Six findings
- The problem has a name, and it is searched everywhere. “shadow ai” has 10,750 searches a month. It grew in all 25 countries where it is measured, with a median change of +150% in a year. Italy is up 519% and Portugal 425%.
- Almost nobody asks permission. Seven questions such as “can i use chatgpt at work” and “is chatgpt allowed at work” add up to 140 searches a month across the 27 countries.
- Worry about ChatGPT and privacy is fading. “chatgpt privacy” fell in 12 of the 14 countries where it is measured. “is chatgpt confidential”, “does chatgpt save your data” and the German “chatgpt datenschutz” fell in every one.
- The safety question moves to the newer tool. “is claude safe” has 860 searches a month and grew in 7 of 8 countries, by 878% in the United States and 750% in the United Kingdom.
- Policies are searched mostly in the United States. “ai policy” has 4,080 searches a month, 2,400 of them in the United States. In Germany “ki richtlinie” has 590, up 50%, and in France “charte ia” has 590, up 22%.
- Companies search more than employees. Terms on the company side add up to 19,680 searches a month, against 7,830 on the employee side.
A problem with a name
Shadow AI is the use of AI tools at work without the company knowing or approving. A year ago few people searched for the term. Now it is growing in every country where it can be measured.
Figure 1
“shadow ai” grew in every country where it is measured
Change in a year, October 2026 against October 2025, in the 25 countries where the term has at least 20 searches a month.
“shadow ai”
Median of the 25 countries
Source: Solorai analysis of Google Ads Keyword Planner data, November 2025 to October 2026.
Who is searching
We sorted the 70 terms into two sides. The employee side asks whether a tool is allowed, private or safe. The company side asks for policies, rules and ways to control use. Employees could have searched for permission. They did not: the terms that draw searches on their side are about privacy, not about being allowed.
Figure 2
Companies search for the problem, employees search about privacy
Average monthly searches, total of the 27 countries, for the terms with at least 300 searches a month.
Company side
Employee side
| “shadow ai” | 10,750 |
|---|---|
| “ai policy” | 4,080 |
| “is chatgpt private” | 2,140 |
| “chatgpt privacy” | 1,650 |
| “chatgpt datenschutz” | 1,180 |
| “is claude safe” | 860 |
| “is chatgpt confidential” | 800 |
| “ki richtlinie” | 690 |
| “schatten ki” | 670 |
| “charte ia” | 630 |
| “does chatgpt save your data” | 560 |
| “ai policy template” | 430 |
| “ai usage policy” | 410 |
| “is copilot safe” | 370 |
| “política de ia” | 320 |
| “ai acceptable use policy” | 300 |
Source: Solorai analysis of Google Ads Keyword Planner data, November 2025 to October 2026.
Figure 3
Worry about ChatGPT and privacy is falling while shadow AI rises
Number of countries where each term rose or fell in a year, among those where it has at least 20 searches a month.
Countries up
Countries down
Source: Solorai analysis of Google Ads Keyword Planner data, November 2025 to October 2026.
Data table: terms with at least 50 searches a month
| Search term | Side | Monthly searches | Countries measured | Countries up | Countries down |
|---|---|---|---|---|---|
| “shadow ai” | Company | 10,750 | 25 | 25 | 0 |
| “ai policy” | Company | 4,080 | 19 | 10 | 4 |
| “is chatgpt private” | Employee | 2,140 | 4 | 1 | 3 |
| “chatgpt privacy” | Employee | 1,650 | 14 | 1 | 12 |
| “chatgpt datenschutz” | Employee | 1,180 | 4 | 0 | 4 |
| “is claude safe” | Employee | 860 | 8 | 7 | 0 |
| “is chatgpt confidential” | Employee | 800 | 2 | 0 | 2 |
| “ki richtlinie” | Company | 690 | 3 | 2 | 1 |
| “schatten ki” | Company | 670 | 3 | 3 | 0 |
| “charte ia” | Company | 630 | 2 | 2 | 0 |
| “does chatgpt save your data” | Employee | 560 | 3 | 0 | 3 |
| “ai policy template” | Company | 430 | 2 | 0 | 1 |
| “ai usage policy” | Company | 410 | 2 | 1 | 1 |
| “is copilot safe” | Employee | 370 | 2 | 0 | 1 |
| “política de ia” | Company | 320 | 1 | 1 | 0 |
| “ai acceptable use policy” | Company | 300 | 2 | 2 | 0 |
| “ai governance policy” | Company | 250 | 2 | 1 | 0 |
| “ai beleid” | Company | 200 | 2 | 1 | 1 |
| “generative ai policy” | Company | 130 | 2 | 1 | 1 |
| “chatgpt policy” | Company | 110 | 2 | 0 | 2 |
| “block chatgpt” | Company | 110 | 2 | 0 | 2 |
| “ki richtlinie unternehmen” | Company | 110 | 1 | 0 | 1 |
| “charte ia entreprise” | Company | 110 | 1 | 0 | 0 |
| “chatgpt confidentialité” | Employee | 90 | 1 | 0 | 1 |
| “company ai policy” | Company | 90 | 2 | 1 | 1 |
| “use chatgpt for work” | Employee | 70 | 2 | 0 | 2 |
| “polityka ai” | Company | 70 | 1 | 1 | 0 |
| “chatgpt at work” | Employee | 50 | 1 | 1 | 0 |
| “política de uso de ia” | Company | 50 | 1 | 0 | 1 |
Country by country
Three terms in each country: monthly searches and the change in a year.
Data table: 27 countries
| Country | “shadow ai” | In a year | “ai policy” | In a year | “chatgpt privacy” | In a year |
|---|---|---|---|---|---|---|
| United States | 3,600 | +128% | 2,400 | +83% | 880 | -32% |
| United Kingdom | 880 | +81% | 720 | 0% | 210 | -46% |
| Ireland | 90 | +100% | 50 | -29% | 20 | -67% |
| Italy | 720 | +519% | 90 | 0% | 140 | -65% |
| Spain | 320 | +243% | 30 | +50% | 20 | -67% |
| France | 1,300 | +69% | 50 | +25% | 70 | -75% |
| Portugal | 90 | +425% | - | - | - | - |
| Germany | 480 | +84% | 110 | +27% | 50 | -67% |
| Austria | 70 | +267% | 20 | +200% | - | - |
| Switzerland | 110 | +55% | 40 | 0% | 20 | 0% |
| Netherlands | 320 | +88% | 90 | +29% | 110 | -76% |
| Belgium | 170 | +50% | 90 | +22% | 30 | -67% |
| Luxembourg | - | - | - | - | - | - |
| Poland | 720 | +156% | 40 | +50% | 20 | -67% |
| Sweden | 140 | +50% | 210 | -19% | 30 | +150% |
| Denmark | 70 | +175% | 20 | +100% | - | - |
| Norway | 50 | +80% | 20 | +100% | - | - |
| Finland | 50 | +150% | 20 | 0% | 20 | -33% |
| United Arab Emirates | 70 | +175% | 40 | -43% | - | - |
| Saudi Arabia | 70 | +25% | 20 | -50% | - | - |
| Qatar | - | - | - | - | - | - |
| Egypt | 50 | +250% | - | - | - | - |
| Mexico | 210 | +271% | - | - | - | - |
| Brazil | 880 | +233% | 20 | 0% | 30 | -25% |
| Argentina | 110 | +143% | - | - | - | - |
| Colombia | 110 | +180% | - | - | - | - |
| Chile | 70 | +250% | - | - | - | - |
What it means for a company
- Do not wait to be asked. Employees are not searching for permission, so the absence of questions says nothing about the absence of use. Start from an inventory of the tools in use. Use the EU AI Act checklist.
- A policy works when there is an approved tool behind it. People stop using personal accounts when the company gives them something better. See how to move teams to Claude.
- Classify the uses you find. Some of what you discover will fall under the AI Act. Check the risk level of one AI use.
Method and limits
- Source: Google Ads Keyword Planner, measured on 14 November 2026.
- 70 search terms in 8 languages, measured country by country in 27 countries: 38 on the employee side and 32 on the company side. The split is ours.
- Volumes are average monthly searches for November 2025 to October 2026. Keyword Planner rounds them.
- The change in a year compares October 2026 with October 2025.
- A country is counted for a term only where the term has at least 20 searches a month.
- Keyword Planner reports 10 as its lowest value. We treat it as not measured.
- “shadow ai” is also searched by vendors, journalists and students, not only by companies with the problem.
- Searches show interest, not behaviour. Few searches about permission do not prove that employees use AI without it.
Cite this research
Solorai (2026). Shadow AI: nobody asks permission. Search data from 27 countries. https://solorai.com/insights/research/shadow-ai-search-data-2026/
Free to quote and to chart, with attribution and a link to this page.