About this post:
Google’s Search Generative Experience is one way that GenAI is being integrated into the tools that we already use. The current version gives added functionality and value to user queries but has limited value for marketeers with a focus on Google Shopping products and issues with attribution of the source information on which the generative result is based. If implemented in its current design, it is likely that the traffic and overall value of high organic rankings will be significantly reduced while increasing the impact of Programmatic, Display and Retargeting campaigns.
Want to find out more?
Access our AI white paper
What is Search Generative Experience?
SGE (Search Generative Experience) is a new experimental feature from Google that leverages AI to analyse search results and create informative summaries. These summaries, appearing as "AI-powered snapshots", offer a quick grasp of complex topics and various perspectives on them, aiming to enhance the user’s search experience. SGE is currently in beta and available through Google Search Labs, and has been live in the USA since May 2023. Google’s latest announcement launches SGE to some accounts in 120 countries and 7 languages as part of the Search Labs. Search Generative Experience promises faster understanding and deeper exploration of topics, but what is it, and what is its impact going to be on Search and your conversion funnel.
So how will it impact businesses? Although it was only launched, in classic Google style as a “beta test”, in May 2023, it is already available in the vast majority of searches. According to a November 2023 study by Content Detector 77.8% of all Google searches have a SGE, AI generated result, and that had grown to 86.8% in Authoritas’s January 2024 Keyword Study. If you have a presence in organic search, your site will be affected by SGE, although a closer look at the data suggests that, at least at present, what sector your business is in will affect the extent to which SGE impacts it.
For Fashion and Clothing, virtually all of the available keywords show SGE results and Technology and Real Estate are not far behind, both with over 90% of keywords showing AI-generated results but in Finance and Sport, only around half of all results have SGE components.
In many cases, the SGE results replace many of the knowledge graph elements that were previously prevalent. Instant answers, People Also Ask, local pack results and FAQs are all integrated into the SGE generative result, so there is still value in optimising your content for these types of query but there is less clarity on how the information from your site is integrated and your website may be a partial source for the information but may not be included in the generative trio of links.
Visually, SGE pushes traditional organic results much further down the results page than they have been, with both the “show more” and “generate” buttons pushing results down the page more than 1.5 times the desktop viewport.
In the screenshots below, for the search “nike shoes” the left result is without SGE and the right result is with SGE enabled.
Without SGE enabled, there are:
With SGE enabled, there are:
Although there are the same number of non-brand URLs in the same space, the SGE results give the brand significantly more visibility while also giving more slots for retailers in the Shopping carousels.
If we just look at the traditional “blue links” however, the difference is more stark; there are half the number of links in the same space but the Brand links are much more dominant.
One thing to note is that, while SGE gives 3 non-branded links as sources for the generated text, these are, currently, rarely the same links as the top ranking organic results without SGE. According to Authoritas’s study, only 4.5% of generative URLs directly matched an organic URL from Page 1 and only 1.6% of the URLs were from the same domain as a Page 1 organic URL.
At the moment there appears to be few traditional PPC ads on SGE results pages, with the main source of paid results being Google Shopping results. At this early stage, this is understandable; Google Shopping has a stream of definitive data that can be linked directly to the specific products that are being searched for, allowing Google to retain quality control and avoid the placement of adverts alongside generated content that the advertising brands might not appreciate. A variation on this can be seen with queries that are contentious or offensive, when the note “An AI-powered overview is not available for this search” is displayed instead of the results. However, with Paid ads being at the core of Google’s business, this is not a state of affairs that we anticipate will continue if the Search Generative Experience version of Google becomes the default.
In 2023, Google's ad revenue was $236.87 billion, which is 78% of its total revenue and although not all of that is via organic and paid search channels, they are not going to fundamentally change their core product without ensuring that it retains the same revenue potential as it does at present. SGE is in very early stages of development and Google is being careful to minimise the possibility of controversy as they try and navigate the changes.
How does Search Generative Experience work?
The way that Search Generative Experience (SGE) works is quite interesting, combining different technologies to understand and condense information.
Retrieving Relevant Information:
When you enter a search query, SGE first uses traditional search algorithms to identify relevant web pages and documents. It then employs a technique called Retrieval-Augmented Generation (RAG). RAG essentially acts like a research assistant, sifting through these retrieved documents and selecting the most pertinent passages based on your query.
Understanding and Processing:
The chosen passages are fed into a large language model (LLM). LLMs are like advanced AI that can understand and process language nuances. The LLM analyses the content, paying close attention to key points, factual information, and different perspectives presented in the passages.
Generating the Summary:
Based on its understanding, the LLM generates a concise and informative summary of the topic. This summary highlights the main points, incorporates diverse viewpoints, and aims to answer your query directly. The summary is presented as an "AI-powered snapshot" at the top of the search results page, allowing you to grasp the core information quickly.
SGE doesn't stop at the summary. It can also identify relevant links and resources within the retrieved documents. These are presented alongside the summary, offering you options for deeper exploration if needed.
SGE aims to bridge the gap between traditional search results and directly answering your questions in a digestible format. It's still under development, but it represents an exciting step forward in how we interact with information online.
As with many innovations in search, Google’s patents are a rich source of information, although in the case of SGE, there only seems to be a single patent; “Generative summaries for search results”, which Google applied for on March 20th 2023, which describes how to selectively use a Large Language Model to generate a natural language summary to a user query.
This patent details how to “selectively use a large language model (LLM) to generate a natural language summary in response to a user query,” and highlights clearly how the accuracy and comprehensiveness of the summaries depend on the quality and diversity of the retrieved documents.
How is Search Generative Experience different from Bard or ChatGPT?
Search Generative Experience (SGE), Bard(now called Gemini), and ChatGPT are all interesting language models, but they have distinct purposes and functionalities:
Search Generative Experience (SGE):
Bard/ Gemini
ChatGPT:
To put a more human face on it, SGE is like a research assistant summarising complex search results, Bard/Gemini is like an information assistant answering your questions comprehensively and ChatGPT is like a creative partner engaging in conversation and generating text formats.
How does Search Generative Experience benefit brands and websites?
Search Generative Experience (SGE) has the potential to benefit brands and websites in several ways, but it's still in its early stages, so the long-term impact is yet to be fully understood. Here are some potential benefits and considerations:
Benefits:
Risks:
What are the negatives of Search Generative Experience for brands and websites?
While Search Generative Experience (SGE) offers potential benefits for brands and websites, it also comes with some potential downsides to consider:
Reduced organic traffic: SGE aims to provide concise answers directly on the search engine results page (SERP), potentially reducing the need for users to visit your website. This could lead to decreased organic traffic, especially for longer or more in-depth content.
Loss of control over brand messaging: The summaries generated by SGE might not perfectly capture the nuances or full context of your content, potentially misrepresenting your brand message or omitting important information. You have limited control over how your website is presented in SGE summaries.
Over Reliance on AI: While AI-powered summaries can be helpful, they are not perfect. Over reliance on SGE for attracting visitors could lead to neglecting traditional SEO optimisation and content creation strategies that ensure your website remains visible and relevant over time.
Competition for snippets: The space available for SGE summaries is limited, making it competitive for brands to get their content featured. This might favour larger, well-established brands with significant online presence.
Potential for bias: As with any AI system, SGE could potentially introduce biases based on the data it is trained on. This could disadvantage websites or brands targeting specific demographics or niche audiences.
Technical challenges: Optimising content for SGE might require technical expertise and resources, which could be a hurdle for smaller businesses or those with limited technical capabilities.
Uncertain long-term impact: SGE is still in its early stages, and its long-term impact on search behaviour and brand visibility is unknown. Adapting strategies based on a constantly evolving landscape can be challenging for businesses.
Ethical considerations: The use of AI-powered summaries raises ethical concerns around transparency, accountability, and potential manipulation of search results. These concerns need to be addressed to ensure responsible development and usage of SGE.
Overall, while SGE presents exciting possibilities, brands and websites need to carefully consider the potential downsides and be prepared to adapt their strategies as the technology evolves.
Impact of Search Generative Experience on Paid search
The arrival of Search Generative Experience (SGE) marks a significant shift in how users interact with search results. This AI-powered tool promises to deliver informative summaries directly on the search engine results page (SERP), potentially impacting how people engage with paid search advertising. While the full picture is still emerging, both potential drawbacks and exciting possibilities exist for advertisers navigating this new landscape.
On the one hand, SGE might present some challenges for paid search. If users find comprehensive answers within the summaries, they might be less likely to click on paid ads, leading to decreased click-through rates (CTRs) and conversions. Advertisers might need to shift their focus towards broader upper-funnel strategies like display and video advertising, while also adapting their keyword-based search campaigns to compete for visibility in a potentially crowded SERP environment. Additionally, the effectiveness of traditional paid search formats within the SGE ecosystem remains uncertain, requiring experimentation and adjustment in targeting and messaging.
SGE also offers potential benefits for paid search. The richer user data and insights generated by SGE could enable more targeted and personalised ad campaigns, ultimately leading to improved relevance and performance. SGE might pave the way for innovative ad formats that seamlessly integrate with search results, creating more engaging and impactful advertising experiences. Importantly, paid search can still play a crucial role in attracting users actively seeking specific products, services, or brands, even if SGE provides general information summaries.
The impact of SGE on paid search is unclear at this early stage and as it is expanded to more markets and more languages, our understanding will improve. While challenges exist, such as reduced reliance on ads and competition for visibility, the potential for enhanced targeting, new ad formats, and a complementary approach to user acquisition creates exciting opportunities for advertisers.
The arrival of Shopping through Search Generative Experience (SGE) promises to shake up the online shopping landscape, potentially transforming how we discover, research, and ultimately purchase products. This AI-powered tool, still in its early stages, has the potential to unlock exciting new possibilities for both shoppers and businesses.
Imagine searching for "cosy winter outfits" and seeing not just a list of products, but an AI-generated mood board showcasing trending styles, curated based on your location's weather and even your past browsing history. SGE could become your personal shopping assistant, not just finding clothes that match your search terms, but sparking inspiration with visually engaging suggestions you might not have considered before.
The potential of SGE goes beyond mere discovery. It could personalise your shopping experience like never before. Imagine searching for "running shoes" and seeing options filtered not just by brand and price, but also by your preferred shoe type and even foot characteristics. This personalised approach could save you time and effort by highlighting products that truly fit your needs and preferences.
No more wading through endless product descriptions and reviews. SGE would summarise key information like features, ratings, and even price comparisons within the search results, making research a breeze. This could be especially valuable for complex products where understanding specifications is crucial.
SGE isn't just about finding the perfect product; it's also about making informed decisions. Imagine seeing a summary alongside a product that not only highlights its strengths but also flags potential drawbacks or common complaints from real users. This transparency could empower prospects to make confident purchases, avoiding products with known issues and ensuring they get the most value for money.
It's important to acknowledge potential challenges. Accuracy and bias in AI-generated summaries could lead to misleading information or unfair promotions for certain products. Over-reliance on SGE might limit exploration and hinder discovery of smaller, niche brands that deserve a chance. Ethical considerations regarding data privacy and potential manipulation of shopping behaviour through AI need to be carefully addressed.
Shopping through Search Generative Experience presents a fascinating glimpse into the future of online retail. It holds the potential to make shopping more personalised, efficient, and even inspiring. It's crucial to approach this technology with a critical eye, ensuring that its implementation promotes fairness, transparency, and a beneficial experience for all. As SGE evolves, it will be fascinating to see how it shapes the way we shop online and how it ultimately impacts the retail landscape.
Limitations and Risks of Search Generative Experience
Search Generative Experience (SGE) promises to revolutionise how we interact with information online, but it's vital to recognize its limitations and potential risks before diving headfirst into its AI-powered summaries. While exciting possibilities abound, we must tread carefully to ensure responsible development and ethical use.
One major concern lies in accuracy and reliability. At this very early stage, SGE’s ability to generate flawless summaries isn't guaranteed. Errors, omissions, and misleading information could occur, especially for complex or subjective topics. The inner workings of SGE are fairly opaque, making it difficult to understand how summaries are created and why certain information is included or excluded. This lack of transparency raises concerns about potential biases and the ability to hold the system accountable for its outputs.
AI systems trained on vast datasets can inherit and amplify existing biases, leading to the potential for unfair or discriminatory outcomes in SGE summaries. Certain perspectives or viewpoints may be unfairly favoured, silencing more diverse voices and creating an incomplete picture of reality. The concise nature of summaries could make it difficult for users to discern fact from fiction, a problem that already exists to a significant extent in social media, and it is easy to see how there is the potential for solutions like SGE to exacerbate this.
Limited control and user agency also warrant consideration. Users have minimal control over the information presented in SGE summaries, potentially hindering their ability to critically evaluate and explore diverse perspectives. Over reliance on these summaries could stifle independent research and critical thinking skills, making us passive consumers of information rather than active seekers of truth.
SGE could significantly impact the search landscape, potentially giving undue prominence to certain websites or information sources featured in its summaries. This could disadvantage smaller websites and niche information that might not be easily captured by the system, creating an uneven playing field. Valuable information could be overlooked simply because it doesn't fit the SGE model, potentially affecting knowledge diversity.
Ethical considerations around data privacy, transparency, and accountability cannot be overlooked. Responsible development and deployment of SGE are crucial to ensure user privacy is respected and trustworthy information access is promoted.
Legal considerations
Search Generative Experience and Generative AI in general, represents a potential paradigm shift in technology, offering immense opportunities for individuals and businesses alike. However, alongside these opportunities come profound and multifaceted legal quandaries.
At the heart of many legal disputes surrounding generative AI lies the question of data usage. For GitHub, their Copilot tool has been trained on vast amounts of open-source code, raising concerns about attribution and adherence to software licensing terms. Similarly, lawsuits have been brought against image generators like Stable Diffusion, Midjourney, and DreamUp, alleging copyright infringement due to their utilisation of internet-scraped images for training AI models. There are many individual artists and creators that have been affected but giants like Getty Images have also found themselves embroiled in legal battles over copyright issues with AI tools like Stable Diffusion using these banks of online images as training data. These cases underscore the complexities of attributing ownership and protecting intellectual property in the realm of generative AI.
The proliferation of generative AI has given rise to a myriad of privacy concerns. AI models trained on vast datasets, including proprietary company data, raise questions about compliance with privacy policies and the potential for unauthorised data usage. Additionally, instances of AI-generated content leading to defamation claims, such as a case involving a mayor in Australia, highlight the broader societal implications of unchecked AI capabilities.
As the legal community grapples with these novel challenges, the application of existing legal frameworks proves inadequate. Copyright laws, while evolving to accommodate AI-assisted creations, still face ambiguities regarding the protectability of works wholly generated by AI. Similarly, patent laws struggle to delineate ownership and inventorship in cases where AI plays a significant role in innovation. The current regulatory environment lacks clear guidelines for protecting AI-generated intellectual property, necessitating reliance on contractual agreements to delineate rights and responsibilities among stakeholders.
For companies navigating the legal uncertainties of generative AI, proactive risk mitigation strategies are imperative. This includes sourcing training data ethically, monitoring AI systems for compliance, and ensuring that employees are well-versed in the legal implications of their work. Contracts with service and data providers should include indemnification clauses to safeguard against potential legal liabilities. Ultimately, while the legal landscape surrounding generative AI remains murky, proactive measures and a nuanced understanding of legal frameworks are essential for navigating this rapidly evolving technological frontier.
SGE: A Glimpse into the Future of Information Synthesis
Search Generative Experience (SGE) represents a significant stride within Google's quest to continue its domination of the search landscape. Since the emergence of ChatGPT3, GenAI has been a buzzword as much as a technology and has been gleefully added to everything by every business and startup looking for VC or Kickstarter funding.
Google however have been much more conservative with their use of this technology, despite their historical advantages in this field. As well as DeepMind, which stunned the world by defeating Lee Se-dol in Go in 2016, Google demoed Duplex in 2018 which could call and make appointments on your behalf, indistinguishable from a human assistant.
Google Search is the core of their offering and the quality of the product is key to the vast majority of Alphabet’s revenue and profits. That they are finally allowing their GenAI capabilities to fundamentally change how the user interacts with this space is significant and this nascent offering is one of the potential paths that this may take.
However it ends up being presented, SGE marks a fundamental shift from simply presenting results to actively distilling and synthesising information with users are no longer left to navigate a sea of text, and instead, AI-powered summaries provide a birds-eye view, highlighting the key points and diverse perspectives there might be on any given subject.
SGE has the potential to expertly cut through a mountain of articles, extracting their essence and presenting it in a readily consumable format. This snapshot, while not a substitute for deep exploration, serves as a springboard, distilling disparate content into a useful overview. Theoretically, this means that the user could have a clearer understanding of the different opinions that their question might prompt, whether it be informational, navigational, commercial or transactional, and a more guided path to their next step.
SGE's potential is undeniable, and as it matures, its ability to synthesise intricate topics, identify potential biases, and adapt to individual preferences will undoubtedly improve. It is easy to see how Google might combine the personalisation that they already deliver based on user profiles, location, device etc and combine it with SGE to tailor summaries to specific needs, offering personalised insights alongside a spectrum of viewpoints.
However, it is crucial to recognize that SGE, like any AI tool, comes with inherent limitations, with accuracy and reliability undergoing constant refinement. As potential biases could inadvertently shape the summaries, maintaining a critical lens is essential and responsible and discerning use remains paramount.
The Search Generative Experience fundamentally does the opposite, obscuring the origin(s) that the answer was synthesised from, to present you with a freshly generated conversational response. Traditional search results pages come with the links that allow you to attribute the information and while some versions of the SGE result page include links, not all do. Even when the links are presented, unless you then delve into all of them, you don’t know which bits of information were taken from where and even then it may not include all the urls that contributed to the answer given.
If you search for “what is the best shoe for running?”, it is as important for the user to know whether the answer is “Nike” or “Adidas” as it is whether the answer is coming from Nike or Adidas…or Sports Direct, runnersworld.com or The Onion
Google has spent years educating businesses and websites that search transcends mere keyword matching and SGE is a huge step down that path, offering a glimpse into a world where information is easier to parse and consume and the customer journey to conversion clarified with the inclusion of intent.
While challenges and ethical considerations remain, SGE's potential to empower users and transform our interaction with information remains undeniable and it represents a step change in how consumers get their questions answered.
