Matthew Bartlett and Geoffrey Roberts
We want people to get reliable, plain-language answers as quickly as possible when they have have legal questions. This vision prompted us at Citizen AI to develop a series of legal help chatbots and is now pushing us to explore search engine optimisation for online legal information. Here we share what we’ve learned so far, because we think it may be helpful to others working to improve access to justice for ordinary people. Warning: jargon ahead!
We know that most people, when they feel unwell, use Google to self-diagnose.* We can make an educated guess that when they have legal problems they largely also turn to Google first — before, and often instead of, seeking professional help.
Google dominates search, with over 92 per cent of market share worldwide, so we have focused our efforts to date on that platform.
As you probably will have noticed, in recent years Google has added many extra features to the basic search engine result page (‘SERP’ in the Search Engine Optimisation / SEO world). Depending on what you search for, you might see a carousel of product images, a map, ads, or some mixture of over 30 other possible special features. We think perhaps 10–15 of those features have potential use in improving access to justice, but in this article we’ll concentrate on just one.
Many of the special SERP features show up above all of the regular search results. In fact in the featured image screenshot you can make out that I had to scroll quite a way down the screen before I got to a vanilla or garden variety search result. That top-most spot is known in the trade as ‘position zero’. On mobile screens, in particular, it forms a big proportion of what you first see when you search.
The information that shows up in position zero is also often used by Google’s voice search features (on, for instance, Google Home speakers, or when people say ‘OK Google’ to Android phones). Voice search is a trend worth paying attention to for potential access to justice possibilities: already roughly half of people (in the USA at least) use it, largely via their smartphones.
Google is getting so good at answering people’s questions that about half of searchers get their answer right there on Google, without clicking through to any other website. The proportion is even higher for mobile devices. This is a difficult reality to grapple with if you’re trying to sell a product, but if your goal is to answer people’s legal questions for free, as ours is, it’s a solution, not a problem.
Depending on the nature of the content, you can sometimes even explore further, still right there within the SERP. Above is a series of screenshots showing how a whole collection of related topics can be interacted with, including links to the source legislation.
Method and results
Our parent organisation, Community Law Wellington & Hutt Valley, publishes a comprehensive plain-english guide to the law, the Community Law Manual, each year in print and online. They gave us permission to make adjustments to the way the online information is presented, so that Google will be more likely to use it for featured snippets.
It was not particularly technically complicated to do this. Google and other search engines set up an organisation and website Schema.org, which develops and hosts schemata to let people structure all sorts of content on the web, making them more easily searchable. Examples of schema include: movie, person, local business, product, recipe, etc. etc. For our purposes we’ve started with the FAQPage schema to describe each page of the online Community Law Manual. Within each page we use Question and acceptedAnswer to make clear to search engines which headings and paragraphs are questions and answers.
Google is already quite good at figuring out what sort of content is on each web page, as it tries to get ever better at serving up content that matches the users’ intents. But marking up the content in this way makes it that much more likely that it will be selected by the algorithm for position zero, and brings that much more rich content to the people who could benefit from it.
Since we’ve begun this sort of experimentation on the Community Law Manual content, we’ve seen visits to the site increase by almost 40 per cent. This won’t necessarily be entirely due to our efforts, as many factors will influence the site’s changing popularity, such as Google’s regular updates to its core algorithm, and events in the real world (like Covid-19). We can be more confident that the schema-based annotations of the content have dramatically increased the number of people getting answers to their legal questions from the Law Manual content: we see roughly 100 per cent extra people accessing that content in the ‘zero click’ mode.**
In a typical month before the changes, we saw around 63,000 visitors; now around 88,000 visitors plus perhaps 68,000 zero-clickers.
We had a big head start with this particular website: it was already very popular (within the New Zealand context), and well-established. For a much newer and more niche legal information site we tested some of these ideas out on earlier, we saw a more modest ~ 30 per cent increase.
Getting the right information is of course only one piece of the access to justice puzzle; and not everyone turns to Google in their moment of need (though 97 per cent of New Zealanders under 65 describe themselves as internet users). Jurisdictional specifics will continue to be a thorny issue — as opposed to, for example, medical information, which generally holds true wherever one is.
On the technical side, Google will continue to roll out more advanced AI search algorithms like the recently released BERT. Google itself says that the algorithm represents ‘the biggest leap forward in the past five years, and one of the biggest leaps forward in the history of Search.’ BERT greatly improves Google’s understanding of ‘long tail’ queries where the word order would have previously confused the search engine.
Spending time trying to better understand and apply Google’s own search guidelines is also on our radar. Google has determined that legal information needs to meet higher quality standards in order to rank in search results and they are getting more fastidious in ensuring only the most authoritative content is served to users.
Finally, being in the business of legal information chatbots we can’t help but see the end game for Google as a conversational search experience. The omniscient librarian is coming to a smartphone near you.
Matthew Bartlett and Geoffrey Roberts, Citizen AI, Wellington, New Zealand
** There are assumptions buried in this number. We’ve arrived at it through this rough-and-ready formula: For FAQ rich results over a certain period from Google Search Console: (search impressions ÷ average position on SERP) – clicks through to website. The rationale is that if a rich result turns up at the top of the page, it’s very likely the person will read it; if it’s in the second position, it’s half as likely; in the third position; one third as likely, etc. And we deduct clicks so as to not double-count those interactions on top of regular website visits.