Written By

AI-Powered Audience Discovery: A Step-by-Step Guide to Finding Your ICP on LinkedIn and Google Ads

AI-Powered Audience Discovery: A Step-by-Step Guide to Finding Your ICP on LinkedIn and Google Ads

It's Saturday, July 26, 2025. A B2B marketer somewhere in the United States is staring at the LinkedIn Campaign Manager interface, facing a dizzying array of targeting options. Job titles, skills, company industries, group memberships—the combinations are infinite. The question is paralyzing: Where do I even start?

For years, the answer has been based on assumptions. We build a persona in a meeting room. We decide, based on our gut feelings and a few customer interviews, that our Ideal Customer Profile (ICP) is a "VP of Marketing at a 500-person SaaS company." We then dutifully plug those criteria into LinkedIn and Google Ads and hope for the best.

This is the old way. It's an educated guess at best, and a costly, misguided shot in the dark at worst. It's the reason why so much of our ad spend is wasted on audiences that will never convert.

But what if you could stop guessing? What if you could let your own data tell you who your best customers really are? What if you could use the powerful Artificial Intelligence already built into these ad platforms to uncover new, profitable audiences you never would have thought to target?

This is the promise of AI-Powered Audience Discovery. It's a strategic shift from assuming you know your ICP to using your customer data as a "seed" to grow a forest of new opportunities. This guide will provide a step-by-step framework for using the AI within LinkedIn and Google Ads to find your true ICP, eliminate wasted spend, and unlock scalable growth.

The Prerequisite: A High-Quality "Seed List" is Everything

Before you can leverage any AI, you must give it high-quality data to learn from. The entire success of this strategy rests on the quality of your seed list. The AI is a powerful pattern-matching machine; if you feed it garbage patterns, it will find you more garbage.

What is a Seed List? A seed list is a curated list of your absolute best customers. It is not a list of all your customers. It's a list of the customers you wish you could clone a thousand times over.

How to Define Your "Best" Customers: This is where you need to move beyond simple revenue numbers. Work with your data or RevOps team to build a list of customers based on a combination of factors that define true business value:

  • High     Lifetime Value (LTV): Customers who have spent the most with you over     time.
  • Low     Cost of Acquisition (CAC): Customers who were efficient to acquire.
  • High     Product Engagement: The users who are true power users of your     platform.
  • Low     Churn Rate / High Retention: Your most loyal customers.
  • Fast     Sales Cycle: Customers who moved through your pipeline quickly.
  • Ideal     Firmographics: They fit the size, industry, and location you want to     target more of.

Data Point: Your goal should be to create a clean, formatted CSV file with at least 1,000 high-quality contacts. While platforms can work with as few as 300, a list of 1,000+ gives the AI a much richer dataset to analyze, leading to significantly more accurate results. This is the most important step in the entire process. Do not skip it.

Part 1: Finding Your ICP with LinkedIn's AI

LinkedIn is the undisputed king of B2B professional data. Its AI can analyze the deep professional characteristics of your seed list to find new audiences with surgical precision.

Step 1: Upload Your Seed List as a "Matched Audience"

This is how you give the AI its learning material.

  1. Navigate     to the "Audiences" tab within Campaign Manager.
  2. Click     "Create audience" and select "List upload."
  3. Upload     your curated CSV file (your seed list).
  4. LinkedIn     will now try to match the emails and company information in your list to     the profiles on its platform. This process usually takes 24-48 hours.     Expect a match rate between 30-60%; it will never be 100%.

Step 2: Create a "Lookalike Audience"

This is where you unleash the AI.

  1. Once     your Matched Audience is done processing, select it from your audience     list.
  2. Click     the "Create lookalike" button.
  3. LinkedIn's     AI will now analyze thousands of data points from the matched profiles in     your seed list—their job titles, seniority levels, skills, group     memberships, company affiliations, education history, and more.
  4. It     then scans its entire network of over 900 million users to find other     professionals who share these same, often hidden, characteristics. The     result is a new, much larger audience of high-potential prospects.

Step 3: Analyze the Lookalike with "Audience Insights"

This is the "discovery" phase. Instead of immediately targeting this new audience, first take the time to deconstruct it.

  1. Select     your new Lookalike Audience and click on the "Insights" tab.
  2. This     dashboard is a goldmine. It will show you a detailed breakdown of the     AI-generated audience, including:
       
    • Top      Job Functions & Seniorities
      •  
    • Top      Skills
      •  
    • Top      Company Industries & Sizes
      •  
    • Years      of Experience
    3.  
  3. This     is where you have your "Aha!" moments. You might discover that     while you thought you were selling to VPs of Marketing, the AI found that     a huge percentage of your best customers actually have     "Director" level seniority. Or you might find a common,     unexpected skill like "Data Analysis" or "Project     Management" that you can now add to your targeting criteria.

Step 4: Refine and Test Your New Audiences

The lookalike audience is a powerful starting point, not afinal destination. Now you can use the insights you just gained to run highly targeted tests.

  • Test     1 (Broad): Run a campaign targeting only the raw Lookalike Audience.
  • Test     2 (Refined): Run a separate campaign targeting the Lookalike Audience,     but layer on an additional filter based on your insights. For example: Lookalike     Audience AND Skills = Data Analysis.
  • Test     3 (Standard): Run a third campaign using your "old" manual     targeting criteria as a control group.

Compare the Cost Per Lead (CPL) and Cost Per Acquisition(CPA) across these campaigns. You will often find that the AI-powered audiences outperform manual targeting significantly.

Step 5: Feed the Machine with Conversion Data

Make sure you have the LinkedIn Insight Tag installed on your website and that you have conversion tracking set up for key actions (like demo requests or trial sign-ups). Every time someone from one of your campaigns converts, it sends a positive signal back to LinkedIn's AI. This teaches the algorithm more about what a good customer looks like, making your lookalike models and campaign delivery smarter and more efficient over time.

Part 2: Finding Your ICP with Google's AI

While LinkedIn is about professional profiles, Google's AIis about behavior—what people search for, what videos they watch, what websites they visit. It provides a different but equally valuable dimension to your ICP.

Step 1: Upload Your Seed List via "Customer Match"

This is Google's equivalent of a Matched Audience.

  1. In     Google Ads, navigate to "Tools and Settings" > "Shared     Library" > "Audience manager."
  2. Click     the "+" button and select "Customer list."
  3. Upload     your curated CSV seed list.
  4. Google     will match the emails in your list to Google accounts. This process can     also take up to 48 hours.

Step 2: Create a "Similar Audience"

This is Google's AI-powered lookalike feature.

  1. Once     your Customer Match list has populated, Google will automatically begin     generating a "Similar Audience" in the background.
  2. Google's     AI analyzes thousands of signals from your seed list, including their     search history, YouTube viewing habits, apps they use, and websites they     visit.
  3. It     then finds a new, larger audience of users who exhibit similar online     behaviors, indicating they are likely "in-market" for similar     solutions.

Step 3: Activate Your Similar Audience in Campaigns

You can use this powerful new audience in two primary ways:

  • Targeting     Mode: Create a new Display, Discovery, or YouTube campaign and set the     targeting to only your Similar Audience. This is a great way to run     a high-potential prospecting campaign to a new audience.
  • Observation     Mode (for Search Campaigns): This is a key tactic. Add your Similar     Audience to an existing Search campaign under the "Observation"     setting. Then, apply a positive bid adjustment (e.g., +25%). This tells     Google, "Continue showing my ads to anyone who searches for my keywords,     but if the person searching is also part of this high-value Similar     Audience, I'm willing to bid 25% more to win that click."

Step 4: Use "Audience Insights" for Discovery

Just like with LinkedIn, you can analyze the composition of your AI-generated audience.

  1. In the     Audience Manager, select your Similar Audience and explore its insights.
  2. Google     will show you a breakdown of its Affinity Categories (e.g.,     "Business Professionals"), In-Market Segments (e.g.,     "Business Financial Services"), demographics, and locations.
  3. This     can reveal surprising facets of your ICP. You might discover your best     customers have a high affinity for certain types of software or follow     specific financial news outlets, providing new ideas for ad placements and     content partnerships.

Step 5: Optimize with Data-Driven Bidding Strategies

Google's AI lives and breathes conversion data. To get the most out of your campaigns, you must have accurate conversion tracking set up. Use AI-powered bidding strategies like Maximize Conversions or Target CPA. When you use these strategies, you are essentially telling Google's algorithm, "Here is my goal. Here is my Similar Audience. You have all the data. Now go find me the most conversions possible within my budget."

Frequently Asked Questions (FAQs)

Q1: How many contacts do I really need for a good seed list? A: The official recommendation from both platforms is typically 1,000 matched users. To get there, you'll likely need to upload a list of 2,000-3,000 contacts, given average match rates. However, you can start with smaller lists (as low as a few hundred) to get the process going. The key is quality over quantity. A list of 500 of your absolute best customers is far better than a list of 5,000 mediocre ones.

Q2: What is a good "match rate" for my uploaded list? A: You should expect a match rate between 30% and 60%.It will rarely be higher because not every business email is associated with a personal LinkedIn or Google account. If your match rate is below 30%, it may be a sign that the emails in your CRM are outdated.

Q3: How is a lookalike/similar audience different from standard targeting? Which is better? A: Standard targeting is you telling the platform exactly who to target (e.g., VPs of Marketing). A lookalike audience is you giving the platform an example of who you want and letting its AI find people like them. Neither is inherently "better"; they are different tools. Often, the best results come from combining them: targeting a Lookalike Audience and then layering on one or two key firmographic filters. Always test both.

Q4: Is using customer data for lookalikes compliant with privacy laws like GDPR and CCPA? A: Yes. The platforms are designed to be privacy-compliant. When you upload your list, the data is "hashed"—a cryptographic process that turns the email into an irreversible, anonymous string of characters. The platform matches these hashed strings. You never see the personal data of the users in the lookalike audience, and the platforms do not share that information with you.

Q5: How often should I update my seed lists and lookalike audiences? A: It's a good practice to refresh your seed lists and generate new lookalike audiences on a quarterly basis. As you acquire new "best" customers, adding them to the seed list gives the AI fresh data to learn from, keeping your audiences relevant and high-performing.

Conclusion: Stop Guessing, Start Discovering

The days of crafting your Ideal Customer Profile based on assumptions in a conference room are over. The most successful B2B marketers of2025 and beyond will be the ones who can effectively partner with AI to let their own data reveal the truth. Your best customers are leaving a trail of thousands of digital breadcrumbs, and the AI engines within LinkedIn and Google are uniquely capable of following that trail to find more people just like them.

By building a high-quality seed list and following this step-by-step process, you can move from guesswork to a data-driven system of audience discovery. You will uncover new pockets of growth, improve your ad efficiency, and gain a much deeper, more nuanced understanding of who your customer truly is. Stop guessing. The answers are in your data. It's time to let AI help you find them.

 

‍