Today, voice search and its cousin, AI-powered search experiences have become more and more prevalent in our daily lives. These include search functions like smart speakers, mobile assistants, and AI Overviews from Google. Increasingly, these are not just taking up more of the search ecosystem but are themselves overlapping. 

Voice searches are search engine queries that are spoken instead of typed and are longer and more conversational in nature. This includes being far more likely to be “full” questions as opposed to shorter keyword-based queries. They’re often made through platforms like Siri or Google’s voice search option. 

About one-in-five internet users use voice search, with those numbers still growing — voice queries skew heavily towards mobile phone and smart speaker users. 

Crucially, these queries further skew heavily towards high-intent, real-world contexts. 

The two both have implications for search and your advertising to that end. For example, AI Overviews are changing users’ behavior when it comes to clicks and shrinking the available real estate for paid and organic results. Because of this, paid search ads and organic results are now being pushed below the fold in favor of AI Overviews. 

Both voice search and AI-related search results come together because the former is more likely to be phrased as longer-tailed, conversational queries, while the latter is mostly likely to appear from these longer-tailed, conversational queries. 

Naturally, this poses a risk to the health of paid search campaigns, as advertisers have been forced to be more aggressive with their bids to appear high in the results while also being outcompeted by Google’s own AI Overviews. 

Because of this, optimizing your paid search for voice and AI-driven SERPs is not just something that should be relegated to a side project — rather, these are necessary steps towards risk management and growth. 

How voice queries differ from typed search (and why that matters) 

For a typical typed query, you’re likely to see a more conceptual search — for example, “Seattle dentist” — whereas for a voice query, they tend to be phrased as complete, full questions: “Who is a highly-rated dentist near me open on Saturday?” 

Voice queries then are longer and heavily question-based, along with being often more conversational. This usually leads to details in the query that act as richer context signals. For example, urgency (“today” or “right now”), location (“near me”) and qualifiers (“affordable,” “for kids,” or “dog-friendly”). 

The implications of this for your campaigns include adapting your keyword strategy and match-type choices, plus how you’re gauging user intent in relation to the marketing funnel. Basically, people are using more voice search with longer, more conversational queries specifically to learn more information, meaning it’s helpful for you in turn to target long-tailed keywords with narrow match types to help ensure your ads appear in response to the most relevant searches. 

Also, note that voice-style queries can behave differently in AI Overviews than in traditional SERPs, which is likely to impact your expected click-through rates and cost per acquisition. 

Finding voice-like queries in your existing search data 

Luckily, there are methods to learn what existing queries are out there that already are similar to what we’re seeing from voice queries. 

For example, you can mine your Google Ads search term reports to find long, multi-word queries and question phrases. These essentially act as proxies for voice searches because they’re structured the same way and have comparable or even the same intent, whatever their medium. 

You can find these by opening the search terms report for a search campaign and checking queries that are, say, five words or longer. By analyzing these, you can start to understand how users’ conversational searches fit within your target market (or markets). 

It’s also possible to use GA4 exploration reports, where you can filter by query length, device type, and question words — i.e. who, what, when, where, how — to use as a proxy for voice searches and comparable queries. Specifically, filtering for queries that are A) five words or longer and B) on mobile will help approximate voice usage, since the vast majority of these searches are made on a phone. 

We also recommend tracking or labelling these queries that are more “voice-like,” and from there creating custom ad groups accordingly. This lets you monitor their performance separately from traditional search queries. 

Over time, patterns to look for in these include queries related to or using terms like “near me,” “open now,” “how much does X cost,” “what is the best X for Y,” and so on — keep an eye on query patterns that may show up more often and provide extra insight. 

Keyword and match-type strategy for conversational searches 

As we’ve alluded to, your keyword strategy when focusing on voice search should shift towards more emphasis on long-tail, conversational phrases — these mirror how people actually speak to real-life human assistants, but now applied to search engines. 

When it comes to organizing question-driven and “how to” phrases into tight, conversationally-themed ad groups, consider the root concern that’s being implied by these long-tailed, conversational queries. Make your ad copy for these groups address this root concern (and of course follow standard copy best practices regarding value propositions, CTAs, and the like). 

Sometimes you may also want to use broad match settings alongside robust negatives instead of just phrase or exact match. This is useful in high-CPA situations, or sensitive verticals. 

This can be useful because broad match uses a wider range of signals that can make it more effective in the right context, but you need a thorough negative keyword list alongside it to ensure it isn’t overreaching and serving ads to irrelevant searches. 

We should note that conversational long-tail terms can lower costs per click — with lower competition and fewer advertisers bidding on these terms — but to do so may require smart bidding and enough volume to be viable. To that end, we recommend leveraging automated bidding strategies like max conversions, max conversion value, and target cost per acquisition, and broad match terms to capture the widest range of available signals. 

Lastly, we’d encourage you to periodically check on your keyword usage and expand when appropriate to capture emerging voice variations, slang, and other ways that search queries evolve. 

Structuring and measuring campaigns around voice-driven intent 

On a bigger scale, this demands that your campaigns are informed by these new factors. 

To begin, we recommend having separate “voice and conversational” campaigns in accounts where budgets and high-enough volumes can justify more granular control. A campaign focused on voice search and the qualities that come with that can be advertising the same product or service as a more conventional campaign, but ad copy, keywords, and the other details can be tailored to target users who are making voice searches versus traditional searches. 

Part of this is combining audience signals — in-market, custom segments, for example — with conversational keywords to prioritize higher-value voice queries. For example, question-based keyword phrases and “near me” keywords, to start, can steer bidding towards higher-intent voice queries. 

We should note, also, that Performance Max can pick up longer, conversational queries with its automation, but we recommend keeping Search-only campaigns when you want stronger control over search terms, negatives, ad messaging, locations, or landing page alignment. 

Measuring the impact of voice search (without a “voice search” column) 

We suggest that you label your voice-focused keywords, ads, campaigns, etc., then run experiments to compare performance against your current structure using the same KPIs. 

Of course, we don’t have a “voice versus text queries” dimension for measuring results on Google or Microsoft, so you need to manually set up campaigns designed to capture these searches separately. 

At Fujisan we generally recommend defining a “conversational query” segment based on search term length and question words, and tracking its performance over time. Label your voice-optimized campaigns and ad groups, monitoring their metrics like CTR, CVR, CPA, and impression share separately from traditional search campaigns based on shorter queries. 
 
Lastly, use GA4 explorations to cross-tab by device, query length, and landing page so you can see where voice-style behavior shows up. 

Bottom line 

At the end of the day, these voice- and AI-driven SERPs are affecting query behavior and ad visibility, so paid search needs to adapt accordingly for this new conversational intent. 

Remember the markers like five-plus word queries, question words, and mobile device signals, and that will help you adapt your targeting accordingly. This can in turn inform your ensuing campaigns and ad groups that separate conversational intent from traditional search and make sure your match types, messaging, landing page, and other factors align. 

And of course, if you want to take these steps but feel you could benefit from professional help, Fujisan Marketing can help with auditing your account, identifying opportunities, and putting a test and campaign plan into action.