If you’ve been running Meta ads the same way you did two years ago, you’re not running them wrong. You’re running them in a world that no longer exists.
On January 15, 2026, Meta removed dozens of detailed interest targeting categories from Ads Manager. Gone. The granular interest stacking that performance marketers spent a decade perfecting, layering “online shopping” on top of “female entrepreneurs” on top of “digital products” to build a perfectly segmented audience, is largely unavailable now. And the advertisers who hadn’t already shifted their strategy to match how Meta’s algorithm actually works in 2026 felt it immediately.
I’ve been running Meta ads for my own digital product business since I launched in May 2026, and I’ve spent close to a decade running paid campaigns for global brands. The fundamental shift in how Meta’s targeting works is the single biggest change I’ve had to adapt my approach around. This post covers what that shift looks like in practice, how to structure audiences that work with the algorithm instead of against it, and the specific setup I use for digital products.
What Actually Changed on January 15, 2026
Meta’s January 2026 update removed a significant number of specific interest categories, particularly those in the areas of politics, religion, and detailed behavioral segments. But the bigger and less-discussed shift happened over the year prior: Meta’s Andromeda algorithm changed how audience inputs work at a fundamental level.
Search Engine Land’s 2026 breakdown of Meta’s Andromeda system put it clearly: Andromeda shifted Meta to creative-first matching, evaluating visuals, themes, hooks, and language to decide relevance. The days of hyper-segmentation are gone.
What this means in practice: the audience inputs you put into Ads Manager are now suggestions, not restrictions. According to AdsUploader’s 2026 targeting breakdown, only location, languages, minimum age, exclusions, and Special Ad Category restrictions are treated as hard controls. Age, gender, detailed targeting, custom audiences, and lookalike audiences are all treated as suggestions that steer Meta’s AI but don’t bind it. The algorithm will show your ads outside your defined audience when it predicts better performance.
This is not a bug. It’s by design. And once you understand it, targeting becomes a different exercise entirely.
The Four Audience Types That Still Exist in 2026
Before getting into strategy, you need to know what your actual options are right now.
Core (Detailed) Targeting
This is the traditional interest and behavior-based targeting that most people learned first. You select demographics, interests, and behaviors from Meta’s library. As of 2026, this is a starting signal, not a hard filter. Stackmatix’s 2026 targeting options list notes that detailed targeting is the default starting point when you have no first-party data to build on yet, but the algorithm now prioritizes AI-driven audience discovery over manual interest selection.
Use 3 to 6 broad relevant interest categories rather than stacking many narrow ones. Accounts that use light guidance with strong creative signals tend to scale faster than those with heavily segmented targeting, according to Conversios’ 2026 Advantage+ guide.
Custom Audiences
Custom audiences are the highest-quality first-party data you can give Meta’s algorithm. They’re built from people who already have a relationship with your business: website visitors, email list subscribers, video viewers, Instagram engagers, and past purchasers.
Pixel Movers’ 2026 targeting analysis is direct about this: the signals that matter now are first-party data from your website, email list, and customer base. This is the data that feeds the algorithm the clearest signal of who your actual buyer is.
For digital product sellers specifically, the custom audiences worth building are: website visitors in the last 30 days, people who initiated checkout but didn’t purchase, your email list, Instagram profile engagers from the last 60 days, and video viewers who watched at least 75% of a Reel. These are your warmest audiences and they should be feeding your retargeting campaigns even when Advantage+ is doing broad delivery on cold traffic.
Lookalike Audiences
A lookalike audience tells Meta to find people who share characteristics with an existing custom audience. The 1% to 10% framework still applies, where 1% is the closest match to your source and larger percentages trade similarity for reach.
The important update: lookalikes are now treated as audience suggestions within Advantage+ Audience rather than hard constraints. Linear Design’s 2026 Advantage+ breakdown notes that even when you attempt to turn Advantage+ off, you can rarely restrict by lookalike inputs, and for nine of the most common performance goals, lookalikes will always be used as suggestions.
What this means for digital product sellers: build your lookalikes from the best source audiences you have, typically your email list or your past purchasers, and use them as suggestions to point the algorithm in the right direction. The better the source list in terms of size, quality, and recency, the better the Advantage+ delivery. A lookalike built on 100 past customers performs worse than one built on 5,000.
Advantage+ Audience
This is Meta’s AI-powered targeting system that has become the default for most campaign objectives in 2026. Instead of you building a specific audience, you provide your ad creative and a conversion goal, and Meta’s AI finds the people most likely to convert using signals from across its entire ecosystem.
Meta’s internal benchmarks via Conversios show Advantage+ Audience delivers up to 32% lower CPA and 13% lower cost per catalog sale compared to manual targeting setups. In an independent 30-day A/B test by Jon Loomer Digital, Advantage+ drove more registrations than both detailed targeting and lookalikes, though the gap over detailed targeting wasn’t statistically significant across all accounts.
Advantage+ works best for accounts with at least 50 weekly conversions and budgets above $50 per day. For newer accounts or lower budgets, detailed targeting with a focused starting point is often more effective because the algorithm needs a tighter direction before it has enough conversion data to optimize freely.
- 1. Core (Detailed) Targeting — interest and behavior-based inputs; now treated as suggestions not restrictions; best used as a broad starting signal for new accounts with no first-party data
- 2. Custom Audiences — first-party data from your website, email list, and on-platform activity; the highest-quality signal you can give the algorithm; feeds both retargeting and lookalike creation
- 3. Lookalike Audiences — Meta finds people similar to your custom audience source; now used as suggestions within Advantage+ rather than hard constraints; quality of source list directly affects quality of lookalike
- 4. Advantage+ Audience — Meta’s AI targeting system; you provide creative and a conversion goal; algorithm finds converters across the entire platform; delivers up to 32% lower CPA vs manual targeting on accounts with sufficient data
Hard Controls vs Audience Suggestions: The Distinction That Changes Everything
This is the concept most advertisers are still confused about, and it’s the most important thing to understand about targeting in 2026.
According to AdsUploader’s Meta audience targeting guide, your inputs split into two categories: hard controls and audience suggestions.
Hard controls are respected absolutely. They are: location, languages, minimum age, exclusions for specific custom audiences, and Special Ad Category restrictions. When you set these, Meta will not show your ads outside these parameters.
Audience suggestions are everything else. Age ranges, gender, detailed interests, custom audiences, and lookalike audiences are all suggestions. They tell the AI where to start, but when the system predicts better performance outside those parameters, it will expand. Even the age range you set is a suggestion by default now, with only the minimum age surviving as a true control.
The practical implication: stop trying to build a precise target audience through interest stacking. You’re not restricting who sees your ad. You’re giving the algorithm a direction and then getting out of the way. The better your creative at self-selecting the right person through the hook, the visuals, and the copy, the better the algorithm will optimize toward your actual buyer regardless of what inputs you put in the audience section.
What This Means for Digital Product Sellers Specifically
Selling digital products on Meta is different from selling physical ecommerce products in one important way: your buyer intent is harder to predict from behavioral data. Someone who buys athletic wear has clear behavioral signals. Someone who is about to buy a digital course on content strategy has a much messier intent profile.
This is actually good news in 2026, because the shift to creative-first matching plays in your favor. The hook and the first three seconds of your ad are doing more targeting work than any interest category. Someone who watches a Reel about how to post content without being on camera and then buys a faceless content collection is being targeted by the creative itself, not by an interest layer.
Here’s how I structure audience targeting for my own product campaigns:
For Cold Traffic Campaigns
Use Advantage+ Audience with 3 to 4 broad interest suggestions relevant to your niche. For digital products aimed at creators, that might be “digital marketing,” “online business,” “content creation,” and “entrepreneurship.” Don’t stack narrow sub-interests. Don’t obsess over the audience size meter. Set your location, set your minimum age, add your interest suggestions, and let the algorithm do the rest.
The recommended audience size according to AdsUploader remains 2 to 10 million people. Bigger is generally better in 2026 because the AI needs room to find your actual buyer.
For Warm Retargeting
Build specific custom audiences from your website visitors, email list, and video viewers. These audiences are small, which is fine, because they convert at a significantly lower cost per acquisition than cold traffic. Run these as separate campaigns with separate budgets and retargeting-specific creative that addresses the specific objection that stopped someone from buying on their first visit.
Jon Loomer’s 2026 targeting guide makes the point that general remarketing should no longer be necessary as a separate campaign because Meta already prioritizes it dynamically within Advantage+. The exception for digital product sellers is high-intent retargeting, specifically people who hit your checkout page or added to cart, where the conversion probability is high enough that a dedicated budget and dedicated creative is worth it.
For Lookalike Campaigns
Build your lookalikes from your best source: past purchasers first, email subscribers second, high-value website visitors third. Use a 1% to 2% lookalike as an audience suggestion within Advantage+ Audience. Don’t create separate ad sets for 1%, 2%, and 3% lookalikes and run them simultaneously. That fragments your budget and prevents any of them from accumulating enough data to exit the learning phase.
Creative Is the New Targeting
This keeps coming up in every 2026 analysis of Meta ads performance and it’s worth stating directly: your creative is doing more audience targeting work than your audience inputs are.
Meta’s Q2 2026 earnings data via AdsUploader shows an 8.3% increase in ad clicks and 15.7% conversion uplift attributed to Meta’s newer GEM ranking model, which evaluates creative relevance to predict conversion likelihood. The algorithm is reading your ad and making targeting decisions based on what it finds there.
This means a hook that speaks directly to your exact buyer (their specific situation, their specific pain, their specific desired outcome) is functioning as a targeting mechanism. A Reel that opens with “if you’ve been posting every day and still not making sales” is going to find the people that sentence resonates with, regardless of whether you’ve put “digital products” in your interest targeting. The creative self-selects.
For digital product sellers, this means: spend as much time on your first three seconds as you do on your entire audience setup. The hook does the targeting. The audience setup just points the algorithm at a starting population.
If you want the full breakdown of what makes Meta ad creative convert for digital product campaigns, including the Triple Hook Formula, the ASCENDANT Ads Manager has an entire module dedicated to creative frameworks and a built-in hook generator.
What to Actually Do in Ads Manager Right Now
Here’s the practical setup for a cold traffic purchase campaign for a digital product in 2026:
One campaign. Advantage+ Shopping or manual Sales objective with Advantage+ Audience turned on. One to two ad sets maximum. Set your location and minimum age as hard controls. Add 3 to 4 broad interest suggestions, not restrictions. Upload 3 to 5 creative variants. Set your daily budget at a level that can accumulate 50 optimization events within 7 to 14 days based on your average purchase conversion rate. Leave it alone for at least 7 days.
That’s it. The era of building 10 ad sets with different targeting combinations and comparing performance is over. That approach fragments your conversion data across multiple ad sets, prevents any of them from exiting the learning phase, and fights against how the algorithm is designed to work in 2026.
For the complete setup walkthrough including campaign structure, budget guidance, what each metric in Ads Manager actually tells you, and when to touch your campaigns versus when to leave them alone, the ASCENDANT Ads Manager covers all of it in an 11-module interactive format built specifically for digital product sellers.
The Meta ads best practices for digital product sellers post covers the foundational setup. And if you’re still trying to figure out why your current campaigns aren’t converting, the most expensive Meta ads mistakes I’ve made covers the specific errors that cost the most, including the audience fragmentation issue in detail.
Frequently Asked Questions
Does detailed targeting still work on Meta in 2026?
It works as a starting signal, not a restriction. Meta removed dozens of specific interest categories on January 15, 2026, and the remaining categories are now treated as suggestions that the algorithm can expand beyond when it predicts better performance elsewhere. Three to four broad relevant interests used as guidance within Advantage+ Audience is the current best practice. Interest stacking to build a precise narrow audience no longer works the way it did before 2024.
Should I use Advantage+ Audience or manual targeting in 2026?
Advantage+ Audience for accounts with 50 or more weekly conversions and budgets above $50 per day. Manual detailed targeting as a starting point for brand new accounts with no conversion history, budgets under $30 per day, or hyper-local campaigns. Meta’s internal benchmarks show Advantage+ delivers up to 32% lower CPA compared to manual setups on accounts with sufficient data, but that advantage depends on the algorithm having enough signal to learn from.
How do I build a custom audience for a digital product business?
Start with your website visitors using the Meta Pixel, segmented by product page visitors, checkout initiators, and purchasers. Add your email list as a customer list custom audience. Build video viewer audiences from anyone who watched 75% or more of your recent Reels. These are your warmest audiences and the highest-quality source data for building lookalike audiences. The quality of your custom audience source directly determines the quality of any lookalike built from it.
How big should my Meta ads audience be in 2026?
Meta’s recommended audience size is 2 to 10 million people for most campaign objectives. Bigger audiences give the AI more room to find your actual buyer. The instinct to narrow audiences to a very specific segment of a few hundred thousand people works against the algorithm’s ability to optimize. Set your hard controls on location and minimum age, add broad interest suggestions, and let the audience size sit in the millions rather than the hundreds of thousands.
What happened to lookalike audiences on Meta in 2026?
Lookalike audiences still exist but are now treated as audience suggestions within Advantage+ Audience rather than hard constraints. For nine of the most common campaign objectives, lookalikes will always function as suggestions even if you attempt to restrict by them. They still provide useful directional signals to the algorithm, particularly for newer accounts where conversion history is limited. Build them from your highest-quality source audiences: past purchasers first, email subscribers second.
Is it worth running separate retargeting campaigns in 2026?
For most campaigns, Meta already handles remarketing dynamically within Advantage+. The specific exception worth running as a separate campaign is high-intent retargeting: people who initiated checkout, added to cart, or visited your product page multiple times. These audiences are small but convert at a significantly lower cost per acquisition, and dedicated creative addressing the specific objection that stopped them from buying is worth the separate budget allocation.
How do I know if my targeting is the problem or if it’s my creative?
Check your CTR first. If your CTR is above 2% but your conversion rate is low, the ad is finding the right people but your sales page or checkout is the issue. If your CTR is below 1%, the creative isn’t resonating with whoever the algorithm is showing it to, which is either a creative problem or a signal quality problem from weak first-party data. Targeting is rarely the isolated problem in 2026 because the algorithm is doing most of the heavy lifting. Creative quality and conversion data quality are the two levers that matter most.
