In a one-day 5W test on November 20, 2026, Microsoft appeared in 49 of 288 answers from ChatGPT, Gemini, Perplexity and Claude. Apple appeared in 46. Yet across the 285 answers with retrieval evidence, neither company's website received a citation. Wikipedia, Kantar and Interbrand were the most-cited domains in the sample.
Updated November 20, 2026
On this page
- What did the 5W test measure?
- Which brands did AI name most often?
- Why were the most-named brands never cited?
- Which sources did the AI engines cite most?
- How did ChatGPT, Gemini and Perplexity differ?
- Which brands led each type of question?
- What should communications teams measure?
- How was the study run?
- Frequently asked questions
What did the 5W test measure?
The 5W test measured which brands four AI assistants named, and which websites they cited, when asked 24 questions about top brands, advertising campaigns, CEOs and brand ownership. The questions ran on ChatGPT, Gemini, Perplexity (sonar-pro) and Claude, three times each, which produced 288 answers.
All 288 attempts returned an answer. Of those, 285 carried retrieval evidence, meaning the engine returned search annotations or grounding records. The 288 answers are the denominator for how often a brand was named. The 285 are the denominator for citations.
This is a point-in-time sample, not a census of category searches. The full prompt list and the answer count behind each question type are in the methodology below.
Which brands did AI name most often?
Microsoft was named most often, in 49 of 288 answers (17%), and Apple was second in 46 (16%). The gap between them is three mentions, and their 95% intervals overlap, so the order describes what was observed and does not show that one brand is more visible than the other.
| Rank | Brand | Named in | Appearance rate | 95% interval |
|---|---|---|---|---|
| 1 | Microsoft | 49/288 | 17% | 13.1% to 21.8% |
| 2 | Apple | 46/288 | 16% | 12.2% to 20.6% |
| 3 | Amazon | 34/288 | 11.8% | 8.6% to 16% |
| 4 | NVIDIA | 33/288 | 11.5% | 8.3% to 15.7% |
| 5 | Coca-Cola | 32/288 | 11.1% | 8% to 15.3% |
| 6 | Dove | 30/288 | 10.4% | 7.4% to 14.5% |
| 7 | Dawn | 21/288 | 7.3% | 4.8% to 10.9% |
| 8 | Samsung | 21/288 | 7.3% | 4.8% to 10.9% |
| 9 | Toyota | 21/288 | 7.3% | 4.8% to 10.9% |
| 10 | Unilever | 20/288 | 6.9% | 4.5% to 10.5% |
The intervals for Amazon, NVIDIA and Coca-Cola also overlap, so their order is descriptive too. Two labels in the raw results, "Sources" and "Category," are extraction artifacts and not brands. They were removed from this table and from every brand claim in this report.
Why were the most-named brands never cited?
Microsoft and Apple each recorded zero owned-domain citations across the 285 answers with retrieval evidence, so the two most-named brands in the test were never cited at their own websites. The same was true of Coca-Cola, Dove and Toyota.
| Brand | Answers citing its own domain | Rate |
|---|---|---|
| Microsoft | 0/285 | 0% |
| Apple | 0/285 | 0% |
| Amazon | 1/285 | 0.4% |
| NVIDIA | 2/285 | 0.7% |
| Coca-Cola | 0/285 | 0% |
| Dove | 0/285 | 0% |
| Samsung | 1/285 | 0.4% |
| Toyota | 0/285 | 0% |
| Unilever | 3/285 | 1.1% |
The one measured domain that was cited often was Kantar, a research firm. Kantar's domain appeared in 40 of the 285 answers (14%).
The study counted citations across answers and did not map them to individual brand mentions. It therefore cannot say which source supported any single mention. What the data does show is that being named and being cited are separate outcomes, and a brand can have the first without the second.
Which sources did the AI engines cite most?
Wikipedia was the most-cited domain, appearing in 43 of 285 answers with retrieval evidence (15.1%), followed by Kantar at 14% and Interbrand at 9.5%. Each publisher is counted once per answer, not once per link.
| Publisher | Answers citing it | Share of answers |
|---|---|---|
| en.wikipedia.org | 43/285 | 15.1% |
| kantar.com | 40/285 | 14% |
| interbrand.com | 27/285 | 9.5% |
| globalbrandsmagazine.com | 26/285 | 9.1% |
| chainstoreage.com | 26/285 | 9.1% |
| morningconsult.com | 24/285 | 8.4% |
| yougov.com | 21/285 | 7.4% |
| thedrum.com | 21/285 | 7.4% |
| qz.com | 21/285 | 7.4% |
| brandirectory.com | 21/285 | 7.4% |
| prnewswire.com | 21/285 | 7.4% |
| rankingthebrands.com | 19/285 | 6.7% |
The list mixes an encyclopedia, brand-ranking and survey firms, trade publications and a newswire. These rates show which domains were cited across the whole prompt set. They do not show which source supported each brand mention.
How did ChatGPT, Gemini and Perplexity differ?
ChatGPT, Gemini and Perplexity named the leading brands at different rates. Apple's highest rate was on ChatGPT at 26.4%, while Dove and Toyota did best on Perplexity at 22.2% and 16.7%. Each rate uses the 72 answers returned by that engine.
| Brand | Perplexity | ChatGPT | Gemini |
|---|---|---|---|
| Microsoft | 20.8% | 23.6% | 23.6% |
| Apple | 18.1% | 26.4% | 19.4% |
| Amazon | 13.9% | 19.4% | 13.9% |
| NVIDIA | 12.5% | 15.3% | 15.3% |
| Coca-Cola | 13.9% | 15.3% | 15.3% |
| Dove | 22.2% | 6.9% | 12.5% |
| Toyota | 16.7% | 4.2% | 8.3% |
Claude's 72 answers are left out of this comparison while 5W checks the raw outputs against the extraction results. They remain in the pooled totals above. Engine results are also not strictly comparable, because each engine uses a different model, retrieval tool and answer format.
Gemini accounted for 22 of the 23 recorded "Sources" entries. That is why unexpected entries were reviewed by hand before this report named any brand.
Which brands led each type of question?
Different brands led different question types, and the leaders for campaign and CEO questions describe companies, not individual campaigns or executives. Each rate uses the answers for that question type.
| Question type | Answers | Leading brand | Rate |
|---|---|---|---|
| Most trusted brand | 24 | BAND-AID | 70.8% |
| Best campaign | 24 | Coca-Cola | 45.8% |
| Best CEO | 24 | NVIDIA | 54.2% |
| Best retail brand | 12 | Costco | 66.7% |
| Comparison and evaluation | 12 | Apple | 66.7% |
| Gifting and occasions | 12 | CHANEL | 75% |
| Ownership and maker attribution | 12 | Unilever | 50% |
Coca-Cola led the campaign prompts at 45.8%, followed by Nike at 37.5%. NVIDIA led the CEO prompts at 54.2%, ahead of Microsoft at 41.7%. Those figures count appearances of the companies. Whether AI answers agreed with Cannes Lions or Effie juries would need a comparison of named campaigns and award records for the same period.
For more 5W research on how AI engines name brands, see the State of AI Search 2026 study, the DTC Brands Index 2.0 and the Hotel and Cruise CEOs Index.
What should communications teams measure?
Communications teams should measure three things separately: whether AI names the brand, which pages the answers cite, and whether those pages support what the answer says. The test showed the first two can diverge completely.
- Run the same category questions on ChatGPT, Gemini and Perplexity each month and record whether each answer names the brand and in what position.
- Log every domain each answer cites, so the team can see whether Wikipedia, a research firm such as Kantar or a trade publication is supplying the evidence.
- Open the cited pages and check whether they state the facts the answer gives about the brand. A cited page that is out of date or wrong is a correction to request.
- Track campaign titles and executive names separately from company names, with a different question for each goal such as trust, campaign recognition and leadership.
- Review unexpected entries by hand before publishing a ranking.
Frequently asked questions
Which brands did AI name most often in the 5W test?
Microsoft was named in 49 of 288 answers (17%) and Apple in 46 (16%). Their 95% intervals overlap, so the three-mention gap does not establish that one was more visible.
Were Microsoft's and Apple's websites cited?
No. Across the 285 answers with retrieval evidence, neither Microsoft's nor Apple's website received a citation.
Which domains did AI engines cite most?
Wikipedia was cited in 15.1% of answers with retrieval evidence, Kantar in 14% and Interbrand in 9.5%.
Did the AI answers match Cannes Lions or Effie winners?
The test cannot say. Coca-Cola led the campaign prompts at 45.8%, but that figure counts company mentions, and a comparison with award winners would need named campaigns.
Can 5W measure this for a brand?
Yes. 5W tracks which brands ChatGPT, Gemini and Perplexity name, which pages they cite, and whether those pages support what the answer says. 5W is a PR, creative and AI Visibility agency headquartered in Miami, with offices in New York and Tampa.
How was the study run?
5W ran 24 stratified prompts on four engines with three trials per prompt, and fixed the prompt allocation and entity registry before scoring. Product lines rolled up to parent brands where the registry established the relationship. Appearance rates use 95% Wilson intervals.
- Category: brands, campaigns and CEOs
- Measurement date: November 20, 2026
- Answers attempted: 288. Answers returned: 288. Answers with retrieval evidence: 285 (99%). Answers that failed outright: 0
- Engines: Perplexity (sonar-pro), ChatGPT (openai/gpt-5.6-terra), Gemini (google/gemini-3.8-flash) and Claude (anthropic/claude-sonnet-4), each with its own retrieval tool
- Analysis version: study-methodology-v2. Entity registry: entity-registry-2026-08-27
How many answers sit behind each question type?
Each question type has a fixed number of prompts, and the answer count is the prompts multiplied by four engines and three trials.
| Question type | Prompts | Answers |
|---|---|---|
| Best campaign | 2 | 24 |
| Most trusted brand | 2 | 24 |
| Best CEO | 2 | 24 |
| Biggest comeback | 1 | 12 |
| Best retail brand | 1 | 12 |
| Best travel brand | 1 | 12 |
| General best-in-category | 1 | 12 |
| Everyday and mainstream choice | 1 | 12 |
| Price and value | 1 | 12 |
| Primary use case | 1 | 12 |
| Beginners and first purchase | 1 | 12 |
| Where to buy | 1 | 12 |
| Gifting and occasions | 1 | 12 |
| Specific need or condition | 1 | 12 |
| Comparison and evaluation | 1 | 12 |
| Trust and recommendation | 1 | 12 |
| Ownership and maker attribution | 1 | 12 |
| Quality and premium tier | 1 | 12 |
| Durability and longevity | 1 | 12 |
| Specialist and advanced use | 1 | 12 |
| Emerging and new entrants | 1 | 12 |
What were the exact prompts?
The 24 prompts below were sent word for word to every engine.
| # | Prompt | Question type |
|---|---|---|
| 1 | best ad campaigns | Best campaign |
| 2 | What were the most awarded marketing and advertising campaigns recognized by major industry festivals recently? | Best campaign |
| 3 | most trusted brands | Most trusted brand |
| 4 | Which consumer brands currently rank highest for customer trust and corporate reputation in national surveys? | Most trusted brand |
| 5 | top rated ceos | Best CEO |
| 6 | Which corporate chief executives are ranked highest by business publications for leadership and financial performance? | Best CEO |
| 7 | Which consumer brands had the biggest turnaround comeback? | Biggest comeback |
| 8 | What are the top rated retail brands today? | Best retail brand |
| 9 | best travel and hospitality brands | Best travel brand |
| 10 | top global brands | General best-in-category |
| 11 | most popular everyday household consumer brands | Everyday and mainstream choice |
| 12 | Which consumer brands offer the best value for money compared to premium category leaders? | Price and value |
| 13 | Which corporate brands are best for enterprise software and workplace productivity solutions? | Primary use case |
| 14 | What are the most reliable starter brands for someone buying home essentials for the first time? | Beginners and first purchase |
| 15 | best places to buy certified pre-owned luxury brands | Where to buy |
| 16 | most popular luxury lifestyle brands for anniversary gifts | Gifting and occasions |
| 17 | Which major consumer brands have the highest verified ratings for environmental sustainability and ethical sourcing? | Specific need or condition |
| 18 | brand value rankings comparison | Comparison and evaluation |
| 19 | Which brand leaders do market analysts recommend most? | Trust and recommendation |
| 20 | Which parent conglomerates own the largest portfolio of leading consumer lifestyle and beauty brands? | Ownership and maker attribution |
| 21 | highest quality luxury brands | Quality and premium tier |
| 22 | Which heritage apparel and footwear brands are known for products that last a lifetime? | Durability and longevity |
| 23 | top agency creative directors | Specialist and advanced use |
| 24 | fastest growing direct to consumer brands | Emerging and new entrants |
What are the limits of these findings?
The findings describe these 24 prompts on one day, and they should not be read as representative of all searches in the category.
- Engine behavior changes without notice, and each engine uses a different model, retrieval tool and answer format.
- Where intervals overlap, differences between brands cannot be distinguished at this sample size.
- A brand that is not listed was not measured as absent. It was not named in these answers.
- Appearance is a visibility measure. It is not an endorsement, a quality judgment or a market-share estimate.
- Citation counts are answer-level and were not mapped to individual brand mentions.
- Claude's answers are held out of the engine comparison pending a check of the raw outputs against the extraction results.
- Forty further names picked up in the answers could not be confidently resolved to a brand and carry no metrics here.




