We have written extensively about how AI is revolutionizing marketing in many ways, culminating in our State of AI in Influencer Marketing: A Comprehensive Benchmark Report. While OpenAI’s ChatGPT generated much of the initial excitement, it doesn’t monopolize the AI space. It did spur Google into increasing its AI investment, however, which has now resulted in the new Google Search Generative Experience (SGE)
Google’s search engine has always been that company’s (now Alphabet’s) premier product. Indeed, it has long been the most common way for people to find things online. The last thing Google needed was a serious competitor in search, and ChatGPT was showing real potential for revolutionizing the search industry. For example, an opinion piece in the South China Morning Post in March 2023 was headed, Move Over Google? ChatGPT And its Like Will Change How We Search Online.
Interestingly, the author of that article, Dr. Yanto Chandra, Associate Professor at City University of Hong Kong, concluded that “search engine optimization or paid searches, ubiquitous on search engines today, will decline or have to evolve”, however, while “ChatGPT delivers results in paragraphs that read well, making it more user-friendly, … experts who can craft complex prompts will have an edge”.
Undoubtedly Alphabet was already developing Google Search Generative Experience (SGE) before OpenAI released ChatGPT, but there must have been much consternation within Googleplex Mountain View, California until Google AI’s product was ready for public testing. At the time of writing, Google Search Generative Experience has yet to go mainstream, however, you can apply to join Google Labs’ waitlist. Google currently restricts participants to a limited number of people in the US and English-language only.
What is Google Search Generative Experience (SGE)?:
The Google Search Engine
For most of its existence, Google has used a process of web crawling to power its flagship search engine. It has the dual distinction of being the most used search engine and the most “searched for” search engine (although Google-owned “YouTube” is the most searched for keyword overall, “Google” comes third overall globally).
Traditionally, Google has delivered search results using what it calls the "PageRank" system. Over time, however, Google has provided many different options for customized searches.
Larry Page, Sergey Brin, and Scott Hassan initially released Google Search in 1996, to search for text in publicly accessible documents offered by web servers. It initially excluded other data, such as images or data contained in databases.
Since then, however, Google has expanded its search engine’s remit, adding many of the features we take for granted today. A major advance came in 2012 when Google introduced a Knowledge Graph semantic search feature. By 2016, Google's search engine began to rely heavily on deep neural networks.
What’s so Important About Generative AI?
Both Google Search Generative Experience and ChatGPT are examples of Generative AI. Generative AI can create a range of content, such as text, imagery, audio, and synthetic data, often with simple intuitive commands in almost real-time.
Although you may think of Generative AI as new, it has been around since the 1960s, when it was used in primitive forms of chatbots. However, it has only been since 2014 that generative AI could create convincingly authentic multimedia assets representing real people. Of course, these capabilities have not always been directed in ways that benefit people – they are the same technologies used to create deep fakes and cyberattacks.
Generative AI requires a prompt, perhaps text, an image, a video, an audio clip, or indeed any input the AI system can analyze. It takes the prompt and applies AI algorithms to create a useful output.
Early versions of Generative AI required that you input data using a complicated process, either using an API or programming languages. This was impractical for a mainstream tool, such as Google’s search engine. However, recent advances in natural language processing have made it possible for people to ask questions in everyday language and use non-text input.
Generative AIs are now self-learning, too. If their first response to your request doesn’t give the results you hope for, you can add clarifying questions and additional stimuli. Modern generative AI can learn to adapt its answers to allow for the additional information you’ve provided.
Google’s Initial AI Claims Were Met with Skepticism
You might wonder how a technology company as market-leading and powerful as Alphabet (Google) is playing catch-up with AI. How did they let ChatGPT become the public face of Generative AI?
Well, it wasn’t always like that. Back in 2016, Google’s recently appointed CEO, Sundar Pichai, proclaimed that Google would now be an “AI-first” company.
Admittedly, that was only two years after Amazon had outflanked Google by releasing its voice assistant, Alexa, but by 2016 Google was ready to release Google Assistant to the world.
But what must have really infuriated Google about all the publicity ChatGPT generated seven years later, however, is that the search giant had already announced an AI/conversational language model in 2021, Google LaMDA (originally developed and introduced as Meena in 2020). However, in June 2022, Google engineer Blake Lemoine claimed that the new chatbot had become sentient, leading to most of the scientific community largely rejecting Lemoine's claims. LaMDA was trained using human dialogue and stories, allowing it to engage in open-ended conversations.
Lemoine based his assertion on responses LaMDA made to questions regarding self-identity, moral values, religion, and Isaac Asimov's Three Laws of Robotics.
Google, however, refuted Lemoine’s claims, placing the engineer on paid administrative leave. This only led to Lemoine claiming that LaMDA was "a person" as dictated by the Thirteenth Amendment to the U.S. Constitution, comparing it to an "alien intelligence of terrestrial origin". He hired an attorney on LaMDA's behalf after the chatbot requested that Lemoine do so. On July 22, 2022, Google fired Lemoine, asserting that Blake had violated their policies.
While Google was facing all this soap opera-level drama with its AI product, OpenAI happily developed ChatGPT without any notable controversy. At the same time, the internal controversy caused by Lemoine’s claims led to Google executives getting cold feet and deciding against releasing LaMDA to the public.
Google Bard
OpenAI’s release of ChatGPT caused Google to issue a "Code Red" alert because they considered that it could have a massive negative impact on their search engine.
In response to ChatGPT Google announced Bard, a conversational generative artificial intelligence chatbot, in March 2023. According to the New York Times:
“For Google, this was akin to pulling the fire alarm. Some fear the company may be approaching a moment that the biggest Silicon Valley outfits dread — the arrival of an enormous technological change that could upend the business.”
CEO Pichai called in founders Page and Brin, and they made it clear that addressing Google’s own artificial intelligence capabilities was now the company’s top priority, regardless of the past problems.
Google’s first major AI release post-ChatGPT was an AI chatbot, Google Bard, which they released on February 6, 2023. It functions similarly to ChatGPT; however, Google's service pulls its information from the web. Bard can code, answer math problems, and help with your writing needs. Bard was initially powered by a lightweight version of LaMDA, making it the first public application of that technology, although it is now powered by PaLM2, Google’s own and most advanced large language model.
Notably, Google’s decision to use its own products in Bard, LaMDA, and PaLM2, differs from most of the opposition (including ChatGPT) who use a language model in the GPT series. Google opened its waitlist for Bard on March 21, 2023, granting access to limited numbers of users in the US and UK on a rolling basis.
Unfortunately for Google, Bard delivered inaccurate information about the James Webb Space Telescope during its initial public demo. Google tweeted an example of Bard in action, which was quickly refuted by experts on the subject:
Bard is an experimental conversational AI service, powered by LaMDA. Built using our large language models and drawing on information from the web, it’s a launchpad for curiosity and can help simplify complex topics → https://t.co/fSp531xKy3 pic.twitter.com/JecHXVmt8l
— Google (@Google) February 6, 2023
As a result of this goof, Alphabet Inc (GOOGL.O) lost $100 billion in market value in one day. Reuters reported Gil Luria, senior software analyst at D.A. Davidson, as saying, "While Google has been a leader in AI innovation over the last several years, they seemed to have fallen asleep on implementing this technology into their search product. Google has been scrambling over the last few weeks to catch up on Search and that caused the announcement yesterday (Tuesday) to be rushed and the embarrassing mess up of posting a wrong answer during their demo."
Although Bard is primarily text-based, in late May 2023 Google updated it to include images where relevant in its answers. It pulls these images from the existing Images section of Google Search.
Google Planning New Search Engine Under Project Magi
Despite its issues with LaMDA and Bard, Google hasn’t given up on AI. It can’t afford to if it wants to keep its traditional position at the top of the search ladder.
As a result of its Code Red alert, Google set up Project Magi to build an “all-new search engine” powered by new A.I. technology, as well as adding new A.I.-based features to its current search engine. Google hopes its new search engine will offer users a far more personalized experience than its current service, attempting to anticipate users’ needs.
Google has announced that Project Magi should contribute to its products in various ways, including:
- helping searchers to complete transactions, such as buying shoes or booking flights (all while still incorporating Google search ads)
- combining Google Earth’s mapping technology with AI
- using AI to generate images in Google Image results
- teaching users a new language through open-ended AI text conversations in Tivoli Tutor
- letting users ask a chatbot questions while surfing the web through Chrome and Searchalong.
What is Google Search Generative Experience (SGE)?
Google Search Generative Experience (SGE) is one of the first outputs of Project Magi to be made available to (some of) the public. It is important to remember that SGE is still experimental, and certainly isn’t ready to replace Google’s existing search engine yet.
If you do qualify to experiment with Google Search Generative Experience, you will find it in Google Chrome (if using a desktop computer) or in the Google app for iOS and Android users. Google sends an email to successful users, along with a notification from the mobile apps welcoming them to Google Search Labs.
Once you have SGE enabled, you will see an additional AI conversation box beneath the main Google search bar when using the Desktop version. On mobile, you’ll see an additional button marked “Converse” beneath the search bar.
Google sums up the current situation with Generative AI by the statement, “You’ll see an AI-powered snapshot of key information to consider, with links to dig deeper.” Google hopes that by adding Generative AI to search, “you’ll be able to understand a topic faster, uncover new viewpoints and insights, and get things done more easily.”
Google provides an example of how searching with Generative AI can simplify a complex search. Suppose you want to know, “What's better for a family with kids under 3 and a dog, Bryce Canyon or arches?” Traditionally you would have had to split this into multiple mini-searches and then combine what you have learned from each, to find the complete answer to your question. However, Generative AI can do much of the behind-the-scenes analysis for you, including making the comparison.
Source: storage.googleapis.com
In addition to it giving you a snapshot answer, Generative AI in search suggests possible next steps you could take. This includes providing additional follow-up questions, in conversational English, so it can further refine the answer it suggests.
How Does Google SGE Alter Your Search Results?
You may have noticed how dramatically Google search results have changed over the years. The days when you merely had an ordered series of listings are well gone.
Even without any thought of Generative AI, there are now multiple sections to Google Search, and these vary depending on whether you’re searching on desktop or mobile devices. Inevitably, Google SGE complicates things even further.
Search Engine Journal tested the placement of Google SGE in detail in each of the following situations:
SGE For SERPs with FAQ, People Also Ask, and Knowledge Panel Features - Google SGE crafts a multi-paragraph answer using multiple sources to answer the question (in Search Engine Journal’s example search it generated a three-paragraph answer, from six sources, and then asked a series of qualifying additional questions. You can toggle a viewing option at the top right of the AI result to see which sources Google SGE used to craft the summary. The standard Google search results appear beneath the AI-generated results, with organic search results, People Also Ask, and a Knowledge Panel. Notably, the sources chosen by SGE do not always match the top organic search results.
SGE For SERPs With the Local Pack Feature – again, the generative AI results appear before the Local Pack and the rest of the traditional search results. In this case, however, Google’s Generative AI suggestions look similar to the Local Pack feature, although don’t necessarily contain the same suggestions.
SGE For SERPs With Sitelinks – Search Engine Journal searched for a site it knew would come up with site links, People Also Ask, and a knowledge panel in a traditional search. In this case, Google asked, “Get an AI-powered overview for this search?” Upon affirmation of this, Google SGE generated a multi-paragraph answer about the site in question, following it up with the usual site links, People Also Ask, and a knowledge panel.
SGE For SERPs With Shopping and Reviews – here, Google SGE generates advice and multiple suggestions about possible products you could purchase to meet the intent of your search, before showing organic search results.
SGE For SERPs With Product Details and Store Listings - Search Engine Journal found that if you inquire about specific features of a product, SGE will give you precise details, reviews, and suggested purchasing options, compared to traditional search which focuses on content from the brand, followed by People Also Ask and organic search results.
SGE For SERPs With Sponsored Ads and Featured Snippets – again, Google SGE begins by building an answer from multiple sources, before moving into the organic search results. This compares to traditional search where Google SERPs typically begin with several sponsored results from advertisers, followed by a featured snippet, People Also Ask, and organic search results. It should be noted that although Sponsored Ads don’t feature prominently at the moment for those trialing Google SGE, this is likely to change before Google makes the Generative AI product available to everybody.
SGE For News and Top Stories – if you’re searching for the latest news on a topic, SGE pulls from the most recent relevant news sources to come up with a series of bullet points highlighting what’s new. It then suggests clarifying questions you could use if you want it to go into more depth about a particular news item. The bullet points listed don’t always match the news stories that traditional Google news searches highlight.
SGE For SERPs With Images - Search Engine Journal found you need to be quite specific if you’re looking for particular types of images. For example, when searching for “social media logos'' Google SGE created a series of paragraphs describing social media logos taken from articles describing social media logos, rather than coming up with visual images of the requested logos. They did include the standard image results further down the page, however.
SGE For SERPs With Video – Google’s results here appear to be highly dependent on the question you ask. In Search Engine Journal’s example, they asked how to repair metal glasses frames. The generative AI took advice from a range of sources to come up with a text-based answer but followed this up with the same relevant videos (split up into sections) demonstrating how to do the task that they offer for traditional video searches.
SGE For SERPs With Events – if you ask Google SGE for upcoming events in your area, it will create a lengthy list of events from multiple sources. It follows this with Google’s normal calendar view of upcoming events, organic results, etc.
SGE With Code Tips – some searches can result in code. Normally these types of Google searches simply give organic search results, suggesting suitable sites to find the code. Google SGE, however, suggests suitable lines of code to meet the intent of your search.
SGE For Unavailable Queries – Google SGE cannot currently answer all possible search questions. In cases where it can’t provide a result, it delivers the message, “An AI-powered overview is not available for this search.”
Wrapping Things Up
It’s important to remember that Google Search Generative Experience is an add-on to Google Search, not a replacement for traditional search methods. While plans for a completely new search engine are part of Project Magi, that doesn’t mean that today’s successful content creators need to worry that their work will become redundant.
AI can produce excellent results for some types of searches, particularly cases that would have required multiple searches in the past to obtain a complete answer to a question. However, most users of Google don’t intend to use it to replace their web browsing. They will still want to go to sites with quality content and in-depth analysis to obtain much of the information they want as well as for other purposes such as entertainment.
Search Engine Land reports about leaked papers that suggest Google has an aim of making their search engine more “visual, snackable, personal, and human”. To that end, our first impression is that Google Search Generative Experience could be a success. Google continues to move further away from its traditional “ten blue links”, creating a layout it feels is more appealing to younger searchers.
Of course, the move to incorporating AI does open up the question of what now counts as “Trusted Content”, particularly in cases where there aren’t objective facts leading to a single “right” answer.
