More and more users are increasingly leveraging speech recognition technology to ask questions and get instant answers. What’s important here for the sake of discussing the impact on Search Engine Marketing (SEM) is distinguishing the difference between the types of voice searches.
First is a query performed through a personal assistant. These searches occur when the user prompts their digital personal assistant (the most common ones being Siri, Google Assistant, and Alexa) with a query, which then performs the search and verbally provides a response to the user’s command or question.
Second is an in-search voice-activated query. What this means is, within a search engine or search bar (often available on most phones and tablets), the user can activate voice search to orally tell the search engine what they are looking for instead of physically typing out their query.
Searches Via Personal Assistants
Where is the information coming from and how does the smart product determine what information to recite back to the user?
The answer to these questions varies for each device. Bing is the default search engine for Alexa while the other majors player, both Siri and Google Assistant, default to Google. Not all searches are limited to just these resources. Bing, for example, may use a site like Yelp to extract information while Alexa leverages Amazon’s ownership of it to fulfill product purchase-related queries. Regardless of where the information is coming from, the voice search program will pull from the top non-paid listing of their results.
Since these programs all pull their information from the non-paid results, that leaves SEM out of the picture in being able to capitalize on this trend. Brands will need to put extra emphasis on their organic rankings and SEO practices if they wish to capitalize in the voice search space.
In-Search, Voice Activated Queries
Here, the user can control where these searches occur by either going directly to their desired search engine or changing the default engine within their settings. They can go directly to a search engine through their browser, perform a search in their browsers URL bar (which also doubles as a search bar and pulls searches from whatever the browsers default search engine is set to), or they can perform a search in their mobile devices search bar, which also performs a search on whichever engine is set as it’s default.
These queries function just as any text query does since what’s actually happening here is that the speech recognition technology is translating the user’s oral query into a text query. These queries will still serve a paid search ad (a.k.a. SEM ad) in the same way a text query would since the user’s voice search has now been migrated into a text search. So, what implications might this type of voice search have on an SEM campaign?
The way in which a user searches via voice search differs from how they search via text. According to Google, almost 70% of requests to the Google Assistant are expressed in natural language. For a search, this means the voice queries are most likely to exist in the form of a question (i.e. “What is the weather in Manhattan today?”) while text queries tend to be shorter and leverage just the keywords that pertain to specific information the user is interested in (i.e. “Manhattan weather”). Certainly, with the rise of voice search, it would be fair to assume a rise in long-tail and/or query-based searches would follow. Proving this has its challenges however because, in evaluating this theory, one must also factor in the overall rise in searches. For example, if trends indicate that the number of searches for “what is diabetes” is on the rise year over year, it would be necessary to also look at the overall rise of searches in the diabetes category to determine if they are on scale or if there’s a greater increase in the question-based queries than the short-form queries. This daunting task would require many hours and resources to properly assess. Also, for greater accuracy, we suggest looking into this data per separate disease state versus the pharmaceutical sector as a whole since the results may likely vary. In theory, a disease that’s much more personal may see fewer voice searches since users may be less inclined to verbally air personal information when others are within hearing distance.
The diagrams below demonstrate that the trend of searches for “what is diabetes” has grown over the past 14 years while the diabetes category has not followed a similar path. This suggests that there may be merit to this theory but would require more evaluation to properly validate.
If ever there is indeed a shift in how people are searching, most SEM advertisers are likely already prepared. In most cases, SEM accounts will use a combination of keyword matching options to ensure they are still serving ads to users who are making queries that are similar, but not identical, to the keywords being bid on. If an advertiser, for example, is bidding on “Manhattan weather” using the match type targeting features available, the advertiser could still have an ad appear on the query “what is the weather in Manhattan” because this query is still pertaining to both weather and Manhattan. As long as the SEM advertiser is bidding on short-term keywords that are relevant to the business, they should be covered on any new trending longer tail queries that are of the same intent. Additionally, an important component to keeping any SEM campaign running effectively is ensuring regular updates are made, allowing the account to identify and incorporate any new keywords that are rising in popularity.
Ultimately even if search terms change, the actual impact on SEM campaigns should be minimal if any because the processes of leveraging the keyword matching features plus identifying keywords and adding them to accounts should be a standard practice for most advertisers. In following these processes, regardless of what users are searching, the SEM expert managing the account will be regularly identifying what those searches are and adding them (if appropriate) to the accounts.
Given the shift to a more natural language within the query, advertisers may have an opportunity here. The more fleshed out query also provides the advertiser with greater context into what a user might actually be looking for. Going back to the weather example provided earlier, if a user were to search “what is the weather in Manhattan” it’s fair for the advertiser to assume they want to see results for the current weather in Manhattan. The same assumption could also be made if a user searched “Manhattan weather” but it’s also possible that user might want a history of the weather in Manhattan or a general overview of what Manhattan weather is like from season to season. If the user wanted either of those other queries, a voice search would be more likely to provide that clarity. This means an advertiser can leverage the longer tail queries to better customize ads and landing pages to what information the user is actually seeking.
At Klick, this theory does not change how ads are written. Best practice for ad copy development states that the advertiser should, when possible, try to cater ad copies to both the keywords and landing pages as much as possible. In following this practice, Klick already curates the user experience so that ads are appropriately matched with the keywords they serve. What this does allow for however is a more comprehensive list of keywords and the intent behind them so the curation of a high-quality user experience can be even better.
One question that may sit on many minds about voice search is whether ad messaging should differ for voice searches. The best way to approach ad copywriting regardless of the keyword is to try and incorporate the keyword or at least the intent of the keyword where possible. Additionally, it’s also even more important to properly set an expectation for the user as to what content they will receive when they land on the website. It’s beneficial for the brand to also have a call-to-action included that tells the user exactly what they want them to do when they land on the website.
Advertisers would likely already target keywords that indicate the different pieces of content a user might be interested in, even if it’s in a simpler, shorter-form fashion. In already having those keywords, the advertiser should have pre-determined the most appropriate ads to serve for them as well. One could suggest that an ad that actually incorporates the actual searched query into the ad has a greater appeal than an ad that does not. This may be true in cases where one ad with the actual keyword is compared to an ad that does not have any remote connection to that query. In a case where an ad includes the actual query in comparison to an ad that has a very close, highly relevant similar copy, the results are very mixed and inconclusive. Example: when someone searches “what is the weather in manhattan”, an ad that states “What is the weather in Manhattan?” will have a higher chance of appealing to a user over an ad that states “Manhattan climate information” but will have similar results and not be consistently better than an ad that states “Manhattan weather information”. As with most things in SEM, this is something advertisers are advised to test if that’s a reasonable and viable option for them. Keep in mind, however, that testing actual queries in ads can be challenging in some cases given that ads have character limits that advertisers need to stay within.
As far as the impact goes, the process of writing ads is still the same. Best practices don’t change and advertisers should continue to test different variations of ads to determine which is the best performing. Also important to keep in mind is that making an ad more appealing to a user is only a fraction of what’s important. Ensuring it’s capturing the right audience, setting proper expectations, and driving valued website engagements (whatever the brand deems that to be) are also important factors that should be layered into the evaluation of an ad’s performance.
In theory, as more people turn to voice search, the amount of text searches performed should decline dramatically, however with the introduction of more different ways to search (mobile devices, personal assistants, media players, video game consoles, etc) and the ever-growing adoption rate of internet usage, the overall volume of searches will continue to grow year over year.
Personal assistants are the only form of voice searching that could cut into the SEM traffic since that’s the only instance where SEM ads are not available. The addition of personal assistants also creates more searches that may not have happened without them so until the industry starts publishing data on the volume of personal assistant searches in relation to all voice searches and all total searches, it remains to be known how much traffic SEM is losing out on to these devices. They are also seen as being early in their product life cycle so as their adoption rate increases, the share of SEM traffic they may take away would also increase accordingly. For now, however, the increase in overall search traffic still continues to grow at a rate higher than whatever amount of traffic might be lost to personal assistants so any impact is negligible.