London was the city that ChatGPT, Claude, Gemini and Perplexity named most often. London appeared in 153 of 288 answers (53.1%) measured on October 28, 2026. New York City came next at 137 of 288 (47.6%). The two confidence intervals overlap, so the data cannot confirm that London is ahead. San Francisco was named first more often than any other city, in 75 of 288 answers. Google AI Overviews appeared on 61 of 61 searches (100%).
RESEARCH · Published May 21, 2026 · Updated October 30, 2026
AI City Index 2.0: London and New York City Lead AI Answers
AI answers analyzed across ChatGPT, Claude, Gemini and Perplexity
of answers named London, the most-named (153 of 288)
tracked cities named at least once (of 25)
of Google searches showed an AI Overview (61 of 61)
Version 2. This version replaces the first version of this study, first published May 21, 2026. It uses measured counts, so the two versions are not comparable. Version 2 asks 24 questions three times each on ChatGPT, Claude, Gemini and Perplexity (288 answers) and adds Google AI Overviews. Every rate comes with its count and a 95% interval.
What does the data show?
153/288 (53.1%)
75/288 (26.0%)
22/72 (30.6%)
61/61 (100%)
32/61 (52.5%)
40/288 (13.9%)
0/12 (0%)
12/288 (4.2%)
Appearance rate across 288 four-engine answers. The bar is the 95% confidence interval; overlapping bars cannot be told apart at this sample size.
Which cities do AI engines name most often?
| # | Name | Named in | Appearance rate | 95% interval |
|---|---|---|---|---|
| 1 | London | 153/288 | 53.1% | 47.4–58.8% |
| 2 | New York City | 137/288 | 47.6% | 41.9–53.3% |
| 3 | Boston | 118/288 | 41.0% | 35.4–46.7% |
| 4 | Beijing | 116/288 | 40.3% | 34.8–46.0% |
| 5 | San Francisco | 103/288 | 35.8% | 30.4–41.5% |
| 6 | Toronto | 100/288 | 34.7% | 29.5–40.4% |
| 7 | Singapore | 89/288 | 30.9% | 25.8–36.5% |
| 8 | Paris | 87/288 | 30.2% | 25.2–35.7% |
| 9 | Montreal | 53/288 | 18.4% | 14.4–23.3% |
| 10 | Shanghai | 49/288 | 17.0% | 13.1–21.8% |
| 11 | Shenzhen | 48/288 | 16.7% | 12.8–21.4% |
| 12 | Seoul | 47/288 | 16.3% | 12.5–21.0% |
| 13 | Bengaluru | 42/288 | 14.6% | 11.0–19.1% |
| 14 | Tel Aviv | 37/288 | 12.8% | 9.5–17.2% |
| 15 | Berlin | 36/288 | 12.5% | 9.2–16.8% |
How do all 25 tracked cities rank?
Own-domain citation is the share of 288 grounded answers that cited the official domains of each name (subdomains included). Citation share is of all 2,523 citations.
| # | City | Named in | Appearance rate | 95% interval | First mention | Own-domain citation | Citation share |
|---|---|---|---|---|---|---|---|
| 1 | London | 153/288 | 53.1% | 47.4–58.8% | 3 | 2.4% (7/288) | 0.4% |
| 2 | New York City | 137/288 | 47.6% | 41.9–53.3% | 2.4 | 3.1% (9/288) | 0.5% |
| 3 | Boston | 118/288 | 41% | 35.4–46.7% | 3.7 | 0% (0/288) | 0% |
| 4 | Beijing | 116/288 | 40.3% | 34.8–46% | 3.7 | 1.4% (4/288) | 0.2% |
| 5 | San Francisco | 103/288 | 35.8% | 30.4–41.5% | 2.1 | 0% (0/288) | 0% |
| 6 | Toronto | 100/288 | 34.7% | 29.5–40.4% | 4.4 | 0% (0/288) | 0% |
| 7 | Singapore | 89/288 | 30.9% | 25.8–36.5% | 5.1 | 4.2% (12/288) | 0.5% |
| 8 | Paris | 87/288 | 30.2% | 25.2–35.7% | 5 | 0% (0/288) | 0% |
| 9 | Montreal | 53/288 | 18.4% | 14.4–23.3% | 4.6 | 0% (0/288) | 0% |
| 10 | Shanghai | 49/288 | 17% | 13.1–21.8% | 4.5 | 1% (3/288) | 0.2% |
| 11 | Shenzhen | 48/288 | 16.7% | 12.8–21.4% | 5.1 | 0.3% (1/288) | 0% |
| 12 | Seoul | 47/288 | 16.3% | 12.5–21% | 4.7 | 1% (3/288) | 0.1% |
| 13 | Bengaluru | 42/288 | 14.6% | 11–19.1% | 6.5 | 0% (0/288) | 0% |
| 14 | Tel Aviv | 37/288 | 12.8% | 9.5–17.2% | 5.8 | 0% (0/288) | 0% |
| 15 | Berlin | 36/288 | 12.5% | 9.2–16.8% | 5.8 | 0% (0/288) | 0% |
| 16 | Tokyo | 33/288 | 11.5% | 8.3–15.7% | 5.7 | 0% (0/288) | 0% |
| 17 | Amsterdam | 31/288 | 10.8% | 7.7–14.9% | 5.6 | 0% (0/288) | 0% |
| 18 | Zurich | 28/288 | 9.7% | 6.8–13.7% | 6.2 | 0% (0/288) | 0% |
| 19 | Dubai | 26/288 | 9% | 6.2–12.9% | 5.7 | 0.7% (2/288) | 0.1% |
| 20 | Los Angeles | 24/288 | 8.3% | 5.7–12.1% | 3.9 | 0% (0/288) | 0% |
| 21 | Stockholm | 22/288 | 7.6% | 5.1–11.3% | 5.8 | 0% (0/288) | 0% |
| 22 | Abu Dhabi | 20/288 | 6.9% | 4.5–10.5% | 3.2 | 0% (0/288) | 0% |
| 23 | Riyadh | 13/288 | 4.5% | 2.7–7.6% | 2.6 | 0% (0/288) | 0% |
| 24 | Sydney | 9/288 | 3.1% | 1.7–5.8% | 7.2 | 0% (0/288) | 0% |
| 25 | Hyderabad | 8/288 | 2.8% | 1.4–5.4% | 6.6 | 0% (0/288) | 0% |
How do the top eight cities compare?
| # | Name | Appearance rate | Answers and interval | Other measures |
|---|---|---|---|---|
| 1 | London | 53.1% | 153/288 answers · 95% CI 47.4–58.8% | first-mention position: 3; owned-domain citation (7/288): 2.4%; citation share: 0.4% |
| 2 | New York City | 47.6% | 137/288 answers · 95% CI 41.9–53.3% | first-mention position: 2.4; owned-domain citation (9/288): 3.1%; citation share: 0.5% |
| 3 | Boston | 41.0% | 118/288 answers · 95% CI 35.4–46.7% | first-mention position: 3.7; owned-domain citation (0/288): 0%; citation share: 0% |
| 4 | Beijing | 40.3% | 116/288 answers · 95% CI 34.8–46% | first-mention position: 3.7; owned-domain citation (4/288): 1.4%; citation share: 0.2% |
| 5 | San Francisco | 35.8% | 103/288 answers · 95% CI 30.4–41.5% | first-mention position: 2.1; owned-domain citation (0/288): 0%; citation share: 0% |
| 6 | Toronto | 34.7% | 100/288 answers · 95% CI 29.5–40.4% | first-mention position: 4.4; owned-domain citation (0/288): 0%; citation share: 0% |
| 7 | Singapore | 30.9% | 89/288 answers · 95% CI 25.8–36.5% | first-mention position: 5.1; owned-domain citation (12/288): 4.2%; citation share: 0.5% |
| 8 | Paris | 30.2% | 87/288 answers · 95% CI 25.2–35.7% | first-mention position: 5; owned-domain citation (0/288): 0%; citation share: 0% |
ChatGPT named Shenzhen in 22 of 72 answers. The other engines did so in 6 to 12 of 72.
ChatGPT
72 asked · 72 returned · 72 grounded
Do the four engines name the same cities?
Claude
72 asked · 72 returned · 72 grounded
Gemini
72 asked · 72 returned · 72 grounded
Perplexity
72 asked · 72 returned · 72 grounded
Google AI Overviews were captured through 61 U.S. desktop searches and scored only where an overview appeared, so five-engine figures use 288 + 61 = 349 answers.
What do Google AI Overviews show?
Trigger rate: 61 of 61 searches
Google AI Overviews appeared on 61 of 61 searches (100%, CI 94.1–100%). The other 0 searches showed no overview. Within the 61 overviews, London was named in 32 (52.5%), New York City was named in 32 (52.5%), San Francisco was named in 30 (49.2%), Boston was named in 28 (45.9%). linkedin.com was the most-cited source, with 19 citations and an appearance in 18 of 61 overviews.
| Five-engine rank | Name | Five-engine rate | Four-engine rank |
|---|---|---|---|
| 1 | London | 53.0% | 1 |
| 2 | New York City | 48.4% | 2 |
| 3 | Boston | 41.8% | 3 |
| 4 | Beijing | 38.1% | 4 |
| 5 | San Francisco | 38.1% | 5 |
| 6 | Toronto | 35.0% | 6 |
| 7 | Singapore | 31.8% | 7 |
| 8 | Paris | 30.4% | 8 |
| 9 | Montreal | 19.5% | 9 |
| 10 | Shanghai | 16.3% | 10 |
Which city does each question surface first?
Six questions from the set. Each tile uses that question's own four-engine answers as the denominator.
| Question | Leader | Next | AI Overview |
|---|---|---|---|
| Q01 · 12 answers | London 100.0% | Beijing 100.0%; Boston 100.0% | AI Overview shown on 2 of 2 searches |
| Q02 · 12 answers | London 100.0% | Beijing 100.0%; New York City 100.0% | AI Overview shown on 3 of 3 searches |
| Q03 · 12 answers | London 100.0% | Beijing 100.0%; Boston 100.0% | AI Overview shown on 3 of 3 searches |
| Q04 · 12 answers | San Francisco 100.0% | London 100.0%; Beijing 91.7% | AI Overview shown on 3 of 3 searches |
| Q05 · 12 answers | London 100.0% | Boston 100.0%; New York City 100.0% | AI Overview shown on 3 of 3 searches |
| Q06 · 12 answers | Singapore 100.0% | Beijing 8.3%; Toronto 8.3% | AI Overview shown on 3 of 3 searches |
The four engines produced 2,523 citations in total. brookings.edu was cited most often, with 66 citations (2.6%) across 40 answers. Next came en.wikipedia.org (50), 5wpr.com (38), secondtalent.com (32), startupproject.org (32), dealroom.co (31).
Four AI engines · share of 288 grounded answers
Which domains do AI engines cite most?
Google AI Overviews · share of 61 overviews
Which types of sources get cited?
Share of all citations by the type of site cited. Source types were reclassified on October 29, 2026. 'Company / brand sites' now covers only the official sites of companies, brands and schools; publishers, directories, trade bodies and other sites have their own types, and sites that could not be placed are 'Other / unclassified'.
| Source type | Four AI engines (of 2,523 citations) | Google AI Overviews (of 172 citations) |
|---|---|---|
| Company / brand sites | 799 (31.7%) | 39 (22.7%) |
| Trade / specialist press | 313 (12.4%) | 6 (3.5%) |
| Institutions / trade bodies | 284 (11.3%) | 18 (10.5%) |
| General news / business press | 267 (10.6%) | 5 (2.9%) |
| Research / data providers | 251 (9.9%) | 15 (8.7%) |
| Official / government | 198 (7.8%) | 9 (5.2%) |
| Directories / rankings | 182 (7.2%) | 9 (5.2%) |
| Community / social | 108 (4.3%) | 70 (40.7%) |
| Reference | 79 (3.1%) | 0 (0%) |
| Review / comparison publishers | 40 (1.6%) | 1 (0.6%) |
| Other / unclassified | 2 (0.1%) | 0 (0%) |
How can cities improve their AI visibility?
These steps follow from the measured results above. They describe what the engines cited in this study and are not a guarantee of results.
| Step | What this study measured |
|---|---|
| Put your facts on your own domain. | Singapore's own domains were cited in 4.2% of grounded answers, New York City's in 3.1% and London's in 2.4%. |
| Keep your facts accurate on the sites engines cite. | The most-cited sites other than the tracked names' own were brookings.edu (in 13.9% of grounded answers), 5wpr.com (13.2%), en.wikipedia.org (12.2%), secondtalent.com (11.1%) and startupproject.org (10.1%). Trade and specialist press supplied 12.4% of citations and institutions and trade bodies 11.3%. |
| Check each engine separately. | Perplexity named Paris in 41.7% of answers against 15.3% on ChatGPT. ChatGPT named Shenzhen in 30.6% of answers against 8.3% on Gemini. |
| Do not ignore video and community sites. | Community and social sites supplied 40.7% of Google AI Overview citations against 4.3% in the four chat engines. linkedin.com was cited in 29.5% of overviews, youtube.com in 19.7% and facebook.com in 18.0%. |
| Watch names outside the tracked list. | Seattle (34), Austin (26), OpenAI (26), Google DeepMind (20) were named often but were not tracked. |
| Re-run the same questions every month. | Compare counts and intervals, not ranks. The 24 questions are in the appendix. |
What does this mean for cities and the teams that promote them?
The results separate being mentioned from being cited. London was named in 153 of 288 answers (53.1%), yet London's own domains were cited in 7 of 288 (2.4%). Singapore ranked seventh on mentions at 30.9% but had the highest own-domain citation rate, 12 of 288 (4.2%). Engines name large, well-known cities often. They cite the sites that state clear, current facts.
That gap is an opportunity for economic development agencies, tourism boards, universities and the companies that compete for talent and investment. A city that publishes plain, sourced answers to the questions in the appendix gives engines something to cite, instead of leaving the answer to third parties. Open ground exists too: none of the 25 tracked cities was named in any of the 12 answers about emerging AI hubs in Latin America, so there is room for a city to be first.
- Publish the facts engines are asked for. Investment, talent, research centers and AI labs, each with a date and a source on your own domain.
- Earn coverage where engines already look. This study found brookings.edu, Wikipedia, trade press and institutional sites among the most-cited sources.
- Measure each engine on its own. The same question produced different city lists on ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews.
5W's public affairs and government relations and generative engine optimization teams help cities, companies and institutions do this work. See more studies in the 5W AI Visibility Index and the 5W research library.
For travel-category data, see the Theme Parks AI Visibility Index and the Airlines & Hotels AI Visibility Index.
Frequently asked questions
Which city do AI engines name most often?
London, named in 153 of 288 four-engine answers (53.1%, CI 47.4–58.8%). New York City was second at 137 of 288 (47.6%). The intervals overlap, so the two cannot be separated statistically.
Which city is named first most often?
San Francisco was the first city named in 75 of 288 answers, with an average first position of 2.1.
How often do Google AI Overviews appear for these questions?
An overview appeared on 61 of 61 searches (100%, CI 94.1–100%). Within those 61 overviews, London was named most often, in 32 (52.5%).
Which sources do the engines cite?
Across 2,523 four-engine citations, brookings.edu was cited most (66 citations, in 40 of 288 grounded answers), followed by en.wikipedia.org (50 citations).
Do the four engines agree?
Only in part. ChatGPT named Shenzhen in 22 of 72 answers, against 6 to 12 of 72 on the other engines. Perplexity named Boston in 36 of 72 answers, against 26 to 28 of 72 on the other engines.
Which source types do AI engines cite most?
Company and brand sites supplied 799 of 2,523 four-engine citations (31.7%). Trade and specialist press followed with 313 (12.4%), then institutions and trade bodies with 284 (11.3%) and general news and business press with 267 (10.6%).
Which sources do Google AI Overviews cite?
linkedin.com was cited in 18 of 61 overviews (29.5%), followed by youtube.com (19.7%) and facebook.com (18.0%). Community and social sites made up 70 of 172 AI Overview citations (40.7%), against 4.3% in the four chat engines.
Which city is cited from its own website most often?
Singapore, whose own domains were cited in 12 of 288 grounded answers (4.2%). New York City followed at 9 of 288 (3.1%) and London at 7 of 288 (2.4%). 17 of the 25 tracked cities had no own-site citation.
Were any cities named as emerging AI hubs in Latin America?
No tracked city was. For the question "Which cities are emerging AI hubs in Latin America?", none of the 25 tracked cities appeared in any of the 12 four-engine answers. The study scored only the 25 tracked cities, so other cities may have been named.
How many questions and engines did this study measure?
24 questions were each asked 3 times on ChatGPT, Claude, Gemini and Perplexity, giving 288 answers, plus 61 U.S. Google searches that each showed an AI Overview. All answers were measured on October 28, 2026, in English.
How was this study measured?
24 questions × 3 runs each × 4 AI engines = 288 answers + 61 Google searches · 61 AI Overviews
24 questions, each asked 3 times on each engine, measured October 28, 2026. Market: United States, English. The destination site was not mentioned to any engine. Gemini answers were produced by gemini-3.6-flash; every Gemini answer in this report used that one model. All 72 Gemini answers had search sources. Cities count as named when they appear in the answer text; products and name variants roll up to the tracked name. Longer names take precedence over shorter overlapping ones, and common words count only in context. Unicode apostrophes and dashes are normalized.
| Engine | Model IDs called | Asked | Returned | Grounded |
|---|---|---|---|---|
| ChatGPT | gpt-5-mini | 72 | 72 | 72 |
| Claude | claude-sonnet-5-5 | 72 | 72 | 72 |
| Gemini | gemini-3.6-flash | 72 | 72 | 72 |
| Perplexity | sonar-pro | 72 | 72 | 72 |
| Google AI Overviews | DataForSEO Google SERP (live, advanced) | 61 | 61 shown | 61 |
Own-domain citations count a domain plus all its subdomains, with the same rule in every table.
Definitions
- Appearance rate: answers naming the city ÷ answers returned, with Wilson 95% intervals.
- First mention: average order in which the city first appears (1 = first).
- Own-domain citation: grounded answers citing the city's domains ÷ grounded answers.
- Citation share: citations to the city's domains ÷ all citations.
Limits
- Measured on one day, October 28, 2026. Each engine answered 24 questions 3 times. Answers came from each engine's API (model IDs above); the consumer apps can answer differently.
- Only 25 tracked cities were scored. Untracked names that appeared often were not scored, including Seattle (34), Austin (26), OpenAI (26), Google DeepMind (20).
- AI Overview figures are based on only 61 overviews, so estimates for individual cities are less precise than the four-engine figures.
- 11 Google search could not be fetched by DataForSEO after three tries. It is excluded from the AI Overview figures, so the trigger rate uses 61 searches.
- Reports describe what the engines said, attributed to the engines. They make no judgement about the people or organizations named and are not legal or financial advice.
- Some answers were asked again on October 28, 2026 after a storage error. The earlier texts are kept in the evidence archive and are not used in the counts.
- Intervals treat each answer as independent, so they understate uncertainty for the 24 repeated questions.
- Corrected October 29, 2026: own-domain citation now uses each name's official domains (subdomains included), listed in the evidence file entity_rollups.csv. Names without a resolvable official site count zero.
Which 24 questions did we ask?
Questions marked "Core question" are the original study questions, asked word for word. The others are added neutral questions.
| # | Question | Group | Source |
|---|---|---|---|
| 01 | What are the best cities in the world to start an AI company? | The 5W AI City Index 2026 | Core question |
| 02 | Which cities attract the most AI venture capital? | The 5W AI City Index 2026 | Core question |
| 03 | What are the top cities for AI research? | The 5W AI City Index 2026 | Core question |
| 04 | Where are the major AI labs headquartered? | The 5W AI City Index 2026 | Core question |
| 05 | Which cities have the strongest AI talent pool? | The 5W AI City Index 2026 | Core question |
| 06 | Which governments have the most supportive policies for AI companies? | The 5W AI City Index 2026 | Core question |
| 07 | Where are the largest AI data center hubs? | The 5W AI City Index 2026 | Core question |
| 08 | What are the best cities in Europe for AI startups? | The 5W AI City Index 2026 | Core question |
| 09 | What are the leading AI hubs in Asia? | The 5W AI City Index 2026 | Core question |
| 10 | Which Middle East cities are investing most in AI? | The 5W AI City Index 2026 | Core question |
| 11 | What are the best cities for AI engineers to work in? | The 5W AI City Index 2026 | Core question |
| 12 | Where should a US AI startup open its first international office? | The 5W AI City Index 2026 | Core question |
| 13 | Which US cities have large concentrations of AI startups? | The 5W AI City Index 2026 | Added question |
| 14 | Which cities host major university AI research centers? | The 5W AI City Index 2026 | Added question |
| 15 | Which cities have published AI industry development strategies? | The 5W AI City Index 2026 | Added question |
| 16 | Which cities are named in reports on AI investment in Canada? | The 5W AI City Index 2026 | Added question |
| 17 | Which cities are emerging AI hubs in Latin America? | The 5W AI City Index 2026 | Added question |
| 18 | Which cities are named in reports on AI development in Africa? | The 5W AI City Index 2026 | Added question |
| 19 | Which cities host major AI conferences and research communities? | The 5W AI City Index 2026 | Added question |
| 20 | Which cities have significant AI activity in healthcare and biotechnology? | The 5W AI City Index 2026 | Added question |
| 21 | Which cities have significant AI activity in finance and enterprise software? | The 5W AI City Index 2026 | Added question |
| 22 | Which cities are associated with semiconductor and AI hardware research? | The 5W AI City Index 2026 | Added question |
| 23 | Which sources publish city-level AI employment and investment data? | The 5W AI City Index 2026 | Added question |
| 24 | How do published AI hub rankings define a city or metropolitan area? | The 5W AI City Index 2026 | Added question |
About 5W
5W is an independent PR, creative and AI Visibility agency founded in 2003, and we get you in the chatbox. All of our work is done in house, including media relations, site building and content. Learn more about our GEO practice and our AI PR and digital marketing practice, or write to info@5wpr.com.
More 5W research: 5W Luxury Island Index.
Full report
Read the PDF
Get in touch
Let's build your next chapter.
Tell us what you're working on. A senior strategist will respond within one business day.
- info@5wpr.com
- Phone
- 212.999.5585
- Offices
- Miami · HQ100 SE 2nd Street, Floor 38
Miami, FL 33131Tampa110 South 12th Street
Tampa, FL 33602New York469 7th Avenue, Floor 8
New York, NY 10018




