Google rewards credit card websites that use language to "frighten" their users about the risk of falling into debt, according to a machine learning study into how the search giant ranks different sectors.

The study by London-based artificial intelligence agency Artios notes that Google is acting as a "regulator by proxy", invoking fear in consumers to protect them from meandering into high levels of debt.

Artios thinks Google rightly treats the credit card sector differently from others, giving sites that emphasise risk over benefit more visibility in their search results, placing them nearer to the top, and mirroring the requirements of industry regulators like the Consumer Financial Protection Bureau in the US and the Financial Conduct Authority in the UK.

When a Google user enters a search term related to credit cards – for example, "best zero percent credit card" – websites that use fear-inducing terms like "risk", "penalty" or "charges" have a better chance of appearing nearer to the top of the results.

Sites that focus on benefits and advantages by putting more emphasis on added extras, bonuses or using words like "best" or "affordable", don't perform as well.

Sean O'Meara, a representative for Artios, said: "This is part of a wider study looking at how Google treats different market sectors. We took credit cards as an interesting example. We noticed that if the text on your page has an emphasis on risk over benefit, then Google will give you a little bump in ranking."

Credit card offerings must not have a promotional line that is more than one click away from the terms and conditions. If you are announcing zero percent interest for instance, that triggers what is called a "risk event", said O'Meara. In this case your promotion should be not more than one click away from the terms and conditions.

"Preferably these will appear on the same page, but obviously if you are looking on it on a mobile you can't display the terms and conditions on screen," O'Meara added.

"Google is doing something virtuous on behalf of its users that the finance industry probably needs to know about. It is going to inform how they maintain their websites going forward."

Using Natural Language Processing (relying on the NLP libraries), the study found sites using language that invokes moderate fear improves Google rankings for credit card websites by at least 0.5 positions on average, according to in-depth analysis of 100 credit card websites.

Andreas Voniatis, the data scientist at Artios who led the study, said: "This finding was perhaps the most interesting of the entire study. As an industry we've known that semantics play a large part in search performance, but this is potentially the first piece of evidence to suggest Google uses emotion of language to judge a site's usefulness.

"Google appears to be applying some degree of regulation by proxy on credit card sites. Industry regulators encourage businesses in the financial services sector to emphasise the risks of their products over the benefits and Google seems to be mirroring those requirements in its own algorithm, typically resulting in a higher degree of fear emotion in the copy in the best performing websites."

The study found that had the largest number of high rankings averaging position #6, on more than 70 keywords, followed by had the highest average ranking of 3.6 although they only represented 14% of the data set.

Voniatis said he used a support vector machine model to predict the Google Ranking for a given variation of a keyword, using all the other variables associated with each keyword.