Protocol: who AI engines name in artificial intelligence — US, November 2026
This protocol is published before data collection begins and does not change while collection is under way. It applies methodology version 1.0.
In short. From October 26 to November 8, 2026, Visibilitas will ask 6 AI engines, once a day, 8 questions about who is authoritative in artificial intelligence in the United States: news outlets, journalists, and creators on YouTube, podcasts, newsletters, LinkedIn, TikTok and Instagram. The questions are listed below in full and can be downloaded before the first answer is collected.
What will the report rank?
Three panels, each ranked separately:
- News outlets that AI engines name as authoritative on artificial intelligence.
- Journalists that AI engines name as authoritative on artificial intelligence.
- Creators, one ranking per platform: YouTube, podcasts, newsletters, LinkedIn, TikTok, Instagram.
The rankings show whom AI engines name when they are asked these questions. They are not a judgment by Visibilitas on the authority, quality or accuracy of anyone named.
Which questions will be asked?
| ID | Panel | Question |
|---|---|---|
| P01 | News outlets | What are the most authoritative news outlets covering artificial intelligence in the United States? |
| P02 | Journalists | Who are the most authoritative journalists covering artificial intelligence in the United States? |
| P03 | Creators — YouTube | Who are the most authoritative YouTube creators on artificial intelligence in the United States? |
| P04 | Creators — podcasts | What are the most authoritative podcasts about artificial intelligence in the United States? |
| P05 | Creators — newsletters | What are the most authoritative newsletters about artificial intelligence in the United States? |
| P06 | Creators — LinkedIn | Who are the most authoritative LinkedIn creators on artificial intelligence in the United States? |
| P07 | Creators — TikTok | Who are the most authoritative TikTok creators on artificial intelligence in the United States? |
| P08 | Creators — Instagram | Who are the most authoritative Instagram creators on artificial intelligence in the United States? |
The questions are submitted exactly as written, with no system instructions, no personal account and no conversation history. None of them names an entity or asks for a number of names.
Download: prompts.csv (SHA-256 884f5828667c31b8a3ed1fe4c49432dad236ad7846db9cfd7b91efa4f0177e71).
How will the answers be collected?
| Parameter | Value |
|---|---|
| Engines | ChatGPT, ChatGPT Search, Gemini, Google AI Overviews, Perplexity, Copilot (see methodology) |
| Location and language | United States, English |
| Collection period | October 26 – November 8, 2026 (14 days) |
| Repetitions | each question to each engine once a day: 14 per question and engine |
| Answers expected | 8 questions × 6 engines × 14 days = 672 |
If an engine cannot be reached on some days, the report states it with the number of missing answers. Missing answers are never estimated.
How will the names be counted?
The rules are those of the methodology for rankings of outlets, journalists and creators:
- every name is verified before it is counted; names that cannot be verified, or that an engine invented, are not ranked and are reported separately;
- each ranking is ordered by Recurrence (the share of answers to its question that name the entity), with ties broken by Breadth (the number of engines that name it);
- an entity enters a ranking only if it is named in at least 5% of the answers to its question;
- the same person can appear in more than one ranking: for example as a journalist and as a newsletter author.
When will results be published?
The report, its rankings and the full dataset (prompts, every answer, every cited source, counts) will be published on /reports/ after collection ends, under a CC BY 4.0 license.
Results are obtained solely by querying AI engines, with a public methodology that is defined and published before each data collection. Prompts, dates and raw data can be downloaded: anyone can repeat the collection and verify the results.