GOV.UK rolls out Google AI-powered semantic search

For the last few months, visitors using the search functionality of government’s website have been given results supported by Google, rather than in-house systems – and more new features may follow

Government has equipped GOV.UK with new generative artificial intelligence tools from Google intended to give users more accurate findings via semantic search capability.

Having implemented the Google Vertex platform several months ago, the Government Digital Service will now explore further upgrades, potentially including the autocomplete suggestions and improved snippets of content from search results, PublicTechnology can reveal.

Until last year, the search function on government’s website was powered by ElasticSearch software and a database hosted by GDS. In late 2022, the Cabinet Office unit revealed its intention to move from this proprietary model and adopt “fully managed search-engine services”.

GDS undertook an exercise – working alongside digital consultancy Kin + Carta – to evaluate the available search tools, including Google Programmable Search Engine, Microsoft’s Bing Custom Search API, AWS CloudSearch, Algolia, and Miso.

In November 2023, Google was appointed to a potential £2.4m contract to provide its Vertex AI search software to underpin GOV.UK search services. PublicTechnology understands that this functionality went live in February this year.

The search service that GOV.UK users are presented with does not yet appear any differently, but it is understood that the Google system provides results based on new semantic search capabilities. This refers to search engines that can surface results based on the inferred meaning of a of a user’s query, rather than lexical tools, which look solely for exact matches of word or terms entered by users.

GDS now plans to explore further additional upgrades, having signed another deal with Kin + Carta, in which the London-based firm will “support ongoing maintenance and improvement of its Google Vertex AI Search and Conversation-based search on GOV.UK”.

Google describes its Vertex Search service as a tool “which lets organisations set up Google search-quality, multimodal, multi-turn search applications powered by foundation models”, while the accompanying Conversation functionality “facilitates the creation of natural-sounding, human-like chatbots and voicebots, powered by foundation models with support for both audio and text”.

Related content

GDS’s agreement with Kin + Carta, which came into effect earlier this month and could be worth £555,000 over its one-year term, will see the supplier “provide a team to… [support] enabling and testing features that could make it easier for users to interact with search, [such as] autocomplete, snippets, [and] extractive answers”.

These examinations will use A/B testing – where visitors to websites are randomly provided with differing designs or functions – to ascertain the impact of potential changes.

If any resultant changes to GOV.UK’s user interface are implemented, the consultancy will work with GDS design professionals to deploy and maintain the updates.

“We will ensure knowledge transfer through pairing work, and one-to-one sessions with GOV.UK developers, and ad hoc knowledge sharing workshops with other teams,” the text of the contract says. “We will monitor the performance of the service and run automated tests to check results quality.”

The deal was awarded via the G-Cloud 13 framework, and according to service-definition documents provided by  Kin + Carta, the firm “designs, builds, and runs Google generative AI platforms using Vertex AI Search and Conversation, to help you create new generative AI-powered products and services”.

As GOV.UK no longer uses internal search services – and is increasingly experimenting with generative AI – it is understood that government believes it has “robust systems and processes” to ensure the site upholds high standards of privacy and security, including higher levels of encryption for some sections and services, and dedicated data-protection agreements with third-party providers.

Sam Trendall

Learn More →

Leave a Reply

Your email address will not be published. Required fields are marked *