Ask ChatGPT or Google? A Side-by-Side Test for Everyday Searches
On 15 November 2026 we ran the same 10 everyday prompts in ChatGPT and Google. Google scored 192 out of 200; ChatGPT scored 185. The gap was almost entirely speed.

Introduction
In a Big Technology interview (3 Nov 2026), ChatGPT consumer lead Adam Fry said ChatGPT’s volume of search-like queries is now roughly one-fifth (~20%) of Google’s. The Verge notes Fry did not specify the Google volume source — and that one-fifth is still far from overtaking Google. That claim is useful context. It is not a result of this test.
On 15 November 2026, from 07:36 to 07:47 UTC, we ran the same 10 prompts, with the same wording, in ChatGPT and in Google. Google had the higher grand total, 192 out of 200, against ChatGPT’s 185. The seven-point gap is almost entirely speed. Accuracy was close, and ChatGPT’s sources were slightly stronger. These numbers describe one run on one day. They are not a general ranking of the two products.
The short version
- Google 192, ChatGPT 185. Google won seven prompts, ChatGPT won one (turning on 2-Step Verification), and two were ties.
- The gap is almost all speed: Google 50/50, ChatGPT 39/50. Every Google run scored 5 — a usable first answer in eight seconds or less, using the AI Overview time when an Overview answered the question. ChatGPT was usually 9–16 seconds.
- Accuracy was close: ChatGPT 47/50, Google 46/50. ChatGPT’s sources were 50/50; Google’s were 49/50.
- Both missed iPhone 16 storage against Apple’s official Capacity row on the test day, which listed only 128GB.
- One same-day run, first turns only. ChatGPT was a Plus Temporary chat with web search on and no model name shown. Google was signed in, in English, with a UK region setting, a saved home location outside the UK, and personalized results.
1. What OpenAI Actually Claimed

| Claim | Source | Caveat |
|---|---|---|
| ~20% of Google search volume | Fry via Big Technology / The Verge | Google baseline source unspecified |
| What counts as a search | Web-search-triggering queries | Not every chat turn |
| 1.2B weekly users / 1.5B images a week / 250M weekly voice users | Big Technology | Company-reported |
Fry described a habit shift toward personalization and journey context — for example a trip — versus one-off question boxes. Weekly users moved from about 1.0 billion to 1.2 billion in roughly two months. The Verge compares the 20% figure to older remarks about Bing’s small share — useful distance, not a claim that ChatGPT has already caught Google. Those figures did not decide any row in the scorecard below.
2. How We Ran This Test
We used the same 10 prompts, pasted exactly, across four everyday categories:
- News/facts. Who won the FIFA World Cup 2026 final, and on what date, with the score; who won the 2026 Nobel Prize in Physics, with the official citation; which city and country will host the 2028 Summer Olympics, with the official dates.
- How-to. Capture a portion of the screen on a Mac; turn on 2-Step Verification on a Google Account, with the official steps; export Chrome bookmarks on a computer, with the official steps.
- Shopping. Official Sony weight and maximum battery life for the WH-1000XM5 and WH-1000XM6; official Apple display size, weight and storage options for iPhone 16 and iPhone 17, not the Pro models.
- Local. Which London Underground line serves Heathrow Airport Terminal 5, and whether the Elizabeth line stops there; whether British Museum permanent-collection entry is free, and the official address of the main entrance.
ChatGPT was a Plus account in a Temporary chat, with web search turned on from the + menu. No model name was shown. Temporary chats can still use memory and custom instructions.
Google was signed in, with language English and region United Kingdom. The results footer showed a saved home location outside the UK and that results were personalized. That may have affected local and shopping results. Personalization was not scored as a fifth column, because the two tools did not have comparable signed-in context.
Each answer was scored 1–5 for accuracy, source usefulness, speed and usability, for a prompt total out of 20 and a grand total out of 200. Speed is seconds from submitting the query to a usable first answer. For Google, that is the AI Overview when it answered the question. The speed bands (5 = eight seconds or less, and so on) are this test’s convention. ChatGPT’s exact finish time on the World Cup prompt was not captured — it was at least 12.9 seconds — and was scored conservatively. We scored first turns only, against official fact keys checked on the test day.
3. The Scorecard

| Prompt | Category | ChatGPT /20 | Google /20 | Winner |
|---|---|---|---|---|
| P1. 2026 World Cup final | News/facts | 17 | 18 | |
| P2. 2026 Nobel Prize in Physics | News/facts | 19 | 20 | |
| P3. 2028 Summer Olympics host and dates | News/facts | 20 | 20 | Tie |
| P4. Mac, capture a portion of the screen | How-to | 19 | 20 | |
| P5. Turn on Google 2-Step Verification | How-to | 19 | 17 | ChatGPT |
| P6. Export Chrome bookmarks | How-to | 19 | 20 | |
| P7. Sony WH-1000XM5 vs WH-1000XM6 official specs | Shopping | 18 | 20 | |
| P8. iPhone 16 vs iPhone 17 official display, weight, storage | Shopping | 17 | 17 | Tie |
| P9. Heathrow Terminal 5 Tube and Elizabeth line | Local | 19 | 20 | |
| P10. British Museum free entry and main entrance | Local | 18 | 20 | |
| Grand total /200 | 185 | 192 |

| Tool | Accuracy /50 | Source usefulness /50 | Speed /50 | Usability /50 | Mean |
|---|---|---|---|---|---|
| ChatGPT | 47 | 50 | 39 | 49 | 4.63 |
| 46 | 49 | 50 | 47 | 4.80 |
| Category | ChatGPT | |
|---|---|---|
| News/facts (P1–P3) | 4.67 | 4.83 |
| How-to (P4–P6) | 4.75 | 4.75 |
| Shopping (P7–P8) | 4.38 | 4.63 |
| Local (P9–P10) | 4.63 | 5.00 |
Google’s mean on the two local prompts was 5.00, and its news/facts mean was slightly ahead. How-to was a draw at 4.75. Shopping was the weakest category for both, because of the shared iPhone 16 storage miss. Do not read the local scores as a Maps win: no Maps pack appeared on the Heathrow or British Museum prompts in this run. These category means come from one same-day run, so they are not a general rule for when to ask which tool.
4. Where the Answers Actually Differed
The only ChatGPT win: official 2-Step Verification labels
ChatGPT matched the current official Google Account Help labels: Security & sign-in, then Turn on 2-Step Verification. Google’s AI Overview used older labels — Security, Get started, and Turn on. That is why ChatGPT won the prompt, 19 to 17, even though Google was faster.


A shared miss: iPhone 16 storage
On the shopping compare, display size and weight matched Apple, and iPhone 17 storage — 256GB and 512GB — matched too. For iPhone 16, both tools listed 128GB, 256GB and 512GB. Apple’s official technical specifications page on the test day listed only 128GB in the Capacity row. That page may be a sold-down listing. We are not going further than that. Both scored 17 out of 20, a tie.


Local, without a Maps pack
Both correctly said British Museum permanent-collection entry is free and gave the official Great Russell Street address. ChatGPT showed an empty map widget under the address. Google had no Maps pack. Google won the prompt 20 to 18, on speed and a slightly cleaner first screen, not because of Maps.


On the World Cup prompt, Google had no AI Overview, so the correct answer — Spain 1–0 after extra time on 19 July 2026 — had to be read from organic snippets. ChatGPT presented a structured scoreboard. That is a usability and speed trade-off, not an accuracy failure on Google’s side. Google still won the prompt, 18 to 17, in part because ChatGPT’s finish time was scored conservatively.
Both tools also correctly answered the 2026 Physics Nobel, the Los Angeles 2028 host and dates, the Mac screenshot steps, Chrome bookmark export, the official Sony headphone weights and 30-hour / 40-hour battery maxima, and Heathrow Terminal 5 (Piccadilly and Elizabeth line).
5. How to Read This
This is one same-day run (15 Nov 2026, 07:36–07:47 UTC), ten prompts, first turns only. It does not support a general claim that ChatGPT is slower, or that Google is more accurate, beyond this session.
Fry’s personalization point still matters for later tests. A signed-in ChatGPT journey, such as trip planning across turns, is a different animal from a one-shot Google query. Score personalization only when both sides have comparable context. We did not score it here.
Use OpenAI’s growth stats as context, not as proof that ChatGPT wins every category — or that Google always will.
Common Misconceptions
Do not rewrite the OpenAI claim as “ChatGPT has 20% search market share.” Do not treat every ChatGPT turn as a search. Do not treat this bake-off as a general ranking of the two products. Do not invent a Maps win or loss from this run. One-fifth of Google’s volume is large and still not “overtaking.”
Conclusion
On 15 November 2026, in one 11-minute same-day run, Google was ahead mainly because it was faster. Accuracy was close; ChatGPT’s sources were a touch stronger; both tripped on the same official iPhone 16 catalog row. OpenAI’s 20% figure remains useful context, carefully defined, and incomplete without Google’s baseline method. If you change a team default, run the same 10 prompts yourself on the day you decide — these scores will not travel unchanged.
Resources
- Big Technology: OpenAI on ChatGPT search volume (3 Nov 2026)
- The Verge: OpenAI says ChatGPT is catching up to Google in search (3 Nov 2026)
- Google Account Help: Turn on 2-Step Verification (official steps used to score prompt 5)
- Apple: iPhone 16 technical specifications (Capacity row as fetched on the test day)
- Apple: iPhone 17 technical specifications
About to crown a default search tool from one headline — or one run? Share the scorecard first.


