Australian online media are suffering as top news stories are being incorrectly flagged as non-brand safe due to the widespread use of keyword blacklists by brands, according to a study by AI ad verification company CHEQ.
CHEQ's study used an industry-standard blocklist of 2000 words which are widely used to prevent brands from appearing online next to harmful, hateful or toxic content.
This lists includes words such as “death” “racism”, “sex” and “injury” associated with controversial topics.
CHEQ ran the words against 225 neutral or positive online articles on a single news day, 18 July, across 15 major news sites including News Corp-owned news.com.au.
In the case of News.com.au the study found 53% of safe or positive content on the site, including stories on entertainment and history, were flagged through the use of over-reaching keywords, such as "attack" and "army" effectively demonetizing the content.
Examples of stories deemed unsafe for advertisers on the Australian news site included mention of a Salvation Army visit which was blocked for the keyword "army", a story about the Ashes test match in which the captain of the Australian women’s cricket team Meg Lanning boasts of having a good "attack", and NSW coach Brad Fittler issuing a “blunt warning to the Maroons over a staggering plan to steal a Blues star from under his nose” which was blocked for the keyword "blunt".
Unable to recognise the difference between news about real wars or violence and entertainment, Tom Cruise's return in Top Gun was also blocked for "gun".
History stories also were unable to be monetised. A brief mention of the Cold War saw “war” blocked.
The wider study included analysis of keyword blocklists on the day on a total of 15 top global news sites, including the New York Times, Financial Times and the Washington Post.
It found 57% - 129 stories - of uncontroversial news reports on a single day, were incorrectly flagged as unsafe due to keyword detection preventing brands from appearing and damaging monetization of news.
The findings reveal that the use of such keywords blocks 73% of neutral stories on gay news sites studied including all mention of “bisexual” and “lesbians”.
Keywords also blocked 75% of safe history news blocked by keywords including speeches by JFK and stories about the moon landing because of keywords such as "war" and "weapons".
They also blocked 65% of entertainment news including Walking Dead and Dead to Me for "dead", a new season of Fargo for "mafia", and Fantastic Beasts: The Crimes of Grindelwald for “crimes”.
The findings are a conservative estimate of the number of daily stories being denied ad dollars with many blocklists going further than the limited examples used in our study blocking entire mentions of politicians such as "Donald Trump" "Westminster" or "Theresa May".
Through the report, CHEQ is calling for the end of the "blunt" instrument of keywords in favor of AI-based detection which can read and understand news in context.
“These findings lay bare the shocking reality of the failings of outdated brand safety measures," CHEQ founder and CEO Guy Tytunovich says.
"It is frankly embarrassing that blunt keyword technology is still being used in 2019 when technology, in particular, AI, can read stories online with the same accuracy as humans. There are many genuine challenges facing the online news industry, but lack of revenue because of the discredited use of unsophisticated technology, needs not be one of them.”
Have something to say on this? Share your views in the comments section below. Or if you have a news story or tip-off, drop us a line at firstname.lastname@example.org