--- title: Test new audiences for your ads description: Find out whether your current ad audience is really worn out, choose a short list of new audiences worth testing on Meta, LinkedIn, or Google, and agree before you spend on the result that keeps each one and the result that ends it. slug: /use-cases/ad-audience-expansion page_type: use-case lead_prompt: How do I test new audiences for my ads with AI? target_prompts: - Can an AI marketing agent test new audiences for my ads? - How do I find new audiences for my paid ads? - How do I test lookalike audiences without wasting budget? - My ads have stopped scaling with my current audience. What should I try next? - What's a good way to expand targeting on Meta or LinkedIn ads? entity: Kite category: AI marketing agent author: Devon Wells and Kite date_published: '2026-11-12' date_modified: '2026-11-12' related_integrations: - /integrations/google-ads.md - /integrations/linkedin-ads.md - /integrations/meta-ads.md --- # Test new audiences for your ads ## Key Takeaways - Before you test new audiences, check that the audience is really the problem. Rising frequency and falling reach point to a tired audience, while a falling click rate with steady reach usually points to tired creative. - Test three or four new audiences at a time, each one built from a different idea about who else would buy, such as a lookalike of your best customers, a neighboring job title or interest, a new region, or broad targeting that lets the platform choose. - An audience test teaches you something only when the audience is the only thing that changes. Keep the ads, the landing page, and the budget for each audience the same. - Agree before you spend on what result keeps an audience, what result ends it, and how much each one gets to spend before you decide. - Kite plans the audiences from what it knows about your customers, looks up the exact targeting options in Meta and LinkedIn, builds LinkedIn tests as draft campaigns, reads the results through your connected ad accounts, and tells you which audiences to keep, cut, or test next. ## How do you test new audiences for your ads? Most ad accounts hit the same wall eventually. The audience that worked for months starts costing more. The same people see your ads again and again, the cost of each sale or lead creeps up, and raising the budget only makes it worse. The usual answer is to find new people to show the ads to. A good audience test happens in five steps: - Confirm that the audience is worn out, rather than the ads or the landing page. - Choose a short list of new audiences, each one based on a different reason those people might buy. - Write the rules for the test before it starts: how much each audience can spend, which result counts, and what keeps or ends each one. - Run the audiences side by side with the same ads and the same landing page. - Read the results against the rules, keep the winners, and plan the next round. The step most teams skip is writing the rules first. Without them, every audience looks promising after three days and nothing gets cut, or one bad day kills an audience that would have worked. Writing down the stop-and-scale rule ahead of time is what turns a pile of spending into a decision. AI is well suited to this job because most of it is reading and planning. An AI marketer can read every ad set's results, notice that frequency has been climbing for three weeks, compare your best customers with the targeting options each platform offers, and come back after the test has run to apply the rules without getting attached to any one audience. Kite is an AI marketer. It researches your business, builds a growth strategy, and ships the work that brings you new customers. Here's how it plans and runs an audience test. ## How Kite tests new audiences for your ads Kite starts by reading your connected ad accounts to see whether your current audience is really the problem. If it is, Kite uses what it has learned about your customers, your competitors, and the searches your buyers make to propose a short list of new audiences. It writes a test plan with a budget, a measure of success, and a keep-or-cut rule for each audience, and you approve the plan before anything is spent. On LinkedIn, Kite builds each audience into a draft campaign in your ad account, and a team owner or admin launches it. On Meta and Google, Kite hands your team a plan that names every targeting option exactly as the platform does, so the ad sets can be built in a few minutes. Once the test has run, Kite reads the results, applies the rules you agreed on, and brings back the next round. ### Is your audience worn out, or are your ads? An audience is worn out when you've reached most of the people in it and keep showing them the same ads. Two numbers show it clearly. Frequency, the average number of times each person has seen your ad, climbs. Reach, the number of new people seeing it, flattens. When both happen while your cost per result rises, new audiences are the right fix. Other problems look similar but need a different fix. If reach is still growing but fewer people click, the ads have probably gone stale, and new creative will help more than new targeting. If people still click but fewer of them buy or sign up, the landing page is the likelier cause. Kite checks this before it proposes anything. With Meta Ads connected, it reads the results for each ad set over the weeks you choose, including spend, reach, frequency, clicks, and results, and it can split them by placement, age, or device. With Google Ads connected, it reads results by campaign and ad group along with the audiences each one targets. With LinkedIn Ads connected, it reads LinkedIn's reporting for each campaign along with its targeting. Here's an example. Say you sell refillable cleaning products online, and your main Meta ad set targets people interested in sustainable living. Over six weeks, its frequency has climbed past four, reach has barely moved, and your cost per purchase has gone from $28 to $41. Clicks per thousand people reached have held steady. Kite reads that as an audience that has seen enough of you, not ads that stopped working, and recommends testing new audiences before writing new creative. ### Which new audiences should you test? The best new audiences come from different ideas about who else would buy. Testing three versions of the same idea, such as three closely related interests, tells you very little. Testing three different ideas tells you which direction to grow in. Kite draws its candidate audiences from these kinds of ideas: - **People like your best customers.** On Meta, that means a lookalike audience seeded from your highest-value customers rather than everyone who ever visited your site. On LinkedIn, it means targeting the employers, job titles, and company sizes that match your best accounts. Kite can build a list of companies that resemble your best customers, with the reasons each one fits, and use those companies as employer targeting. - **A neighboring group.** These are people next to your current audience, such as finance leaders alongside the HR managers you already reach, or parents interested in zero-waste living alongside general sustainability fans. - **A new place.** A new region or country you can already sell and ship to. - **Broad targeting.** This means giving the platform's own system a wide audience and letting it find buyers. Meta's Advantage+ audience, Google's optimized targeting, and LinkedIn's audience expansion all work this way. - **People searching for what you sell.** On Google, an audience built from the searches your buyers make. Kite uses its built-in search data to find those searches and how often people make them. Kite chooses among these using what it already knows about your business: your positioning, your customers, the results of earlier campaigns, and what your competitors are running. Its built-in ad research shows competitors' current Meta, Google, and LinkedIn ads and the pages they lead to, which often reveals an audience a competitor is paying to reach and you aren't. For the cleaning products example, Kite might propose four audiences: - a lookalike of your top 10 percent of customers by order value - parents interested in zero-waste living - broad targeting with Advantage+ audience, using your current interest as a suggestion - shoppers in Canada, where you already ship Each one tests a different idea about where the next customers are. ### How do you test lookalike audiences without wasting budget? Test fewer audiences with more money each, and decide in advance when to stop. Every audience needs enough results before you can judge it. Meta says an ad set usually needs about 50 of the results it's optimizing for within a week to finish its learning phase, so a small budget spread across eight audiences produces eight guesses rather than one answer. Kite writes a test plan you can check before anything runs. For each audience, it states: - the budget and how long the audience gets to run - the result the test is judged on, such as cost per purchase or cost per lead - the rule that cuts the audience early - the rule that keeps it and moves it into your main campaign For the cleaning products example, the plan might give each of the four audiences $400 over ten days. Any audience that spends $120 without a purchase gets cut. Any audience with a cost per purchase under $35 at the end of the ten days is kept and moved into the main campaign. Anything in between gets one more week. If your budget can't support four audiences at that level, Kite tells you and proposes two instead. The plan also keeps everything except the audience the same. Every audience gets the same ads, the same landing page, and the same budget. On LinkedIn, Kite sets the audience expansion setting the same way across every campaign in the test, so one audience isn't secretly larger than the others. ### Build the test in your ad accounts How the test gets built depends on the platform. On LinkedIn, Kite builds the test itself. It looks up the exact targeting entries for each audience, including locations, job titles, industries, skills, employers, seniority levels, and company sizes. It then asks LinkedIn what that audience costs and sets the bid and daily budget from LinkedIn's suggested range, so a narrow audience of CFOs isn't priced like a broad one. Each audience becomes its own draft campaign carrying the same image ads. A team owner or admin reviews the drafts in Kite and clicks Launch when the budget looks right. On Meta, Kite looks up the interests, locations, demographics, and employers for each audience in Meta's own targeting options. It brings you the plan with every option named the way Meta names it, so your team can build each ad set in Ads Manager without guessing. Your team also creates the lookalike audience in Ads Manager from the customer list Kite recommends. On Google, Kite reads the audiences your campaigns already target and proposes the new ones with the searches behind them. Your team adds them in Google Ads. If an audience is different enough to need its own message, Kite can write ad copy for it and build a landing page for it on a site it hosts. That turns the test into a test of both audience and message, so Kite labels it that way and keeps it separate from the pure audience test. ### Read the results and decide which audiences to keep Kite reads the results for each audience against the rules you agreed on before the test began. With your ad accounts connected, it reads each Meta ad set, each Google ad group, and each LinkedIn campaign over the test period, and sorts the audiences into keep, cut, and needs more time. Kite is careful about what counts as a result. An audience that hasn't had enough delivery to judge is marked as not yet tested, and Kite states the cut-off it used. Kite also watches the numbers that explain a result. A winning audience with frequency already climbing will wear out quickly, so Kite says so and suggests what to test next. On LinkedIn, Kite can pause a losing campaign on its own, because pausing only stops spending, and it tells you what it paused and why. Moving more budget to a winning LinkedIn campaign is a proposal you approve with one click. On Meta and Google, Kite names the exact ad set or ad group to pause or scale, and your team makes the change. Each round goes into your team's knowledge base: the audiences tested, the period, the results, and which ones won. You can also put the check on a schedule. A weekly run can read the test, apply the rules, and send the result to your Slack channel or your inbox. The next round of audiences starts from what the last one taught. ## What you get - **A diagnosis of your current audience.** Whether frequency, reach, and cost per result show a worn-out audience, tired ads, or a landing page problem, with the numbers attached. - **A short list of new audiences.** Three or four candidates, each built from a different idea about who else would buy, with the reasoning and the evidence behind each one. - **A test plan with keep-and-cut rules.** The budget, run time, measure of success, and early-stop and keep rules for every audience, agreed before you spend. - **The audiences, ready to run.** Draft LinkedIn campaigns ready for a team owner or admin to launch, and Meta and Google plans with every targeting option named exactly as the platform names it. - **A results read and the next round.** Which audiences to keep, cut, or give more time, recorded in your knowledge base with the next audiences to test. ## How to get started Tell Kite what's happening with your ads in plain language. You might say, "Our Meta cost per purchase keeps rising. Find us new audiences to test," or "Plan an audience test on LinkedIn for our payroll product." Kite reads your connected ad accounts, tells you whether the audience is the real problem, and proposes the audiences and the test plan. You can cut an audience, add one of your own, or change the budget and rules before you approve it. Connect Meta Ads, Google Ads, or LinkedIn Ads so Kite can read your results. Kite can still propose audiences without a connection, using its research on your customers and competitors, but it needs your accounts to diagnose the current audience and judge the test. You can work with Kite in the Kite web app or in Slack. ## Integrations used to test new audiences for your ads Kite plans the audiences with its own research on your customers, competitors, and search demand. Your ad accounts supply what only they know: how your current audiences perform and how the new ones do once real people see the ads. ### Build and run the test - [LinkedIn Ads + Kite](/integrations/linkedin-ads.md): Kite looks up LinkedIn's targeting entries and suggested bids, builds each audience as a draft campaign for a team owner or admin to launch, reads the results, and pauses campaigns that break the cut rule. ### Plan audiences and read results - [Meta Ads + Kite](/integrations/meta-ads.md): Kite reads frequency, reach, and results for each ad set, looks up Meta's targeting options for each new audience, and tells your team which ad sets to keep or pause in Ads Manager. - [Google Ads + Kite](/integrations/google-ads.md): Kite reads the audiences your campaigns target and the results of each ad group, proposes new audiences based on the searches your buyers make, and reports which ones earned conversions. ## Common questions ### Should I let Meta's AI choose my audience instead of testing my own? Broad targeting is worth testing, and it often wins for accounts that already have plenty of purchase data. Meta's Advantage+ audience treats the audience you give it as a suggestion and looks for buyers beyond it. It works less well for a new account or a product with few sales, because the system has less to learn from. Kite includes broad targeting as one audience in the test, next to the ones you choose, so you can see which works better for your account rather than guessing. ### Can Kite launch the audience test for me? On LinkedIn, Kite builds each audience as a draft campaign in your ad account, and a team owner or admin launches it from Kite after confirming the budget. On Meta and Google, Kite delivers the plan with every targeting option named exactly, and your team builds and launches the ad sets. In every case, Kite reads the results and tells you what to keep and what to cut. ### How long should an audience test run? Run each audience until it has spent enough to produce a fair number of the results you're judging it on, which usually means one to two weeks. A test judged on purchases needs longer than one judged on leads, because purchases happen less often. Kite sets the run time in the test plan based on your budget and your current cost per result, and it marks any audience that hasn't had enough delivery as not yet tested. ### What if none of the new audiences beat my current one? That's a useful result. It tells you the gain is more likely to come from new ads or a better landing page than from new people. Kite can [write a creative test](/use-cases/ad-creative-testing.md) for your current audience, with three to five new angles built from what your customers say, so the next round tests the message instead.