Across 9 studies (N = 6,282), identical AI-written empathic replies were rated more empathic and supportive when labelled “human” than when labelled “AI”. People who suspected AI had helped write the “human” replies rated them lower too.
Module: psyche · grade A · 2025-11-01. Rubin, Li, Zimmerman, Ong, Goldenberg & Perry (2025). Comparing the value of perceived human versus AI-generated empathy. Nature Human Behaviour 9(11), 2345–2359
In 13 experiments, people who disclosed using AI were trusted less than people who did not. Being exposed by a third party cost even more trust than disclosing.
Module: psyche · grade A · 2025-05-01. Schilke & Reimann (2025). The transparency dilemma: How AI disclosure erodes trust. Organizational Behavior and Human Decision Processes 188, 104405
In two preregistered experiments (N = 4,976, US and UK), labelling headlines “AI-generated” lowered perceived accuracy and willingness to share, even when the headlines were true or written by humans. The AI label cost 2.66 points, about a third of a “False” label (9.33 points).
Module: psyche · grade A · 2024-10-01. Altay & Gilardi (2024). People are skeptical of headlines labeled as AI-generated, even if true or human-made. PNAS Nexus 3(10), pgae403
In US retailer transaction data, adopting buy-now-pay-later raised purchase incidence by about 9% and amount by about 10%, more for small-basket shoppers and heavy credit-card users. Three preregistered experiments back the mechanism.
Module: psyche · grade A · 2025-05-01. Maesen & Ang (2025). Buy now pay later: Impact of installment payments on customer purchases. Journal of Marketing 89(3), 13–35
A meta-analysis of 71 studies (N = 98,299) links heavier short-form video use with worse attention (r = −.38), worse inhibitory control (r = −.41) and worse mental health (r = −.21). Measures of addictive use correlate more strongly than time spent.
Module: psyche · grade A · 2025-09-01. Nguyen et al. (2025). Feeds, feelings, and focus: A systematic review and meta-analysis of short-form video use. Psychological Bulletin 151(9), 1125–1146
In a preregistered experiment with 900 people, GPT-4 given basic demographic data on its opponent raised the odds of the opponent moving toward its position by 81.2% versus human–human debates (95% CI +26.0% to +160.7%), and out-persuaded humans in 64.4% of non-tied pairs.
Module: psyche · grade A · 2025-08-01. Salvi, Horta Ribeiro, Gallotti & West (2025). On the conversational persuasiveness of GPT-4. Nature Human Behaviour 9(8), 1645–1653
Across 202,898 adults in 22 countries, mean flourishing is flat from age 18 to 49 (7.03) and rises only later (7.36 at 80+), unlike the U-shape seen in life-satisfaction studies. Without financial security in the index, Indonesia (8.47), Mexico (8.19) and the Philippines (8.11) rank highest; Japan (5.93), Turkey (6.59) and the UK (6.88) rank lowest. Meaning shows a negative relationship with GDP per capita.
Module: culture · grade A · 2025-04-30. VanderWeele, Johnson et al. (2025). The Global Flourishing Study: Study Profile and Initial Results on Flourishing. Nature Mental Health 3, 636–653
Among 466,089 Europeans aged 20–29 in 32 countries (Eurobarometer, 1990–2023), 14 countries, Romania, Poland and Hungary among them, show no ideological gender gap, 7 a stable one and 11 a widening one; young men moved right in only 6 of those 11. In the US, Gallup finds liberal identification among women aged 18–29 rose from 28% (2001–07) to 40% (2017–24), while young men stayed near 25%.
Module: culture · grade B · 2025-06-13. Nennstiel & Hudde (2025). Is there a growing gender divide among young adults in regard to ideological left–right self-placement? Evidence from 32 European countries. European Sociological Review 41(6), 862–883
Pew’s 2023–24 Religious Landscape Study (36,908 US adults) puts Christian identity at 62%, down from 78% in 2007 but steady at 60–64% since 2019; the unaffiliated have plateaued at 29%, and 86% believe in a soul, including about 82% of 18–24-year-olds. In Britain, YouGov withdrew the data behind the Bible Society’s “Quiet Revival” (monthly attendance 8% to 12%) after finding fraudulent respondents concentrated in the groups the survey emphasised (Research Live, April 2026).
Module: culture · grade A · 2025-02-26. Pew Research Center (2025). Decline of Christianity in the U.S. Has Slowed, May Have Leveled Off. 2023–24 Religious Landscape Study
Gallup finds 54% of US adults drink, a record low (62% in 2023; peak 71% in the late 1970s), and 53% now say moderate drinking is bad for health, up from 28% in 2018; among 18–34s, 50% drink and 66% call moderate drinking unhealthy. ESPAD 2024 (113,882 students aged 15–16 in 37 countries) shows past-month alcohol use at 42%, down from 55% in 1995.
Module: culture · grade B · 2025-08-13. Saad, L. (2025). U.S. Drinking Rate at New Low as Alcohol Concerns Surge. Gallup
Among 236,270 American Time Use Survey respondents (2003–2023), the share reading for pleasure on a given day fell from 28% in 2004 to 16% in 2023, about 3% a year. Readers themselves read longer, from about 1h23 to about 1h37 a day, and only 2% read with children on the diary day, unchanged over 20 years.
Module: culture · grade B · 2025-08-20. Bone, Bu, Sonke & Fancourt (2025). The decline in reading for pleasure over 20 years of the American Time Use Survey. iScience 28(9), 113288
US vinyl revenue reached $1,042.9M in 2025 (+9.3%) on 46.8M units (+7.9%), about 75% of physical revenue, while CD revenue fell 7.8%. Streaming still made up about 82% of a record $11.5B total. UK vinyl grew for an 18th consecutive year (BPI, reported by RouteNote).
Module: culture · grade C · 2026-03-16. RIAA (2026). 2025 Year-End Music Industry Revenue Report. Recording Industry Association of America
US toy sales rose 17% in the first half of 2026; sales to adults 18+ grew 25%, the largest contributor, and adult-only households accounted for 55% of toy sales (+16%). Teens aged 12–17 were the fastest-growing group (+33%), and teens plus adults generated nearly 60% of the incremental dollars.
Module: culture · grade D · 2026-08-12. Circana (2026). U.S. Toy Industry Posts Strongest First Half in Six Years. Circana
In a UNFPA–YouGov survey of about 14,000 adults in 14 countries, nearly 20% of reproductive-age adults believe they cannot have the number of children they want. 39% say financial limits have affected or would affect their family size, and nearly 1 in 5 cite fears about the future such as climate change, war or pandemics.
Module: culture · grade D · 2025-06-10. UNFPA (2025). State of World Population 2025: The Real Fertility Crisis. United Nations Population Fund
In 2025, 50.6% of EU women aged 25–34 held a tertiary degree versus 39.3% of men (EU total 44.8%), and the gap widened from 10.6 points in 2015 to 11.3 points. Romania has the lowest tertiary share among 25–34-year-olds, below 25%.
Module: culture · grade C · 2026-04. Eurostat (2026). Educational attainment statistics. Statistics Explained
Global employee engagement fell to 20% in 2025 (21% in 2024, peak 23%), and manager engagement dropped from 27% to 22%, in a 2025 sample of 263,810. Europe is lowest: 12% engaged, 15% actively disengaged, 39% stressed daily. Under-35s are at 19% versus 21% for those 35+, and 22% of employees felt a lot of loneliness the previous day.
Module: culture · grade A · 2026-04. Gallup (2026). State of the Global Workplace 2026. Gallup
Only 31.84% of Romanians aged 16–74 had basic or above basic digital skills in 2025, the EU's lowest (EU 60.40%, BG 38.26%, PL 50.42%, HU 57.32%, CZ 70.45%, NL 83.61%), while 88.18% used the internet daily (EU 89.54%). Inside Romania the share falls from 44.33% in cities to 30.30% in towns and 22.12% in rural areas. By age it runs from 53.28% at 16–24 (EU 74.55%) to 17.95% at 55–64 and 7.26% at 65–74.
Module: nations · grade C · 2026-04-17. Eurostat (2026). Individuals' level of digital skills (from 2021 onwards), isoc_sk_dskl_i21; daily use from isoc_ci_ifp_fu. Eurostat
In 2025, 60.28% of Romanians bought something online in the previous 12 months, against 73.56% in the EU (BG 51.20%, IT 56.22%, PT 60.50%, PL 69.73%, CZ 83.67%, IE 95.05%). Within Romania the share is 69.70% in cities and 51.65% in rural areas, 80.11% at ages 25–34 and 23.58% at 65–74. It is 36.89% among people with low education and 87.49% among the highly educated.
Module: nations · grade C · 2026-04-17. Eurostat (2026). Internet purchases by individuals (2020 onwards), isoc_ec_ib20. Eurostat
In 2025, 19.2% of Romanians aged 15–29 were neither in employment nor in education or training, the EU's highest (EU 11.0%, BG 13.8%, GR 13.6%, IT 13.3%, PL 9.2%, NL 5.3%). Most are outside the labour force (13.7%), and 12.1% say they do not want to work, against 4.1% in the EU. The rate is 25.0% for young Romanian women and 13.8% for men (EU 12.1% and 10.0%), barely changed from 20.3% overall in 2021.
Module: nations · grade C · 2026-09-10. Eurostat (2026). Young persons neither in employment nor in education and training by labour status (NEET rates), edat_lfse_20. Eurostat
In 2025, 27.4% of Romanians were at risk of poverty or social exclusion (EU 20.9%, BG 29.0%, GR 27.5%, PL 15.0%, CZ 11.5%). The rate is 13.7% in Romanian cities, below the EU average, 24.3% in towns and suburbs and 39.8% in rural areas. Only Bulgaria's rural rate is higher (40.8%), against an EU rural average of 21.4% and 20.4% in rural Poland.
Module: nations · grade C · 2026-06-08. Eurostat (2026). Persons at risk of poverty or social exclusion by age and sex (ilc_peps01n) and by degree of urbanisation (ilc_peps13n). Eurostat
On 1 January 2025, Romanian citizens resident abroad numbered 1,053,042 in Italy, 771,276 in Germany, 609,270 in Spain, 158,209 in France (provisional), 155,715 in Austria, 130,007 in Belgium, 56,708 in the Netherlands and 47,979 in Ireland. Romania itself counted 18,674,247 resident Romanian citizens (estimated).
Module: nations · grade C · 2026-08-10. Eurostat (2026). Population on 1 January by age group, sex and citizenship, migr_pop1ctz. Eurostat
In June 2025, 55% of Romanians said they had encountered disinformation „very often” or „often” in the past seven days (23% + 32%), against 35% in the EU27 (12% + 23%), with samples of 1,040 in Romania and 26,121 in the EU. Among people who use social media or video for news, 47% of Romanians follow news influencers or creators versus 37% in the EU, rising to 87% at ages 15–24 (EU 74%) and 26% at 55+ (EU 14%).
Module: nations · grade A · 2025-10. European Commission (2025). Flash Eurobarometer FL014EP, Social Media Survey 2025: Information Habits and Sources, Romania factsheet. European Commission
Trust in news overall in 2026 is 23% in Romania (−3, a new low for the fifth year running), 21% in Bulgaria, 17% in Hungary and 39% in Poland, against 63% in Finland and a 37% global average. Romania has the highest share of the four CEE markets paying for online news (15%, +3), and 47% sometimes or often avoid the news. 32% of Romanian respondents get content from news-focused creators, and Pro TV reaches 60% weekly online and offline.
Module: nations · grade D · 2026-06-16. Radu, R.-N. et al. (2026). Digital News Report 2026: Romania (with Hungary, Bulgaria, Poland and Finland country pages). Reuters Institute for the Study of Journalism
In 85 of 136 countries, under-25s rate their lives higher in 2023–2025 than in 2006–2010, but youth life evaluations fell by an average of 0.86 points (0–10 scale) in the US, Canada, Australia and New Zealand, by 0.312 in Western Europe and 0.416 in the UK and Ireland. The report says youth wellbeing has fallen only in those regions, while the top five gainers overall are all in Central and Eastern Europe: Serbia, Bulgaria, Georgia, Latvia and Bosnia and Herzegovina. Finland ranks first for the ninth year, Czechia 20th and Slovenia 18th of 147, and Romania is reported 34th.
Module: nations · grade A · 2026-03-19. Helliwell, J. F. et al. (2026). World Happiness Report 2026, Chapter 2: International evidence on happiness and social media. Wellbeing Research Centre, University of Oxford / Gallup
In 2024, cash made up 52% of euro-area point-of-sale transactions by number (59% in 2022) and 39% by value, ranging from 67% in Malta and 64% in Slovenia to 22% in the Netherlands and 27% in Finland, where cash was also 17% of value. Consumers under 40 used cash for less than half of their in-store payments, against 57% for those aged 65 and over. Although 55% prefer cards or other cashless means, 62% say having the option to pay cash is important, across 40,981 respondents.
Module: nations · grade A · 2024-12-19. European Central Bank (2024). Study on the payment attitudes of consumers in the euro area (SPACE) – 2024. ECB
Across 25 countries, a median of 34% are more concerned than excited about AI, 42% equally concerned and excited and 16% more excited; about half are mainly concerned in the US, Italy, Australia, Brazil and Greece, against as few as 16% in South Korea. In Greece, 59% of adults 50+ are more concerned versus 18% of 18–34s. Trust in one's own country to regulate AI is 89% in India, 74% in Indonesia and 72% in Israel but 22% in Greece, while a median of 53% trust the EU to regulate it (US 37%, China 27%), with Poland among the least trusting EU members.
Module: nations · grade A · 2025-10-15. Pew Research Center (2025). How People Around the World View AI. Pew Research Center
A National Academies consensus report concluded that dividing the workforce into generations “is not strongly supported by science and is not useful for workforce management”. It found that many findings attributed to generations “may actually reflect” age-related variation or changing work, and that the research relies “too heavily on cross-sectional surveys and convenience samples”. In May 2023 Pew said it “won't always default to using the standard generational definitions and labels” and will discuss generations only when it can compare them at similar life stages.
Module: cohorts · grade A · 2023-05-22. National Academies of Sciences, Engineering, and Medicine (2020). Are Generational Categories Meaningful Distinctions for Workforce Management? National Academies Press; with Parker, K. (2023). How Pew Research Center will report on generations moving forward. Pew Research Center
In Pew's February 2026 survey (n=5,119), 66% of adults aged 18–29 had ever used an AI chatbot, against 61% at 30–49, 42% at 50–64 and 23% at 65+ (2024: 55%, 41%, 25%, 12%). Daily use was 31%, 34%, 19% and 7%, and ChatGPT use was 61%, 55%, 37% and 19%. Only one in five Americans 65+ have heard “a lot” about chatbots, and 29% of them are unsure how AI will affect them personally.
Module: cohorts · grade A · 2026-06-17. Gottfried, J., Bishop, W., Anderson, M., Faverio, M., Park, E., McClain, C. (2026). How opinions and use of AI differ by age. Pew Research Center
Pew's survey of 1,458 teens aged 13–17 (Sept–Oct 2025) found that 97% go online daily, 4 in 10 almost constantly, and 21% use TikTok almost constantly (16% in 2022). 64% have used AI chatbots (66% in households earning $75k+ vs 56% under $30k), and boys lead on YouTube (94% vs 89%) while girls lead on Snapchat (61% vs 49%). Pew's February 2026 report “How Teens Use and View AI” adds that 57% of teens have used chatbots to search for information, 54% for help with schoolwork, about one in five for news and 12% for emotional support or advice.
Module: cohorts · grade B · 2025-12-09. Pew Research Center (2025). Teens, Social Media and AI Chatbots 2025. Pew Research Center; with McClain, C., Anderson, M., Sidoti, O., Bishop, W. (2026). How Teens Use and View AI. Pew Research Center
In a probability-based Ipsos KnowledgePanel survey of 1,578 parents (August 2024), children aged 0–8 averaged about 2.5 hours of screen time a day: 1:03 under 2, 2:08 at 2–4 and 3:28 at 5–8, with 3:48 in households under $50k against 1:52 at $100k+. 40% have their own tablet by age 2 and about one in four have their own phone by 8, while gaming time is up 65% since 2020. 62% of parents sometimes co-view YouTube but only 17% co-view TikTok, and 80% worry about excessive screen time.
Module: cohorts · grade A · 2025-02-26. Common Sense Media (2025). The 2025 Common Sense Census: Media Use by Kids Zero to Eight. Common Sense Media
In Gallup data pooled from 2023 and 2024, 25% of US men aged 15–34 said they felt lonely “a lot of the previous day”. That compares with 18% of young US women, 15% of men aged 35–54, 16% of men 55+ and 18% of all US adults. The OECD median for young men is 15%.
Module: cohorts · grade A · 2025-05-20. Vigers, B. (2025). Younger Men in the U.S. Among the Loneliest in West. Gallup
In the Federal Reserve's October 2025 survey (n=12,934), 16% of adults used buy now, pay later in the past year (10% in 2021): 22% at 18–29, 21% at 30–44, 17% at 45–59 and 9% at 60+, with women at 19% and men at 14%. 26% of users paid late, including 32% of users aged 18–29 and 40% of users with incomes under $25k, and 11% had a payment trigger an overdraft or NSF fee. 29% used it because it was the “only way I could afford it” (40% of those under $25k), about one in five used it for groceries or food delivery, and only 14% answered both credit-reporting questions correctly.
Module: cohorts · grade C · 2026-05. Board of Governors of the Federal Reserve System (2026). Economic Well-Being of U.S. Households in 2025: Credit. Federal Reserve
In FINRA's 2024 National Financial Capability Study investor survey (n=2,861), 60% of investors aged 18–34 used social media for investing information and 61% followed finfluencers, against 37% and 31% at 35–54 and 9% and 6% at 55+. Social media users scored lower on knowledge quizzes (42% vs 47% for non-users) but rated their own knowledge higher (63% vs 53%), and among investors targeted by fraud, 68–69% of users and followers lost money against 26–29% of non-users. 46% of users agree that “people like me aren't usually investors”, against 23% of non-users.
Module: cohorts · grade B · 2026-04. Ko, K. J., Green, S., Mottola, G. R., Valdes, O. (2026). Finfluencer Followers and Social Media Scrollers: The Profile, Patterns and Pitfalls of Social-Media-Informed Retail Investors. FINRA Investor Education Foundation
In 2024, adults 50+ generated $12.5 trillion of US economic activity, 43% of GDP (2018: $8.3 trillion, about 40%), and households they head spent $10.7 trillion, “more than half of all consumer spending”. AARP projects this spending will reach $22 trillion, more than 61% of consumption, by 2060, as the 50+ share of the population rises from about 36% to 41%.
Module: cohorts · grade D · 2026-06-02. Huddleston, C. (2026). Older Americans Are Boosting the Economy for All (Longevity Economy Outlook 2026). AARP
In 2025 private label sold €387bn across 17 European markets (+€15.3bn), a 38.8% value share; it grew 4.1% versus 3.2% for the total market and 2.7% for manufacturer brands. Private label units rose 1.3% while branded units dipped slightly. Switzerland is above 50%, Portugal at 48.2% and Norway lowest at 23.6%, and private label outgrew brands in 12 of 17 countries.
Module: sectors · grade C · 2026-04-21. PLMA International / NielsenIQ (2026). A Mosaic of Markets: PLMA's 2026 Report on the Status of Private Label Across Europe. PLMA International
Households with at least one GLP-1 user cut grocery spending 5.3% within six months of adoption (8.2% for higher-income households); savory snacks fell 10.1% and spending at fast-food chains, coffee shops and limited-service restaurants fell 8.0%. Yogurt showed the largest rise among the few categories that grew, and effects persist through year one but weaken after six months. In the Numerator panel of about 150,000 households, homes reporting a GLP-1 user rose from about 11% (late 2023) to more than 16% (mid-2024).
Module: sectors · grade B · 2025-12-18. Hristakeva, S., Liaukonytė, J., Feler, L. (2025). The No-Hunger Games: How GLP-1 Medication Adoption Is Changing Consumer Food Demand. Journal of Marketing Research 63(3):427–444
Luxury consumers fell from about 400 million in 2022 to about 340 million in 2025; new-customer acquisition dropped 5% from 2024 to 2025, and active shoppers fell from about 60% of the addressable base to 40–45%. Personal luxury goods are about €358bn (−2% at current rates, flat at constant), with Europe at −1% to −3% and China at −3% to −5%. Reuters reports Bain expects another 20–30 million lost clients and 3–5% growth in 2026, and quotes a Bain partner saying high-end clients “feel betrayed”.
Module: sectors · grade D · 2025-11-20. Bain & Company, Altagamma (2025). Global luxury stays resilient despite economic headwinds and shifting consumer trends that reshape market (Luxury Study 2025). Bain & Company press release
In 2025 the US secondhand market grew nearly 4x faster than the broader retail clothing market; the US market is projected to reach $78.8bn and the global market $393bn by 2030, growing about 2x faster than overall apparel. Gen Z and Millennials are projected to deliver 71% of market growth through 2030. Nearly 50% of shoppers discover secondhand finds through social media, creators and influencer feeds (survey of 3,268 US adults, Jan–Feb 2026).
Module: sectors · grade D · 2026-04-02. ThredUp, GlobalData (2026). ThredUp's 14th Annual Resale Report. ThredUp Newsroom
IDC forecasts 2026 shipments down 16.7% to just over 1 billion units, the steepest drop on record, while average selling price rises 27.6% to $581 and market value 6.3% to $613bn. NAND and DRAM costs are up more than 300% year on year, sub-$100 phones fell almost 60% year on year in Q2 2026, and second-half shipments are forecast at −27.2%. Android is forecast at −24.3% versus −1.3% for iOS, and IDC says consumers will likely hold devices longer.
Module: sectors · grade D · 2026-08-26. Popal, N., Jeronimo, F., Reith, R. (2026). Smartphone Shipments Set for Record 16.7% Drop in 2026, as the Memory Crisis Hits Full Force. IDC
Revolut says it has more than 5 million retail customers in Romania after adding over 500,000 in under 12 months; Finance Magnates puts this at about one-third of the estimated 15.5 million adults, noting these are registered, not active, accounts. From Sep 2025 to Aug 2026 Revolut cards averaged more than 56 million payments a month, 77% of them at merchants in Romania, and the company's internal “primary account” measure rose 40%. In February 2026 Romania Insider reported Revolut had overtaken Banca Transilvania (about 4.6 million active cards) by number of card users.
Module: sectors · grade D · 2026-09-30. Marica, I. (2026). Revolut reaches 5 mln retail customers in Romania, plans new Bucharest headquarters. Romania Insider (company data; see also Finance Magnates, 1 Oct 2026)
International arrivals rose 0.4% in H1 2026 to about 690 million (Q1 +2%, Q2 −1%, June −3%); Europe grew 3% but Western Europe fell 6% in June amid a heatwave, while the Middle East dropped 22%. Asia-Pacific is up 1% but still 11% below 2019. UN Tourism cut its 2026 forecast to +1–2% from 3–4% in January and expects travellers to seek value for money and closer-to-home or domestic trips.
Module: sectors · grade C · 2026-09-17. UN Tourism (2026). International tourism holds steady with 0.4% growth in first half of 2026 as Middle East conflict and rising costs weigh on momentum. World Tourism Barometer 24(3)
Netflix says its ad-supported plan has more than 250 million global monthly active viewers, with over 80% of ads members watching every week, and that the plan expands to 15 more countries from 2027, including Poland, the Netherlands, Belgium, Sweden, Austria and Ireland. Variety reports the previous figure was 190 million (November 2025), a 31% rise.
Module: sectors · grade C · 2026-05-13. Netflix (2026). Netflix Upfront 2026: Get Closer. Netflix newsroom
12% of US adults currently take a GLP-1 drug and 18% have ever taken one (KFF, n=1,350, Oct 27–Nov 2, 2025, ±3 pp): 15% of women, 9% of men, 22% of 50–64-year-olds and 9% of those 65+. Current use is 45% among adults with diabetes, 29% with heart disease and 23% among those overweight or obese in the past five years; 56% of users say the drugs were hard to afford and 27% of insured users paid the full cost. Gallup separately reports 12.4% taking a GLP-1 for weight loss, up from 5.8% in February 2024, and US obesity at 37%, down from 39.9% (via NPR).
Module: sectors · grade C · 2025-11-14. KFF (2025). KFF Health Tracking Poll: 1 in 8 adults say they are currently taking a GLP-1 drug for weight loss, diabetes or another condition, even as half say the drugs are difficult to afford. KFF
Total beverage alcohol volume fell 2% in 2025 (beer −2%, wine −5%, spirits −3%, RTDs +3%), while RTD cocktails grew 14% and no-alcohol beer 8%. IWSR forecasts volumes declining until 2031 and ending 2035 about 1% below 2025, with wine −14% and RTDs +17% over 2025–2035. No-alcohol beer is forecast to rise from 2% to 4% of global beer volume by 2033.
Module: sectors · grade D · 2026-06-11. IWSR (2026). Global beverage alcohol to drop further before recovering. IWSR press release
In 2025 the 16-market European games market was worth €29.7bn, 91% of it digital, with 199 million players aged 6–75 (51% of that population) playing 8h59m a week on average; the average player is 34 and 46% are women. 82% of players use smartphones or tablets, 47% consoles and 34% PC, and revenue splits 49% mobile, 30% console and 12% PC; in the five largest markets revenue fell 4% versus 2024. Romania: €449m in revenue, 8.6 million players, 52% of people aged 6–75.
Module: sectors · grade C · 2026-08. Video Games Europe (2026). 2025 All About Video Games: European Key Facts (data: GameTrack Ipsos, GSD). Video Games Europe
Meta says every public Reels and Feed post on Instagram is now processed by an LLM for ranking. Its largest Reels ranking release added 15 basis points to sessions. On Instagram Feed, more than half of recommended content is under one day old, over double a year earlier. Facebook video time rose 9% globally year over year.
Module: algo · grade C · 2026-07-29. Meta Q2 2026 Prepared Remarks (Meta Investor Relations)
Meta introduced the Meta Generative Recommender and put its first generative model into ads retrieval. With GEM, it reported +8.3% ad clicks and +15.7% conversions on Facebook. Advantage+ passed a $75B annual run-rate, and 9 million small businesses use at least one AI creative tool.
Module: algo · grade C · 2026-07-29. Meta Q2 2026 Prepared Remarks (Meta Investor Relations)
A public Threads post starting with “Dear Algo” adjusts that user's feed for three days, and others can repost the request to apply it to their own feed (US, UK, Australia, NZ). Instagram is testing ways to make “Your Algorithm” central to the app: pull-down on Feed, swipe-up on a Reel, more-like-this buttons.
Module: algo · grade C · 2026-02-11. Control Your Threads Feed With New Dear Algo Feature (Meta Newsroom)
The TikTok USDS Joint Venture (Silver Lake, Oracle, MGX at 15% each; ByteDance 19.9%) “will retrain, test, and update the content recommendation algorithm on U.S. user data”, secured in Oracle's US cloud, and holds authority over US trust & safety and moderation.
Module: algo · grade C · 2026-01-23. Announcement from the new TikTok USDS Joint Venture LLC (TikTok Newsroom)
YouTube says Shorts recommendations will “further prioritize original content and reduce the reach of content re-uploaded from other creators”. Channels that mainly aggregate or re-upload others' videos without significant changes get less distribution; “minor technical edits or template-based changes” do not count as original. No numbers or thresholds given.
Module: algo · grade C · 2026-10-01. YouTube is cleaning up one of the worst parts of Shorts (Android Authority)
LinkedIn replaced its separate retrieval sources (chronological, trending, collaborative filtering) with one pipeline built on LLM-generated embeddings, and ranks with a sequential Generative Recommender over more than 1,000 past interactions per member. Retrieval runs under 50 ms; recall@10 rose 15%.
Module: algo · grade C · 2026-03-12. LinkedIn rebuilds its feed from scratch with LLMs (PPC Land, on LinkedIn Engineering post of 12 Mar 2026)
X open-sourced its For You code. A transformer ranker (“Phoenix”) predicts the probability of about 20 actions, positive and negative, and the score is a weighted sum Σ(weight × P(action)), followed by an out-of-network discount, author-diversity decay and a new-author boost.
Module: algo · grade C · 2026-01-20. X open sources its algorithm (TechCrunch); repo github.com/xai-org/x-algorithm
Pinterest says delivery updates gave a 28% ROAS improvement in testing for SMB advertisers, and Smart Assembly (an AI-assembled ad per impression) gave +6% CTR in alpha. Its open models beat closed models by 34 points on shopping-recommendation relevance. About 30% of lower-funnel revenue ran through Performance+ as of Q1.
Module: algo · grade C · 2026-08-05. Pinterest Q2 2026 Earnings Call Transcript (Pinterest IR)
Google is testing two Gemini-built ad formats in AI Mode: Conversational Discovery ads (creative generated for the specific query) and Highlighted Answers (ads placed inside AI Mode recommendation lists). Direct Offers, piloted since January 2026, is adding Gemini-built bundles and native checkout for Universal Commerce Protocol merchants.
Module: algo · grade C · 2026-05-20. A new generation of ads for the AI era of Search (Google Ads & Commerce Blog)
ChatGPT shows labelled ads, visually separated and placed after the answer, to logged-in adult Free and Go users, matched to the conversation topic and past chats. After the US (Feb 2026) it expanded to 31 European markets in August; self-serve buying (CPM/CPC bids, conversion optimisation, geo and custom audiences) has been open there since 31 Aug. No ads near health, mental health or politics, or for users predicted to be under 18.
Module: algo · grade C · 2026-08-31. ChatGPT ads expands across Europe (OpenAI)
Social and video networks are now the most-used online news route (54%), ahead of news websites and apps (51%). 52% of 18–24s say social, video or chatbots are their main news source. 27% get news weekly from news-focused creators, 46% from creators of any kind. In Romania, 32% get news from news-focused creators.
Module: intel · grade B · 2026-06-16. Digital News Report 2026, Executive Summary and Romania page (Reuters Institute)
Trust in news overall is 37%, the lowest since 2015; trust in news on social media is 22%, in AI chatbot answers 20%. 10% use chatbots for news weekly (up from 7%), 16% of under-35s. 42% sometimes or often avoid the news. In Romania trust is 23% (down 3 points, a fifth consecutive record low) and avoidance 47%.
Module: intel · grade B · 2026-06-16. Digital News Report 2026 (Reuters Institute)
In tracked browsing of 900 US adults (March 2025), Google searches showing an AI summary led to a click on a traditional result in 8% of visits, versus 15% without one. Only 1% clicked a link inside the summary. Sessions ended after 26% of summary pages vs 16% without. About 18% of searches produced a summary.
Module: intel · grade A · 2025-07-22. Google users are less likely to click on links when an AI summary appears in the results (Pew Research Center)
Non-work messages grew from 53% to more than 70% of ChatGPT use. Practical Guidance, Seeking Information and Writing together account for nearly 80% of conversations, measured when about 10% of the world's adults had adopted ChatGPT (July 2025).
Module: intel · grade B · 2025-09-15. Chatterji, Cunningham, Deming, Hitzig, Ong, Shan & Wadman (2025). How People Use ChatGPT. NBER Working Paper 34255
On Claude.ai (Nov 2025), augmentation was 52% of conversations (+5 points) and automation 45% (−4). Directive delegation fell to 32% from 39% in August 2025. Coding-type tasks were 34% of Claude.ai conversations vs 46% of API traffic; task success was 67% on Claude.ai vs 49% on the API.
Module: intel · grade C · 2026-01-15. Anthropic Economic Index report, January 2026 (Anthropic)
Edelman's 2026 Trust Barometer finds 70% globally unwilling or hesitant to trust people with different values or information sources, while 78% trust their employer. About 34,000 respondents in 28 countries, online, Oct–Nov 2025.
Module: intel · grade D · 2026-01-18. 2026 Edelman Trust Barometer: Society Slides into Insularity (Edelman)
Cloudflare's crawl-to-refer ratios, Jan→Jul 2025: Anthropic 286,930→38,066 crawls per referral, OpenAI 1,217→1,091, Perplexity 55→195, Google 3.8→5.4. Training was 79% of AI crawling in July 2025, up from 72%. In 2025, non-Google AI bots made 4.2% of HTML requests, against 4.5% for Googlebot alone.
Module: intel · grade C · 2025-08-29. The crawl-to-click gap: Cloudflare data on AI bots, training, and referrals (Cloudflare)
Imperva's 2026 Bad Bot Report: bots are more than 53% of web traffic and 40% of bot traffic is malicious. AI-driven bot attacks rose 12.5×; APIs and identity systems are 27% of attacks.
Module: intel · grade D · 2026-05-05. The 2026 Bad Bot Report is here (Thales / Imperva)
Across 76,977 UK participants, 19 LLMs and 707 political issues, post-training made models up to 51% more persuasive and prompting up to 27%. The same methods that raised persuasion lowered factual accuracy (466,769 claims fact-checked).
Module: papers · grade A · 2025-12-11. Hackenburg, Tappin, Hewitt … Rand & Summerfield (2025). The levers of political persuasion with conversational AI. Science
In a preregistered 10-day field experiment with 1,256 X users during the 2024 US campaign, a browser extension reranked feeds in real time to change exposure to antidemocratic and partisan-animosity posts. Out-party animosity moved by about 2 points on a 100-point feeling thermometer, with no difference between parties.
Module: papers · grade A · 2025-12-02. Piccardi, Saveski, Jia, Hancock, Tsai & Bernstein (2025). Reranking partisan animosity in algorithmic social media feeds alters affective polarization. Science
Removing political ads from the feeds of 36,906 Facebook and 25,925 Instagram users in the six weeks before the 2020 US election had no detectable effect on political knowledge, participation (including donations), candidate favourability or turnout.
Module: papers · grade A · 2026-03-02. Allcott, Gentzkow, Levy … Tucker (2026). The Effects of Political Advertising on Facebook and Instagram before the 2020 US Election. Nature Human Behaviour 10, 884–895
Users who deactivated for six weeks before the 2020 US election improved on a happiness, depression and anxiety index by 0.060 SD on Facebook (19,857 users) and 0.041 SD on Instagram (15,585 users), versus controls who deactivated for one week. Exploratory: the Facebook effect sits in users over 35, the Instagram effect in women under 25.
Module: papers · grade D · 2025-04-25. Allcott, Gentzkow, Wittenbrink … Tucker (2025). The Effect of Deactivating Facebook and Instagram on Users' Emotional State. NBER Working Paper 33697
In a Google Display Network field experiment on cosmetics ads, fully AI-generated ads beat human-expert ads by up to 19% on click-through. An AI-disclosure label cut click-through by about 31.5%. Editing human ads with AI did not help.
Module: papers · grade D · 2025-11-06. Lee, Todri, Adamopoulos & Ghose (2025). The Impact of Visual Generative AI on Advertising Effectiveness. SSRN working paper (NYU Stern summary)
In a preregistered RCT with 467 iPhone users, blocking mobile internet for two weeks (calls and texts still worked) improved sustained attention by dz = 0.23, comparable to about ten years of age-related decline. 70.5% said their mental health improved.
Module: papers · grade A · 2025-02-18. Castelo, Kushlev, Ward, Esterman & Reiner (2025). Blocking mobile internet on smartphones improves sustained attention, mental health, and subjective well-being. PNAS Nexus 4(2)
ChatGPT personas built from ANES respondents (3,614,400 synthetic responses) matched human averages within 1 SD, but 48% of regression coefficients differed significantly from the real data and 32% of those had the wrong sign. The same prompt gave different results in April, June and July 2023.
Module: papers · grade B · 2024-05-17. Bisbee, Clinton, Dorff, Kenkel & Larson (2024). Synthetic Replacements for Human Survey Data? The Perils of Large Language Models. Political Analysis 32(4), 401–416
Across 6 studies, some preregistered, a chat with an AI companion lowered loneliness: post-chat loneliness was 25.62 vs 36.91 for control (d = 0.40, N = 713). In a one-week longitudinal study (N = 922), people underestimated how much it would help.
Module: papers · grade D · 2024-07-26. De Freitas, Uguralp, Uguralp & Puntoni (2024). AI Companions Reduce Loneliness. arXiv 2407.19096 / HBS Working Paper 24-078
IN FORCE. The Digital Omnibus on AI (Regulation (EU) 2026/1744, in force 27 July 2026) pushes high-risk AI rules to 2 Dec 2027 (standalone) and 2 Aug 2028 (embedded), while Article 50 transparency duties (tell people they are dealing with AI, label deepfakes) apply from 2 Aug 2026. Systems already on the market have until 2 Dec 2026 to add labels or watermarks.
Module: wire · grade D · 2026-07-27. EU AI Act: AI Omnibus formally adopted (Addleshaw Goddard); Digital Omnibus on AI enters into force (ERTICO)
ADOPTED. The Commission fined Temu €200 million under the DSA for failing its risk-assessment duty: it used generic e-commerce data, underestimated EU users' exposure to illegal products, and did not assess how recommender systems and influencer-led promotion amplify them. Second DSA fine after X (€120M, Dec 2025).
Module: wire · grade D · 2026-05-28. European Commission fines Temu €200 million under the DSA (Slaughter and May, The Lens)
PRELIMINARY FINDING. The Commission found Meta in breach of the DSA: a false birth date is accepted at sign-up, reporting an underage user takes up to 7 clicks and often goes nowhere, and Meta's risk assessment ignores evidence that roughly 10–12% of under-13s use the platforms. Possible fine up to 6% of global turnover. A separate probe into addictive “rabbit hole” design is open.
Module: wire · grade C · 2026-04-29. Commission preliminarily finds Meta in breach of the DSA for failing to prevent minors under 13 (European Commission)
NOT YET PROPOSED. The European Parliament's Legislative Train lists a Commission initiative planned for Q4 2026, covering dark patterns, addictive design, influencer marketing and unfair personalisation, with subscription cancellation and renewal named as concerns. Consultation closed 24 Oct 2025; reporting expects negotiations into 2027 and phased application over 2028–2030.
Module: wire · grade C · 2026-10-01. Digital Fairness Act, Legislative Train Schedule (European Parliament)
IN FORCE. Since 1 July 2026 the EU charges a flat €3 customs duty on goods worth up to €150 from outside the bloc, per tariff category within a consignment: five T-shirts cost €3, a T-shirt plus a watch €6. It runs until 1 July 2028, when a permanent regime is expected.
Module: wire · grade D · 2026-07-01. Shein and Temu hit by new EU customs duty on cheap parcels (Retail Gazette)
MOSTLY PILOTS. OpenAI moved away from in-chat Instant Checkout (launched Sep 2025): merchants can use their own checkout and the focus shifted to discovery. Forrester cites Walmart reporting about 3× lower conversion inside Instant Checkout than with a click-out to its site. Google launched the Universal Commerce Protocol (11 Jan 2026); Europe's first live agent payment (Mastercard, Santander, 2 Mar 2026) was a controlled test, not a rollout.
Module: wire · grade D · 2026-03-25. Agentic Payments In B2C Commerce: Where We Are Now (Forrester)
DATA. INS: Romania's retail sales volume fell 5.6% in H1 2026 vs H1 2025; non-food −7.3%, food, drinks and tobacco −4.2%, fuel −4.0%. June was −7.3% year on year (seasonally adjusted). BCR says real wages have been negative for 11 consecutive months; the Finance Minister expects 5–6% inflation at year-end.
Module: wire · grade C · 2026-08-06. INS: Volumul cifrei de afaceri din comerțul cu amănuntul a scăzut cu 5,6% în primul semestru (AGERPRES)
ANPC joined an ICPEN international influencer sweep (23 Mar–8 Apr 2026) and checked 10 Romanian influencers: sponsored posts were not consistently labelled, and 70% of those who used AI-generated or AI-altered content did not disclose it. Across all countries, 228 influencers were checked: 92% posted commercial content, about 1 in 5 labelled it properly. No fines reported.
Module: wire · grade D · 2026-07-02. Activitatea influencerilor, verificată de ANPC (Digi24)
Non-experts identified AI poems as AI only 46.6% of the time, below chance, and judged them human more often than poems by famous poets. With the source hidden they rated AI poems higher (d = 0.67). Being told a poem was AI-written lowered ratings (d = −0.51).
Module: anomaly · grade A · 2024-11-14. Porter & Machery (2024). AI-generated poetry is indistinguishable from human-written poetry and is rated more favorably. Scientific Reports 14, 26133
In a global tournament, no intervention increased effortful pro-environmental behaviour (a tree-planting task) versus control, and several reduced it, including inducing negative emotion and writing a letter to future generations. The same interventions slightly raised belief (~2.3 points), policy support (~2.6) and sharing (~12.1).
Module: anomaly · grade A · 2024-02-07. Vlasceanu, Doell, Bak-Coleman … Van Bavel (2024). Addressing climate change with behavioral science: A global intervention tournament in 63 countries. Science Advances 10, eadj5778
Short personalised conversations with GPT-4 reduced 2,190 believers' belief in their chosen conspiracy theory by about 20%; the effect was still there two months later and spilled over to unrelated conspiracies. Fact-checkers rated 99.2% of the AI's claims accurate.
Module: anomaly · grade A · 2024-09-13. Costello, Pennycook & Rand (2024). Durably reducing conspiracy beliefs through dialogues with AI. Science
In a preregistered RCT (N = 2,159 US adults), ChatGPT fact-checks did not improve discernment between true and false headlines, although the model flagged about 90% of false ones. Belief in true headlines it wrongly flagged fell 12.75%; belief in false headlines it was unsure about rose 9.12%. Human fact-checks did help.
Module: anomaly · grade A · 2024-12-04. DeVerna, Yan, Yang & Menczer (2024). Fact-checking information from large language models can decrease headline discernment. PNAS 121(50), e2322823121
Lower AI literacy predicts higher AI receptivity across 27 countries (B = −8.80, p < .001) and in six individual-level studies (e.g., B = −.10, t(1,004) = −10.80). It reverses for tasks that do not need distinctly human qualities. Most forecasters, executives included, predicted the opposite.
Module: anomaly · grade A · 2025-01-01. Tully, Longoni & Appel (2025). Lower Artificial Intelligence Literacy Predicts Greater AI Receptivity. Journal of Marketing 89(5), 1–20
In 4 preregistered experiments (N = 556), third-party judges rated AI-written empathic replies as more compassionate than human ones and preferred them, including against expert crisis-line responders and even with authorship disclosed.
Module: anomaly · grade B · 2025-01-10. Ovsyannikova, Oldemburgo de Mello & Inzlicht (2025). Third-party evaluators perceive AI as more compassionate than expert humans. Communications Psychology 3