Categories
Artificial Intelligence Google Semantic SEO

Ask A Follow-Up Question

What is “Ask a Follow Up” Question in Google’s SGE?

Within Google’s new AI-powered Search Generative Experience (SGE), new Question Answering SERPs are emerging; like “Ask a follow up”.

It just takes a click on the “Ask a follow up” button or on one of the suggested next steps under the snapshot. This feature empowers you to ask any next question that comes to mind on your topic. SGE already has the background reference of your initial query, so it will answer in context. (Unless you hit “reset” to start a new query thread.)

A person searching on Google doesn’t always find what they want on the first display page. SGE’s conversational mode is engineered to answer follow-up questions that are relevant to the initial query. Its “memory” maintains the context of your original question as it lets you refine your search with an additional query.

Table of Contents

I’ll present my logic here as to why the new “ask a follow up” is so rich with lead generation opportunity.

Search Engine Optimization is a complex area where engineering, the ability to resolve technical issues, creativity, content creation, business strategy, and science need to work in tandem. Google has long been an “Search Answer Engine“. Yes, I intentionally placed “business strategy” here. In truth, SEOs are business developers. Indeed SEO is your best investment to get your business in front of new buyers. Since most people’s purchases start with some form of QA research, winning a SERP presence at this early stage is vital.

Google searches don’t always end after the first query. SGE’s conversational mode is designed to answer follow-up questions with better ability to puzzle out the intent of your search. You engage SGE’s conversational mode by tapping “ask a follow-up” or one of the suggested next steps under the snapshot.

Here is how Google explains its ‘ask a follow up’ feature:

“People can tap to ‘ask a follow up’ or on one of the suggested next steps under the snapshot. This will launch the new conversational mode, where they can naturally ask Google more about the topic they are exploring.

 

SGE is not designed to respond in the first person, for example, and we finetuned the model to provide objective, neutral responses that are corroborated with web results.” – An Overview of SGE, Google [1]

I find it interesting that the document projects that its AI answers will “provide objective, neutral responses”. Meaning, a conversational tone and also “that are corroborated with web results”. Perhaps, later on, this means that its AI algorithms will spin their own blended search responses. And perhaps feel no need then to link to the answers’ original sources. We’ll see.

Need a visual of how this button is meant to be useful? Think of how you “stack” something, like towels in a linen closet. This button lets you “stack” further search queries and gain more highly filtered information. It triggers Google’s AI capabilities to intelligently combine your questions to answer contextually going forward. It is meant to save the searcher time.

Multimodal Question Answering Capabilities

Google’s advancement in AI text generation have better conceptualization of question answer pairs.

So far in Google Search Labs, we see its ability to combine text and visuals seamlessly in order to answer questions involving multiple data types. Its Gemini project, which may come next, summarizes QA information with ease. It capably combines multiple types of data to produce what it believes is the best answer. This may also power it follow-up question options for complex or ambiguous quries.

This aligns with Google’s commitment to provide the best answer to searcher’s questions. For many niches, a ton of People Also Ask QA options populate. The insurance niche is rich with PAA SERPs. With heavy user interaction within QA SERPs, this makes search technology less reliant on keywords and better at understanding the semantic meaning of queries.

Search engines provide answers. How they use https://www.hillwebcreations.com/using-ai-and-knowledge-graphs-to-maximize-content/s is rapidly changing their ability to understand and categorize queries in order to dynamically generate accurate answer results. To outshine competitive search engines requires more advanced technology in order to provide the best question answer search experience.

What is a Google answer snapshot?

SGE uses AI to generate unique responses to users’ multi-angle queries. It draws information from multiple sources simultaneously. Google calls this a snapshot.

You control your web content that has chances to be included in a Google SGE snapshot.

“When appropriate, SGE will show an AI-powered snapshot to help people quickly get an overview, with factors to consider and a helpful synthesis of relevant insights and information. These snapshots serve as a jumping-off point from which people can explore a wide range of content and perspectives on the web.” – A New Way to Serach with Generative AI: An overview of SGE

“We believe everyone benefits from a vibrant content ecosystem. Key to that is web publishers having choice and control over their content, and opportunities to derive value from participating in the web ecosystem. However, we recognize that existing web publisher controls were developed before new AI and research use cases.” – A principled approach to evolving choice and control for web content, Google June 2023 [2]

Google’s SGE Sequence Approach to Interactive Question Answering

It has long been natural for people to ask search engines questions. Voice search has shaped new approaches to conversational queries, and Google Search is advancing to respond in a similar fashion. Its SGE “Ask a follow up” feature gives us a glimpse into future conversational-type answers.

The best way to make your content marketing on a shoestring budget effective ➡ invest in evergreen content. Publishing one exceptional piece of content per month is likely more effective than 10 pieces of content that readers skim and ffers no added-value to re-read. If you effectively answer your audience’s question, Google may pull your content (with a source link) into its SERPs. That may be in Google’s Knowledge Graph Question Answering feature, an answer card, or one of several answer SERP formats that searchers are provided.

 

Interactive Question Answering (IQA) requires an intelligent agent to irespond within a dynamic, conversational environment in order to gather information necessary to answer questions. We learned from the BERT model of natural language process of where Google Search was headed.

The Association for Computational Linguistics published an abstract in July 2022 that I read and helped shape my perspective on QA content strategies.

“The BERT model achieved a higher QA accuracy than sufficient information score during evaluation, indicating that the context window spanning multiple states was a boon to QA accuracy. During training, the BERT model achieved almost perfect scores for question-answering on the held-out set of offline trajectories.” – A Sequence Modelling Approach to Question Answering in Text-Based Games

Google needs great content to train its large language models (LLMs). They build on the Bert algorithm. It relies on exceptional content from the web to “feed” these language models. This means creating not only valuable and engaging content for search engines, but foremost, for the human readers who use search to ask questions.

Follow-up Questions meaning and examples of use

I like these examples from the Cambridge English Corpus of how follow-up questions are useful and why your content should be optimized for them. It suggests why Google may be introducing this feature in its advanced search technology.

– Moreover, including a follow-up question increases the precision of the result.

– Thus, efficiency was seen to be improved by adding a follow-up question.

– In addition, including a follow-up question increases the precision of the result.

– Following each stimulus-sentence presentation, subjects were asked a follow-up question, which differed according to the training condition being delivered.

– We begin with an elementary example and a follow-up question which serve to motivate much of the material to follow. (Source: Follow Up Question article by Cambridge English Corpus [3])

Read or download the Google’s New Way to Search With Generative AI PDF to learn more. Google Bard AI seems to be advancing in tandem with its emerging SGE. Daily, new announcements emerge. This makes it a fast and exciting learning adventure.

Google’s SGE seeks to provide highly relevant AI answers

Google is testing its SGE answers to see if users click through to sites for more information. Currently, unique keywords show different formats and levels of detail in the AI answers it provides.  The question of which niches will be hit the hardest by SGE remains until it is fully rolled out. Without leveraging AI, Google was challenged to efficiently and effectively answer longtail queries.

 

LLMs and generative AI are already resolving much of this issue. They source SGE results with trusted, traditional web search results to improve relevancy, accurate, and fact-checked AI answers. We see this especially occurring for informative queries. Effective content provides the customer-centric answers they want. In other words, on-page SEO highly influences what searchers click on in SERPs.

On top of obtaining a tailored answer to your question, SGE Search will also display images, links, and products that it thinks are related to your query. You can simplify or refine your query – without you needing to think through how to conduct additional searches on your own. And while new SGE features are rolling out, Google has also vastly decreased FAQ rich results in SERPs.

Semantic matching applications for question answering systems

Semantic matching is a technology in computer science to identify semantic-related information. At present, the most common semantic matching applications in the deep learning sphere are Question Answering System (QA), Semantic Relatedness, and Natural Language Inference (NLI).

They each may have a role in what technology is behind the “Ask a follow up” question feature. We know that SGE is an experimental Google search engine beta project that uses artificial intelligence to generate contextual answers to complex questions. So, what has Google told us that offers glimpses of how this works? Here is a quote from CEO Thomas Kurian of Google Cloud that I find helpful.

“Text Embeddings API is a new API endpoint that lets developers build recommendation engines, classifiers, question-answering systems, similarity matching, and other sophisticated applications based on semantic understanding of text or images” – At Google I/O, generative AI gets to work, May 10, 2023

Growing Importance of Question Answering SERPs

It’s fascinating to follow what patents Google files – and obtains – for processing query variants that may impact its evolving question answer features. Since SGE loads content dynamically, in-the-moment, during run-time it relies on a trained generative model based on tokens from the original queries and additional input entities.

A Google patent granted on the 30th of May 2023, tells us something that I think may relate to this. “Instead of providing the answer response, multiple variants that are ‘follow-up’ variants are generated in order to verify the accuracy of the response to the original query.”

Its system has the capacity to generate query variants for novel, new, first-time, or tail queries. The SGE’s ask a follow up currently lacks sufficient data for such queries. However, everything you/anyone publishes may be used in the future. Google hopes to reduce the 15% of queries submitted that have never been previously asked. Its new generative model intends to predict which query variants to generate even for infrequently asked queries by leveraging a neural network.

Here is another article by Google that I also learned a lot from – A new way to search with generative AI – Get AI-powered overviews and ask follow ups, right in Search.

Google Patents Possible Behind its Ask a Follow Up FunctionGenerating query variants using a google-trained generative AI model for ask a follow up question in SGE

Patents provide clues to Google’s SGE question answering system

With the SGE being so new, we can find clues to how its QA features may work from patents.

The SGE is meant to be a question answering system that selects one or several results from a large number of alternative answers to best match a given query. After all, QA search responses are one of the key technologies in the field of information retrieval. Bard AI and the SGE are attempts at improved methods that consider the semantic information of the contexts.

Google patented the pre-training language model that the BERT Question-Answering uses by leveraging masking. Extended context modeled by BigBird benefits various NLP tasks like question answering, summarization, long document classification, etc. [5]

It’s also interesting that Google’s Google patent US20110125734A1: Questions and answers generation was abandoned after being published May 2011. However, the Google patent US9213748B1 Generating related questions for search queries is still active after being granted in December 2015. [6] Here is what we learn from it.

Method described for identifying related questions for a search query:

  • Receiving a search query from a user device.
  • Obtaining a plurality of search results for the search query provided by a search engine, wherein each of the search results identifies a respective search result resource.
  • Determining one or more respective topic sets for each search result resource, wherein the topic sets for the search result resource are selected from previously submitted search queries that have resulted in users selecting search results identifying the search result resource.
  • Selecting related questions from a question database using the topic sets.
  • Transmitting data identifying the related questions to the user device as part of a response to the search query.

And here is a quote from a more recent patent application granted on May 30, 2023.

“A generative model is productive, in that it can be utilized to actively generate a variant of a query based on application of tokens of the query to the generative model, and optionally based on application of additional input features to the generative model. In this manner, a generative model can be utilized to generate variant(s) of any query, even if the generative model was not trained based on the query. Accordingly, the generative model can be utilized to generate variants for novel queries and for so-called “tail” queries (i.e., those with a submission frequency and/or submission quantity below a threshold). As a result, queries can be processed more effectively as the richer query input can lead to more efficient identification of relevant results.” – Google Patent US11663201B2: Generating query variants using a trained generative model

The above patents are an indication of Google’s intent to expand its question answering SERPs and entity knowledge graphs.

Bard and SGE let You “Build” Your Follow-UP QuestionsGoogle's SGE Ask a follow-up is visually rich

Google’s latest article on Bard conversational advancements sets the backdrop by stating, “That’s why we created Bard: to help you explore that curiosity, augment your imagination and ultimately get your ideas off the ground — not just by answering your questions, but by helping you build on them.”

It’s interesting to me how that accurately describes the new “Ask a follow up” question-answering feature. You can build on your first question by refining it with a second question and so on.

“Pin and rename conversations: We’ve heard you want to be able to revisit prompts (queries/questions), so we’re adding new ways to pin and rename your conversations with Bard. Now when you start a conversation, you’ll see options to pin, rename and pick up recent conversations in the sidebar. For example, if you ask Bard to help you compare outdoor sports for the summer, you can revisit the tips later. This feature is now live in over 40 languages.” – Bard’s latest update: more features, languages and countries

It’s possible that when fully rolled out, SGE may use another button name for this feature. It will be exciting to see. But we know from its patents and statements that it is building better question-answering capabilities. Recently, on July 7, 2023, it published a Modular visual question answering via code generation post. [7] This aligns with how image-rich SGE’s query answers already are.

Why this matters to you: SGE is a radically different search experience. This changing SERP may impact both your paid and organic results. This could cause fluctuations in your rankings, web traffic, content generation and advertising costs.

I’ll wrap up by sharing some of my personal thoughts from my experience using SGE and Bard AI.

Content Marketer’s Tips for “Ask a Follow Up” Inclusion

1. Build your E-E-A-T and add value. Content writers and SEOs relying largely on keyword-loaded content to target algorithms versus searchers’ needs should be rewriting those pages. Authors should lean into their personal experience and expertise. This is the best way to offer new insights to your audience and provide comprehensive content that answers searchers’ questions.

2. Use FAQ schema markup to boost your question-answering content. If you are unfamiliar with creating custom schema code, try a JSON-LD FAQ code generator. For example, you could use AIPRM, an AI Prompt Marketplace for ChatGPT, Midjourney, and more. [4]

3. Plan your content creation ahead and resolve question-answer content gaps. While currently Google’s Search Generative Experience or AI answers are not available for all searches across industries, the future will be expansive. The cost of providing AI-generated answers is higher than for displaying traditional blue links at this time. Just as Google has already shortened display load times, it will likely evolve to optimize costs. It may opt to caching AI answers to achieve this.

4. Customize QA conversations to align with user needs. Use sub-graphs to provide specific responses to specific audience questions. In this way you can tailor both traditional and AI generated conversations to your users’ related needs.

5. Conduct ongoing market research to assess how various SERP features display. Generated AI query variants for “People also search for” (PASF) and PAA currently display in Google’s SGE lab, however, the “Ask a follow up” feature is above both of them. It is important to prepare your SEO techniques today for tomorrow. I find it fascinating and rich with opportunities.

6. Be an early AI adopter. Google Launches NotebookLM; it summarizes documents and answers questions. It formerly was called Project Tailwind and was announced at I/O in May. Product Manager of Google Labs Raiza Martin and Editorial Director of Google Labs Steven Johnson stated its intent is to let you dive deeper and “use the power and promise of language models paired with your existing content to gain critical insights, faster.”

It helps you get a document summary, ask questions about the documents you’ve uploaded, generate ideas for Q&A content inspirations, and more. The July 12, 2023, Google blog post is just one of many to inspire and assist people aspiring to move forward and maximize SGE’s potential. [8]

Can you or I afford to wait on this?

I’ll quote a concise statement from an email I received today from Nicola Agius, Paid Media Editor at Search Engine Journal. “The biggest threat to your job isn’t AI – it’s other marketers that have figured out how to use AI to their advantage.”

And below is another quote from Google.

“We’re in the middle of a third paradigm shift. The first shift was the arrival of the internet, with its revolution in information and services. The second was the explosion of mobile computing, driven by the adoption of smartphones. And today, advances in artificial intelligence (AI) promise to deliver transformations that are even more profound, for both consumers and businesses. – Why marketers must be bold, creative, and curious about AI, Kristen O’Hara, June 2023 [9]

“SGE shortens your path to the info you seek. You’ll be able to ask follow up. People often come and ask a question and then they refine their question… but they have to start over. Now you’ll be able to ask a natural question like you would a normal person…..and have same sense of context that you have in a conversation.

 

If we cannot corroborate it, then we will not output it.” – Liz Reid, Google’s VP of search [10]

SUMMARY: Take Advantage of “Ask a Follow Up” (Questions)

You can take significant steps to make your website and digital brand smarter and more accessible. We help clients transform their knowledge graph into interactive conversations, that enhance the user experience. Today you can pave the way for more effective SEO implementations and leverage “Ask a follow up” opportunities.

To make the most of Google’s incorporation of generative AI in your marketing strategy, reach out to Hill Web Marketing. 851-206-2410

We can conduct your personal Google SGE SERP analysis

 

Resources:

[1] static.googleusercontent.com/media/www.google.com/en//search/howsearchworks/google-about-SGE.pdf

[2] blog.google/technology/ai/ai-web-publisher-controls-sign-up/

[3] dictionary.cambridge.org/example/english/follow-up-question

[4] forum.aiprm.com/t/seo-faqpage-schema-markup-json-ld-generator-through-live-crawling/38610

[5] analyticsvidhya.com/blog/2022/11/an-introduction-to-bigbird/ and ai.googleblog.com/2021/03/constructing-transformers-for-longer.html

[6] patents.google.com/patent/US20110125734A1/ and patents.google.com/patent/US9213748B1/

[7] ai.googleblog.com/2023/07/modular-visual-question-answering-via.html

[8] blog.google/technology/ai/notebooklm-google-ai/

[9] thinkwithgoogle.com/marketing-strategies/automation/ai-in-marketing/

[10] https://fortune.com/longform/google-ai-chatbots-bard-search-sge-advertising/