The Fact-checking Observatory is an automatic service that collects misinforming content on Twitter using URLs that have been identified as potential misinformation by fact-checking websites. Using this data, the Fact-checking Observatory automatically generates weekly reports that updates the state of misinformation spread of fact-checked misinformation on Twitter.
This analysis is limited to URLs identified by Fact-checking organisations. The collected data only consist of non-blocked Twitter content and may be incomplete.
This report updates the status of misinformation spread between Monday 13 February 2023 and Monday 20 February 2023.
New:+48 Trend:-2,109
91,218 Fact-checking Tweets
New:+188 Trend:-101
5,776 Fact-checks
143 Fact-checking Organisations
Key Content and Provenance
During the period between Monday 13 February 2023 and Monday 20 February 2023, 48 new URLs have been identified as potential misinforming content. Out of the 288 domains identified by Fact-checking organisations (Figure 1), most of the new shared URLs were from afp.com with an increase of +48 compared to the previous total spread for the same domain The domain that saw the least increase in spread compared to the previous period total spread was 1tv.ru with a change of +0 compared to the previous total spread for the same domain
In relation to the previous week, the domain that saw the biggest relative spread change was correctiv.org with a change of +12 compared to the previous total spread for the same domain whereas the domain that saw the least relative change was twitter.com with a change of -2,076 compared to the previous period.
The all time most important domain is twitter.com with a total of 182,968 URL shares and the least popular domain is 24-post.com with 1 shares (Figure 2).
Figure 1: Domain importance.
Figure 2: Amount of domains shares per week.
The top misinforming content and fact-checking articles shared since the last report are listed in Table 1 and Table 2.
Table 1: Top misinforming content.
Table 2: Top fact-checked content.
Fact-Checkers and Spreaders Location
The data used for creating the Twitter dataset is obtained from 143 fact-checking organisations.
The largest amount of fact-checked content comes from euvsdisinfo.eu (840 fact-checks) and the least from Verify Sy (1 fact-checks). Most fact-checked content are from VoxCheck (352) followed by Factcheck.ge (277) and AFP fact checking (255) (Figure 3).
Figure 3: Amount of fact-checks by fact-checkers.
Figure 4: Identified location of users spreading fact-checks and misinformation.
Locations and Mentions
Using automatic entity extraction methods, we identify key locations and persons mention in the fact-checking articles in order to identify what location and person are the most discussed in misinforming content.
The top mentioned locations and persons are listed in Table 3 and Table 4.
| Location | Description | Current Week | Previous Week | Total |
|---|---|---|---|---|
| U | [‘Country in eastern europe’]. | 205 | 2529 | 133643 |
| G | [‘Country in central europe’]. | 55 | 2165 | 14967 |
| B | [‘Capital city of iraq’]. | 45 | 26 | 13313 |
| E | [‘Continent’]. | 40 | 41 | 17785 |
| I | [‘Country in western asia’]. | 32 | 2127 | 24592 |
| P | [‘Country in central europe’]. | 31 | 2160 | 15060 |
| R | [‘Statue in chadron, united states of america - frédéric auguste bartholdi - 1950’]. | 19 | 29 | 8092 |
| M | [‘Country in north america’]. | 18 | 42 | 5333 |
| S | [‘Sculpture on liberty island in new york harbor in new york city, new york, united states’]. | 17 | 13 | 8390 |
| L | [‘Sovereign state in western asia’]. | 12 | 2121 | 4579 |
Table 3: Top locations mentioned in misinforming posts.
| Person | Description | Current Week | Previous Week | Total |
|---|---|---|---|---|
| V | [‘President of russia (1999–2008, 2012–present)’]. | 76 | 107 | 60937 |
| J | [‘American educator, wife of joe biden and first lady of the united states’]. | 17 | 5 | 15331 |
| H | [‘Ukrainian energy minister’]. | 16 | 14 | 2248 |
| O | [‘Ukrainian politician and entertainer’]. | 14 | 11 | 2125 |
| E | [‘President of france since 2017’]. | 12 | 2124 | 7833 |
| U | [‘President of the european commission since 2019’]. | 12 | 2121 | 4907 |
| R | [‘President of turkey since 2014’]. | 11 | 1 | 2622 |
| M | [‘Ukrainian politician’]. | 8 | 9 | 3066 |
| A | [‘Austrian-born german politician, dictator of germany from 1933 until his death in 1945’]. | 6 | 29 | 4825 |
| B | [‘King of thailand (1927-2016)’]. | 4 | 8 | 2200 |
Table 4: Top people mentioned in misinforming posts.
Demographic Impact
Using automatic methods, Twitter account demographics are extracted for user age, gender and account type (i.e., identify if an account belong to an individual or organisation).
Figure 6 displays how misinformation and fact-checks are spread by different demographics.
Figure 5: Misinformation and Fact-check spread for different demographics. Top: Gender, Center: Age group, Bottom: Account type.
Data Collection and Methodology
The full methodology and information about the limitation and dataset used for this analysis can be accessed in the methodology page.