Topical authority & genAI in search (SGE)

For those of us outside of the US, you have likely seen Google's AI overview finally hit your search results. Here is an overview of how GenAI will influence search behaviours and results, crucial information for anyone who regularly develops?digital content (e.g. website content, thought leadership articles, videos and images, whitepapers, infographics, webinars).

Many of us appreciate that our healthcare audiences search for information through various channels, including browsers, online newspapers, video streaming platforms, and social media. AI, as the latest channel for individuals to conduct those searches, is rapidly gaining popularity, with many folks turning to voice assistants and generative artificial intelligence chatbots.

But have you considered how your content development needs to change to stay relevant in the AI era?

For a while now, search engines have been telling us to develop content in a way that answers frequently asked questions. Those platforms have also moved us away from competing for the most competitive keywords and focussed us on becoming the authority on niche topics. If you haven’t started on those shifts yet, see below for an explanation of SGE and why you need to start immediately.

What is SGE?

Generative AI in Search (SGE) is a new search engine experience in which users can get a single yet layered answer to their query in a single search.

Features include (source):

• Ability to ask questions that are more complex and more descriptive.
• Get the gist of a topic faster, with links to relevant results to explore further.
• Make progress easily by asking conversational follow-ups or trying to suggest the subsequent query.
• Get started on tasks quickly directly where you're searching, like writing drafts or generating imagery.

AI technology is not new in search; it has been powering search for decades.

Here is a brief background on AI in search.

• 2001: One of Google's first machine learning applications was spelling correction.??????
• 2019: Bidirectional Encoder Representations from Transformers (BERT) helped search engines understand the context in which searchers used words, allowing them to ask more conversational queries.
• 2021: Multitask Unified Model (MUM) is a large language model (LLM) used across many search features to improve the quality of results, such as serving related videos (even though the video doesn't mention the specific keyword).
• Today: Generative AI in Search (SGE) is being piloted and will continue to be rolled out to more users by the end of 2024.

What functionalities make SGE work?

Functionalities include (source):

AI-powered overviews

Search engines will answer a query in a single AI-generated answer summary. From there, users can explore the information further by clicking links.

Prompts for follow-up queries

Users have the power to ask a follow-up question (e.g., "Why is this?") to refine their search without restating the context. The search engine will then suggest follow-up questions, giving users more control over their search journey.

Vertical experiences

GenAI in search will also be helpful for information journeys related to verticals. For example, genAI will create a snapshot for product searches using product options, descriptions, reviews, ratings, prices, and product images.

Search ads will remain

Search ads will continue to appear in search results, clearly labelled as "Sponsored," maintaining the balance between user experience and advertising needs.

Creativity and productivity

Eventually, genAI in search will include a creativity element, allowing users to find information and use it in the creative process.

Image generation

Users can generate images to visualise an idea or search for inspiration. Expect images to include metadata stating they are AI-generated and an invisible digital watermark.

User experience

The interface will include recognisable elements from the current search experience, such as website links. Still, additional best practice elements will be introduced, like call-to-action buttons directing users how to navigate the new experience (such as the prompts for follow-up queries mentioned above).?

Experimental phase

SGE has known limitations (see below), so it was first available in Google's Search Labs, where a small group of users tried out the experience and provided feedback.

Limitations:

As with all AI, there are limitations in the functionality. The following are some limitations observed during testing that future users of SGE may experience as well, although Google is working to address all these limitations:

Misinterpretation?

Misinterpretation of language can change the context of the original information used to answer a query, although the original information appropriately answers the query asked. This means that the AI may not fully understand the nuances of the language used in the query, potentially leading to a misrepresentation of the intended meaning.

Hallucination

The experience can misrepresent facts or inaccurately identify insights similar to other LLMs. This means that the AI may generate information that is not supported by the original data, leading to potentially misleading results.

Bias

Google trained the model using publicly available web content, which often narrowly represents the world's population. Therefore, biased results are produced, similar to those observed in search results today. This means that the AI may favour specific perspectives or sources over others, potentially limiting the diversity of search results.

Opinionated content implying a persona

There have been instances when the output reflects opinions shared on the web, giving the impression of the model displaying a persona.

Contradiction with existing search features

The output can contradict other information in search results because current search results reflect the perspective of a single source, while SGE will represent a perspective formulated from a range of sources.

SGE in application: the healthcare focus

• Search quality systems will still apply: SGE will maintain the high standard for responses to sensitive queries applied in search today, such as health, finance, or civic information. Answers to the questions about these topics will only come from reliable sources (e.g., the NHS or CDC). The outputs will also include disclaimers. For example, health-related queries will include a disclaimer saying people should always talk to their healthcare professional.
• Content policies: The answers provided should not violate Google's existing content policies, including their policy towards medical content.

Tips and tricks

Now, for the most critical part of this article: the tips and tricks. Although this list of tips will evolve over the coming months, here are some initial considerations for preparing your digital content for this shift in the search experience.

• Develop content in a conversational format that would answer frequently asked questions.
• Convince executives and authors to become the authorities on niche topics that support the organisation's content pillars.???
• Ensure long-form content includes short sentences summarising the key messages.
• Use actively used keywords related to your topic in your content instead of jargon, which will help LLMs use the content for training.??
• Assume users won't visit your website to "read more" in the future and expect their questions to be answered directly in search results.?
• Anticipate that generated answers will use both owned and competitor information alongside each other. The cause of this combination of information is that search result outputs will use many sources vs. current search results only using one source. This combination will make competitor content audits even more critical moving forward.
• Consider how media outlets will cover your key messages. Digital media content will train LLMs as well.
• Health –
• a. Expect the SGE to adhere to the current agreements between Google and national health systems. This agreement will result in these authorities being presented as the most authoritative answer to all health-related queries (e.g., the NHS or the CDC), not pharma companies.???
• b. Complex searches, including academic or personal health searches, will progress in a single experience rather than over time, as we have seen happen in the past. This change is because search engines continually suggest a follow-up search query.

It doesn't stop there

I have some more SEO tips for content developers to familiarise themselves with. This year, a massive discovery of info indicated how Google ranks its content. The information "leaked" provides further valuable guidance for creating content so Google's search engine will rank content highly. In the ongoing theme of driving "topical authority", the leaked info suggests content developers should be developing fresh, original, and topical content.

Google Search Doc leak

On March 13th, 2024, a trove of Google documents leaked, providing an unprecedented view into the inner workings of Google search. The documents unveil potential Google Search ranking factors (over 14,000 factors), such as clicks, links, content, entities, and Chrome data, that could guide content development (in partnership with technical considerations such as site structure and tags). While the documentation is quite long, here is a short list of tips you should apply to all digital content moving forward.?

Tips for content development:

• Links: Make sure links are relevant and diverse
• Successful links: Create great, ORIGINAL content and user experiences so users click through to your content and stay
• Brand: Your brand should be popular and recognisable outside of Google
• Authors and entities: Google looks at the content author and determines if the entity is the author?
• Site authority: Create strong authoritative content and avoid weak content
• Allowlists: Google allows certain domains, such as government health sites
• Freshness: Google determines how recent content was published
• Page title: Google weighs how well a page title matches a search query
• Topical authority: Google determines whether content is a part of the core topic of the website
• Page quality: Google is using an LLM to estimate “effort” for article pages
• Website linking strategy: Low-quality and high-quality pages linking to other pages on the website could taint the destination page. News pages linking also improve content rankings
• Images: High-quality images that receive a large number of clicks are ranked highly
• Weight: Bolding text and the size of text have an impact on content scores
• Length: Put essential messages early in long content. Short content should be original and robust

Let me know if you want to learn more about my POV for driving Topical Authority for corporate comms, executive visibility, and thought leadership. I can share what info we have.