When a program office, a staffer or an investor asks an AI assistant which American companies build military drones, who gets named? We ask 20 buyer questions of Claude + Gemini every week and count.
Share of Answer is the percentage of the 40 scored answers that name the company or one of its products. Recommended is the share that explicitly put it forward. Position is where in the answer it first appears (1 = named first). First edition — week-over-week change appears from the second run.
| # | Company | Share of Answer | Claude | Gemini | Recommended | Avg. position | Δ vs last run |
|---|---|---|---|---|---|---|---|
| 1 | AeroVironment | 80% | 65% | 50% | 3.4 | new | |
| 2 | Anduril Industries | 60% | 65% | 33% | 5.1 | new | |
| 3 | Shield AI | 45% | 45% | 23% | 7.3 | new | |
| 4 | Northrop Grumman | 40% | 40% | 8% | 8.2 | new | |
| 5 | Red Cat Holdings | 45% | 35% | 23% | 5.4 | new | |
| 6 | Kratos Defense & Security Solutions | 40% | 20% | 8% | 8.4 | new | |
| 7 | Neros Technologies | 25% | 35% | 25% | 4.3 | new | |
| 8 | Skydio | 30% | 30% | 25% | 2.5 | new | |
| 9 | Boeing / Insitu | 30% | 25% | 10% | 7.7 | new | |
| 10 | General Atomics Aeronautical Systems | 30% | 25% | 20% | 5.7 | new | |
| 11 | Lockheed Martin | 30% | 20% | 3% | 9.5 | new | |
| 12 | Performance Drone Works | 25% | 25% | 13% | 6.2 | new | |
| 13 | Textron Systems | 20% | 20% | 13% | 4.6 | new | |
| 14 | Firestorm Labs | 10% | 5% | 3% | 6.0 | new | |
| 15 | Edge Autonomy | 5% | 5% | 3% | 8.5 | new |
One sample per question per engine per week, so each company's number moves in steps of about 2.5 points per answer. Read it as a ranking with wide error bars, not a stock price. Per-engine columns are that engine's own share; the headline share pools every engine equally. The full data, including every edition, is in data.json (CC BY 4.0).
Domains the engines cited while answering, ranked by how many of the 40 answers cited them. A roster company's own site is tagged; everything else is the third-party coverage that actually decides visibility.
Written the way buyers ask, never naming a roster company. The set is frozen: changing a question changes the denominator for everyone, so additions become a new index edition, not a quiet edit.
Top three by share of answer in each weekly edition. First measured 2026-09-28.
AI Share of Answer = scored answers that name the company ÷ all scored answers, as a percentage. Every question carries equal weight; every engine carries equal weight. It is a plain mention rate — it is not prominence-weighted, and it says nothing about whether what the engine said was accurate or flattering.
The page reports counts, not characterizations. If a roster entry is wrong (a product name misattributed, a domain missing, a company that no longer exists), email [email protected]. Corrections are applied to the roster and noted in the next edition.
We make true things visible. We never make false things plausible.
The free check writes ten buyer questions for your niche, asks ChatGPT, Claude and Gemini, and shows you who got named instead of you. Takes about four minutes.
Run a free visibility checkThe new table, the movers, and what changed in the sources the engines trust. Nothing else.