Google Releases Topics API Into the Wild: What You Need to Know

Tim Cross 17 July, 2023 

Google’s deadline for phasing out third-party cookies from its Chrome browser remains set for Q3 of next year (for the time being). But Google’s aim is that by the time this deadline hits, advertisers, publishers, and ad tech companies will all already be familiar with the alternative tools created within its Privacy Sandbox. And on Wednesday Topics API, one of the most relevant Sandbox tools for the advertising world, started rolling out to users broadly via the release of version 115 of Google Chrome.

Topics had already been available for testing as part of Google’s ‘Origin’ trials, which it’s conducting for most of its Sandbox APIs. These trials saw Topics deployed across a small portion of Chrome users. But as Chrome 115 rolls out across Google’s user base, Topics will become available much more widely – and can be used as a legitimate way to target ads on Chrome devices.

Still a work in progress

In simple terms, Topics analyses Chrome users’ web-browsing behaviour, looking at the types of content and websites which each user is visiting. Instead of remembering exactly which websites and pages each user visited, Chrome will maintain a list of topics that the user seems to be interested in (hence the name ‘Topics’). A few of these topics will then be presented to ad tech tools and advertisers whenever an ad impression is available for that user, helping to inform ad targeting.

To do this, Google has created a finite list of topics and sub-topics which can be used for ad targeting. This list is fairly long – there are currently 470 topics within Google’s taxonomy. While some of these topics are broad, others are quite specific – for example one currently available topic is ‘/Games/Computer & Video Games/Driving & Racing Games’. Many of these topics have also been chosen specifically to relate back to brand categories, making them easier for advertisers to pick from (like topic 499, ‘/Home & Garden/Home Storage & Shelving/Cabinetry’). This list isn’t final – Google has said it wants the taxonomy to become a resource maintained by trusted parties, and says the final number could be between a few hundred and a few thousand.

Whenever a user visits a website, it chooses the topic most relevant for that site, and logs that topic within the browser. To choose the most relevant topic, Google uses machine learning to analyse the hostname and pick an appropriate topic. For the topic 10,000 most popular domains, these ML-chosen topics can be overridden by manual curation.

Through this process, a Chrome user’s browser will collect a list of topics which that user seems to be interested in over a period of time. By seeing which topics appear most frequently, Chrome can then deduce which topics each user is most interested in. This list is refreshed over time – each week a new list of topics is saved for each user.

Then when an ad is shown to that user, Chrome can make topics available to those who ask for them – though any company calling the Topics API will only be shown topics which they themselves have already observed. This last aspect is designed to ensure Topics doesn’t spread information more freely than third-party cookies. With cookies, an ad tech company would have to be present on each website and drop a cookie there in order to be able to recall later on that the user had visited that website earlier.

A contentious topic

The fact that Topics API is being made available widely within Chrome doesn’t mean work on the tool is complete. Google says that for all of its Sandbox tools, it expects to make refinements and optimisations over time as more companies test and use the APIs.

But with Topics now out in the wild, it looks like those who had been holding their breath for a comprehensive rework of Google’s new targeting tool are set to be disappointed. As soon as Topics was announced, some in the industry expressed concerns about Topics’ usefulness for advertisers, as well as scepticism around its privacy credentials.

There have been suggestions that Topics could be used by bad actors for fingerprinting. Google itself has acknowledged the risk here, and for the time being its solution is to ask all developers which enroll to use the API to attest that they won’t abuse it (which doesn’t sound like a watertight solution).

And earlier this year the World Wide Web Consortium (W3C), a body which creates standards for the open internet, publicly rejected Topics API. The W3C said that Topics enables “inappropriate surveillance”, calling on Google to cease development of the project.

Google – with its self-imposed deadline approaching, and having already pushed back this deadline twice already – decided to push on with the project. But the tech giant had originally hoped that other web browsers might implement the API themselves, making it more useful to advertisers by allowing them to use the same methodology across different browsers. This has not turned out to be the case.

But Google will care most whether publishers, advertisers, and ad tech companies themselves take to Topics, and will likely be watching closely how usage of the API picks up now it’s widely available.

Follow VideoWeek on Twitter and LinkedIn.


About the Author:

Tim Cross is Assistant Editor at VideoWeek.
Go to Top