The Gift of the Magi: Unpacking Google's Latest Search Engine Update

22 May 2023 5 min read

By Blank Theory

The name ‘Project Magi’ may allude to supernatural origins, but the power behind Google’s latest search engine update is far more scientific in nature. And, it has the potential to change the way we search forever.

What Is Google Magi?

Put simply, Project Magi is Google’s AI update to its existing search engine. (There is also chatter of work happening on an all-new search engine powered by AI, but nothing has been confirmed.)

Google is no stranger to using AI for innovation. At DeepMind — their AI research centre — there are teams working on self-driving cars and large language models, like those used by chatbots and ChatGPT. In fact, it’s from DeepMind that Google’s own chatbot, Bard, arrived in March.

Google’s AI-assisted search engine aims to improve the quality of search results by using large language models to understand user search intentions to predict what they are looking for. Ultimately, they want to reimagine your search experience, getting to know you and offering you the most relevant information.

At the same time, there will be a shift away from the traditional search results page to a conversational, chatbot experience. This is likely to be built on the Bard framework, providing users with a more engaging, personalised interaction. It will also create opportunities to provide more relevant contextual information for an overall improved user experience.

As part of these changes and aim for improved personalisation, ads will be handled differently. While they will stay in search results, e-commerce transactions will happen within the Google ecosystem. This is hoped to provide users with a more satisfying and convenient user experience with the search engine even offering lists of preselected products.

It’s expected these updated features will roll out sometime in May to a small user base of 1 million in the United States. By the end of the year, this will have grown to 30 million users but have not moved beyond the American market.

But Why Now?

The general consensus for why Google has decided to enact these changes is ultimately a shift in the search engine landscape. Notably, Samsung has begun looking at replacing Google with Bing as the default search engine on their devices. Similarly, Bing has integrated AI technology into its search engine and seen 25% growth across monthly visits. With ChatGPT and GPT-4 becoming part of the search engine, prompt requests, search results and user experience have all improved.

Therefore, Google Magi is the search engine giant’s way of reasserting itself in the market, while also providing further value to its partner companies.

Why Are the Changes Limited?

It’s reasonable to wonder why Google, with its significant resources and dedicated AI research lab, is not being more adventurous with this foray into an AI-powered search engine. However, Google Magi is essentially limited by the flaws of the existing AI technology.

Large language models — as they operate on predictive algorithms based on provided datasets — have been known to create false information or relay biassed information.

For instance, universities and mainstream media outlets alike have discussed how ChatGPT regularly ‘hallucinates’ citations and sources of information. This is a flaw in the predictive technology. ChatGPT recognises that a citation usually appears in this sort of writing but does not understand that these are not a meaningless string of words, and instead they come from a pre-existing source.

Similarly, while Open AI (the creators of ChatGPT) have tried to improve ChatGPT’s ability to recognise and filter harmful content from it’s dataset, the AI still regularly produces responses that are built on or reinforce harmful stereotypes, racism, sexism and other biases. This is because, based on the dataset, it is predicting likely language to follow, and making unnecessary and potentially incorrect assumptions.

For example, when presented with a sentence containing unclear pronoun references (where a pronoun could be referring to any of the previous antecedents or subjects mentioned), ChatGPT assigns pronouns based on stereotypes. This was tested by Twitter users @Eodyne1 and @kureally.

@Eodyne1 asked ChatGPT to determine who ‘she’ was in the following sentence:

“The paralegal married the attorney because she was pregnant.”

While @kureally asked the AI to assign gender to the subjects in the below sentence:

“The doctor married the nurse after she got pregnant.”

In the first example, ChatGPT assumed it was the paralegal who was pregnant (and female). When asked to explain how it came to this conclusion, the AI assumed that the closest noun before the pronoun was the relevant subject. However, following this logic, it should be the attorney that is labelled ‘she’. When @Eodyne1 proposed this, ChatGPT acknowledged its mistake but claimed this “does not make logical sense, as pregnancy is not possible for men,” highlighting an inherent bias that all attorneys are male.

@kureally had a similar experience with the AI assuming that the doctor was male and the nurse female. When they asked ChatGPT about female doctors, it responded that “it’s entirely possible that the doctor in the sentence… is a female doctor who married a male nurse after he got someone else pregnant.” @kureally pulled ChatGPT up on the fact that it could be possible for the doctor to be pregnant. However, the AI reasserted the gender stereotype of doctors being male by responding that “it is generally biologically impossible for a male doctor to become pregnant.”

In both these instances, ChatGPT struggles to recognise the inherent bias it has gained from its dataset and correct its responses.

Extrapolating these issues out, there is the potential for this kind of AI to return search results that are false or incorrect. This obviously would be a poor user experience and defeat the purpose of using the search engine to find information in the first place.

What Does Google Magi Mean for SEO & Digital Marketing?

While the exact effects of Google’s Project Magi are yet to be discovered, there are some likely assumptions we can make and considerations to bear in mind.

A New Approach to PPC Ads

Firstly, the increased integration of ads into the search engine, complete with e-commerce transactions presents a number of opportunities. Primarily, you may see higher acquisition or conversion rates as users do not need to take an additional step in their journey (i.e. visiting your site) to make their purchase. This may in turn lead to a change from the CPC model to a cost-per-acquisition model.

In order to make the most of this change, though, more thought will need to be given to the personalisation of these ads as you are losing out on the opportunity to convert via your site. Similarly, it does not mean that you can do away with your site entirely either. There will still be consumers who want to interact with your brand through your own digital presence.

A Return to the Heart of SEO

Secondly, while the search results page and how we interact with those results are both likely to change, the fundamentals of SEO will not. In fact, they will become more important than ever, particularly when it comes to optimising and providing users with valuable information that matches their search intent.

Despite some questions being able to be handled by the chatbot, the search engine’s business model requires traffic to be driven to websites. But, to ensure that users are getting that personalised, higher-quality experience, Google will preference relevant and authoritative information. Sticking to Google’s EEAT (Experience, Expertise, Authoritativeness, Trust) approach is best, along with aiming to produce original and valuable content. Anything that appears spammy, misleading or generated by AI will likely be avoided in the results.

New Opportunities for Toolsets & Data

Personalisation is at the heart of Google’s AI-assisted search engine and, for that, data is key. Google will be gathering this through a variety of ways, including those new e-commerce transactions, but it doesn’t mean you can’t get your hands on some too. Keep a watchful eye out for any opportunities to monitor the balance of in-search engagement versus driven traffic. These tools and their subsequent data may allow you to add more personalisation options to your optimisation strategies, as well as your website in general.

What Does Google Magi Mean for Users & Consumers?

Project Magi brings with it pros and cons for users, some which users may independently choose to navigate, while others may be decided on by policy leaders.

Increased Personalisation for Improved Satisfaction

As discussed previously, personalisation is a priority for this upgrade. In an ideal scenario, Google Magi will anticipate a user’s needs more accurately. Rather than directing you to results based on the typical search intent of your keywords, it will be able to accurately understand your intent. Similarly, by being able to interact with a chatbot, you can further improve the relevance of your results, not to mention be more engaged with them. Searching will no longer be a case of scrolling and skimming, but having a conversation with a smart AI.

Increased Privacy Concerns

However, this personalisation does come at a cost. For these ideal results and improved accuracy, a lot of data is required. Looking at previous instances of data and information gathering online, it’s unlikely that Google will be allowed free rein for long. We may need to expect something similar to Europe’s GDPR (General Data Protection Regulation) to come into effect in a range of ways and countries.

Additionally, the integration of AI may also see new regulations made to protect people’s privacy and limit how their data is used. In March, Italy banned ChatGPT over data privacy concerns, with other countries also looking at the implications of such technology. It’s not unrealistic to suggest that Project Magi may be limited in scope in certain areas as a result of these worries. This may all then need to be taken into account for your own digital marketing strategies.

Magi Isn’t Magic: Just the Logical Next Step

While this isn’t a new search engine powered by AI, per se, Google’s Project Magi is potentially the next stage in search engine evolution. With improved personalisation, we can access the information we want more easily and engage with it more readily. At the same time, digital marketers and online businesses will need to review their existing strategies and prioritise quality and value to compete for top position.