.

Earned media in the publishers that AI systems cite most. Founder content programs. Generative Engine Optimization. Wikipedia audit and remediation. Quarterly Citation Share measurement. The methodology used in this report is available for client benchmarks against named competitive sets.

Request the Full Dataset Talk to 5W
Related Research

More from the 5W AI Visibility Index series.

Volume II · Financial Services

Venture Capital AI Visibility Index 2026
Volume III · Financial Services

Credit Cards AI Visibility Index 2026
Volume IV · Technology

AI Companies AI Visibility Index 2026