Skip to content

Search Intents & Generative AI

By Matthew Edgar · Last Updated: April 15, 2024

Google’s Search Generative Experience (SGE) will likely not affect organic search traffic consistently across all types of search queries. As we evaluate the potential impact of SGE on our websites, it is important to consider each query intent separately.

Query Intent

People enter different types of queries on search engines today, each with distinct intentions and motivations. One of the more common classification systems for search intent comes from Semrush. They categorize four types of search intent: informational, navigational, commercial, and transactional.

  1. Informational:  People want to learn something. This might be a fact or figure, like a query looking for the population of a specific country. Or it might be a query related to a broader topic, like a query asking the best time to travel to a specific country.
  2. Navigational: People want to find a specific page or website. For example, people may go to Google and search for “Amazon” instead of typing Amazon’s URL into the address bar. As another example, people may search for a company’s address, such as a query for the location of the nearest grocery store. Some break local searches, like those for a company’s address, into a separate query type.
  3. Commercial: People want to find information about specific companies, products, or services. For example, people may want to find reviews of dishwashers, find information about the best dishwasher manufacturers, or may want to find reviews comparing two different dishwashers. Commercial queries differ from informational queries because the focus is learning about something within the context of making a purchase.
  4. Transactional: People want to find out how to take some type of action. These queries often involve words like “buy” or “purchase”. For example, a query might be where to purchase a dishwasher, how to book a trip, or how to get groceries delivered.

A widely cited research paper from 2008 found that 80% of searches were informational and a 2011 study found that 82% of searches were informational. However, these studies only grouped queries as informational, transactional, and navigational. Under this definition, navigational searches exclude local queries, only considering queries looking for a website. The definition of transactional terms is slightly expanded from Semrush’s definition to include any terms of obtaining, downloading, or interacting with something as well as terms related to movies, music, and other forms of entertainment. This definition of transactional would capture some of the terms Semrush defines as commercial. Informational searches in this paper were a catch-all for all other queries that did not fit transactional or navigational terms.

Google’s Query Classifications

One other way of understanding search intent is from Google’s Quality Rater Guidelines. In these Guidelines, Google describes four query types: Know, Do, Website and Visit-In-Person.

  • Know: People are searching to find information about a given topic. Under Semrush’s classification, these would be referred to as informational and commercial queries. There is a special category, though, referred to as Know Simple. Know Simple queries are informational queries looking for a specific answer, like a fact or figure. A query looking for the population of a specific country is a Know Simple query.
  • Do: People are searching for how to take a specific type of action. Semrush’s classification system would consider most Do queries a transactional query. However, Google’s definition broadens the definition beyond making a purchase to include any type of interaction with a website or app.
  • Website: People are looking for a specific website, including searches for a specific URL. These are a form of navigational queries.
  • Visit-in-Person: People are looking for information about a nearby place or location. Under Semrush’s definition, these would be considered a navigational query while others would refer to these as local queries.

How Answer Engines Address Search Intent

Google’s SGE is a shift away from “search” engines toward “answer” engines. An answer engine powered by generative AI. It returns a direct response to the search query. Unlike a search engine, an answer engine downplays the list of results and gives more emphasis on the generated response. Perplexity and Bing’s AI integration are also examples of answer engines.

Informational Query Impact

An answer engine would likely cause the most disruption for informational queries, or what Google refers to as Know Queries and Know Simple Queries. Know Simple queries have a specific answer that can be provided with a very short answer. Many of these Know Simple queries are already addressed directly by Google in the search engine with a featured snippet. So, there is already little need to click to a website. For example, a query for a country’s population will return that population in a featured snippet and there is little reason to click to a website to answer this question. A generative AI tool could provide a similar type of response to shorter queries.

Other informational queries require an in-depth answer. Currently, most of these in-depth informational queries require the user to go to a website after conducting a search to obtain or find whatever information they were seeking. For example, a traditional Google search for details about a country’s demographics returns a list of articles containing more details and the user would have to pick one or more of the websites listed to get the answer. In contrast, an answer engine powered by generative AI can return a detailed answer that provides the necessary information. This likely negates the need to visit a website. For example, in the following image, the traditional search result on the left requires clicking to Wikipedia to get a complete answer to the search query. However, the generative AI result on the right answers the question—why would somebody click to Wikipedia?

Traditional search vs. generative AI search for an informational query

Commercial Query Impact

Commercial queries could also be impacted by an answer engine powered by generative AI. Take the example query in the next image as an example. In the traditional search result on the right, there is no response from generative AI and the user would need to visit Energy Star, Consumer Reports, or one of the other websites listed to research these dishwashers on their own. The generative AI response on the right pushes the traditional search results further down the page and provides much of the same information that is contained on those websites. For many users, the information provided in the generated result might be satisfactory. Would people see a need to visit any websites to gather additional information?

The question is a bit more nuanced than that. The generative AI response links directly to places where the dishwasher can be purchased. So, if the user were to click from the generated response, they would bypass the review websites. This means the answer engine response would impact different types of websites differently. In this case, review websites might lose traffic while ecommerce websites could gain traffic.

Of course, neither website may get the traffic. The user could continue the conversation with the generative AI tool to answer follow-up questions instead of visiting any website. As well, Google or future answer engines may try to complete more of the transaction by having users converse with the generative AI system.

Traditional search vs. generative AI search for a commercial query

Navigational Query Impact

Currently, it is less clear how an answer engine could impact navigational search queries. There are two main types of navigational queries that Google defines as Website and Visit-In-Person. Website queries are searching for a specific website, so it is likely that generative AI would not satisfy this type of search. People are searching for a website and want to visit that website; searchers would not want to see a conversational response discussing that website.

Generative AI may satisfy Visit-in-Person queries, though, as seen in the next example. The list of locations provides an address, hours of operation and highlights about the venue. Along with the list, the generative AI result provides a map like a traditional local search result. Interestingly, most of the websites cited are the local coffee shops. However, some of the “Highlights” information provided in the generated answer is not from the coffee shops’ websites. Instead, this data comes from other websites that have curated lists of these coffee shops, like local magazines or Yelp. The lists from Yelp and local magazines are presented in the traditional search results but are not given as much attention in the generated response. In the following screenshot, none of those websites are included in the citations though other tests of local queries like this do occasionally show those list websites in citations. Still, though, it seems unlikely the websites providing the lists will see the same type of benefit from an answer engine as they do from a traditional search engine.

Bing Copilot’s generated response for a local navigational or Visit-In-Person query
Traditional search results on Bing for a local navigational or Visit In-Person Query

Transactional Query Impact

It is also unclear how transactional queries will be affected by answer engines. The response from Google’s SGE for two different transactional queries is shown in the following image. For the query “buy a new monitor” traditional search results return a list of stores, online and offline, where the searcher can buy a monitor. However, the generative AI response, shown on the left of the image, provides background information.  In doing so, treats a transactional query more like a commercial query. Because it misunderstood the search intent, that generated response may not satisfy the person conducting the search and the searcher may still need to use the traditional search results to find their desired information.

The generated response to a search on the right may provide a more satisfying answer. The searcher wants to find “pickup trucks for sale” and the generated response returns a list of locations along with a map. Traditional search results show a list of websites that would provide the same information as the generative AI’s response; with the generative AI response, people may no longer need to use the traditional search results or visit the websites listed in those results.

Comparing generative AI responses to different transactional queries

Adapting Your SEO Strategy

Using query intent information, you can begin to assess your risks and opportunities as Google changes over to an answer engine. Here are the main steps you need to take right now.

  1. Review and classify your search queries. Pull a list of the search queries bringing traffic to your website. You can find these search queries (or keywords) in Google Search Console or with tools like Semrush or Ahrefs. As you find these terms, classify them by search intent (pro tip: ChatGPT, Gemini and Claude can help with this classification). You want to find out what types of search intent brings traffic to your website. Do you have a wider distribution across query intents? Or are most of your queries primarily a single intent type?
  2. Review generated responses for queries of each intent type. Conduct those queries with Google’s SGE, Copilot on Bing, and Perplexity. Google’s SGE is not shown as the default experience for most queries or for most users currently. However, you want to know what an SGE response would look like if (or when) it becomes the default experience. Is your website still listed in the generated response? Are websites like yours still listed—or is generative AI presenting a different type of website? For example, on the commercial query example shown earlier, SGE listed the ecommerce site to make a purchase and not the review website. This could be a big problem for review websites.
  3. Interview or test your users. If your website (or websites like yours) are listed in the generated response, the next question is if people still will want to visit a website from those generated responses. Or will the generated response be enough? There is no sense getting your website listed in the generated response (or defending your website’s current appearance in that generated response) if nobody will click. The only way to find out if people will click to a website from the generated response is by asking your users. Conduct user testing and interviews to find out whether people would click or not. You want to focus your optimization efforts on appearing in any generated responses where people would still want to click to your website (or a website like yours).

Need Help?

The shift toward answer engines is a big change for SEO. This requires adapting your SEO strategy. SGE will still list websites, but which websites are listed will change. The amount people need click to a website from a generated response will also change. If you need help assessing your situation and adapting your SEO strategy, please contact me today.

You may also like

What Are Generative AI Hallucinations?

One of the biggest problems with generative AI responses is hallucinations. What are hallucinations? Why do they happen? More importantly, as generative AI starts to dominate search results, how do we factor hallucinations into our future SEO strategies?

Will People Click Links in SGE Results?

Will people click links in SGE results? How many clicks could we lose with SGE rolling out? We conducted an in-depth study to find out.

Understanding Prompt Types

How do we optimize for prompts and make sure our companies show up in generative AI search responses? How is this different than optimizing for search queries? It begins by understanding what types of prompts people use.