Should I Use Interest Targeting in 2026?
Digital marketing
Meta Ads
If you run Meta ads, you've probably asked at least one of these out loud this month:
Is interest targeting dead in 2026?
Does broad targeting actually work, or is it just what everyone says now?
Why have my lookalike audiences quietly stopped pulling weight?
How many audiences am I supposed to be testing at once?
You're not imagining it. Targeting on Meta looks nothing like it did two years ago. iOS killed a chunk of the signal Meta used to lean on, Advantage+ rolled across most account types, and the algorithm got good enough to make a lot of old playbooks worse than doing nothing.
This is the working version of how audience targeting actually behaves now, written for people spending real money, not for blog SEO bots. (Well, also for blog SEO bots. Hi.)
Should you still use interest targeting in 2026?
Short answer: yes, but stop treating it as your main lever.
Interest targeting used to be the whole game. Pick "yoga," pick "small business owners," pick "digital marketing," let it run. In 2026 it works very differently because two things changed at the same time.
First, Meta's algorithm got significantly better at finding buyers on its own. It doesn't need you to hand it the audience anymore. Most of the time, when you over-restrict with interests, you're making the model's job harder, not easier.
Second, the interest data itself is weaker. Fewer people like pages. Fewer people sit in groups. Self-reported interests from 2019 are still in the system. When you target "entrepreneurship," you're reaching anyone who ever tapped a motivational business post in 2017. The signal is noisy and old.
That doesn't mean interest targeting is useless. It still earns its keep in a few specific cases:
New accounts with no pixel data. Before you have ~50 conversions a week, broad has nothing to optimize toward. Interests give the algorithm a starting box.
Genuinely niche products. If you sell left-handed mechanical keyboards or sourdough starter kits, narrow interests find the small pocket of buyers fast.
Awareness and top-of-funnel campaigns. When your goal is reach inside a defined group (say, marketing managers at SaaS companies), interests still let you point at the rough shape of that group.
For everything else, my honest take: use interests as a guardrail, not a strategy. They shape the room. Your creative does the actual selling.
Does broad targeting actually work?
Yes. With caveats.
Broad targeting means stripping out the interest layers, opening up the age range, and letting Meta's algorithm decide who sees the ad. In 2026, this is the default for most accounts that have any pixel history. Advantage+ shopping campaigns essentially force you into a version of it anyway.
When broad targeting wins, it wins because:
The pixel has enough conversion data to actually optimize (you want 50+ weekly conversions per ad set as a rule of thumb)
The product has wide enough appeal that the algorithm has room to find pockets
The budget is high enough to escape the learning phase (running broad on $10/day is just lighting money on fire slowly)
The creative is doing the targeting for you
That last one is the one most people miss. When you go broad, your creative becomes the filter. The headline, the hook in the first three seconds, the on-screen text, the offer, the image, all of it has to repel the wrong people and pull in the right ones. If your ad could be for any brand in any category, broad will spend your money showing it to everyone and converting nobody.
Most "broad doesn't work for me" stories are actually creative problems wearing a targeting costume.
A practical way to start: run broad for at least 7 to 14 days with a budget that gives you meaningful daily conversions, pair it with two or three strong creatives, and only judge it after the learning phase exits. If it's still flat after two weeks with good creative and adequate spend, then layer in some interest signals or rebuild the funnel.
Why your lookalike audiences aren't performing the way they used to
This one stings for advertisers who built whole agencies on lookalikes. The 1% lookalike used to be money. Upload a customer list, click a button, watch the CPAs drop. That magic faded for three concrete reasons.
1. The source data got worse after iOS
App Tracking Transparency took a chunk out of the pixel data Meta uses to build lookalikes. The model is now training on a partial picture of your converters. A lookalike built from incomplete data finds people who look like a sample of your customers, not your customers. The math just doesn't work as cleanly.
2. Your source audience is probably too small or too old
A 1% lookalike off 200 past customers is not really a lookalike. It's a guess wearing a 1% label. Most case studies I trust point to 1,000+ as the floor for source size, and 5,000+ being where you actually get stable results.
The other half of this is freshness. A customer list from 2022 builds a lookalike of who used to buy from you. Buyer behavior shifts. Your offer probably shifted. Run the lookalike off purchasers from the last 90 to 180 days instead.
3. The algorithm now does what lookalikes used to do
This is the uncomfortable part. Broad targeting in 2026 often finds an audience that looks a lot like your lookalike anyway, because that's literally what the algorithm is optimizing for. Adding a lookalike layer on top can just shrink your reach without changing the quality.
If you still want to run lookalikes, the version that actually works:
Rebuild the source every quarter using recent buyers
Use a value-based source (upload purchase values, not just emails) so the model targets high-value customers, not just any customer
Test 2-5% before assuming tight 1% will outperform (it usually doesn't anymore)
Pair lookalikes with strong retargeting so warm traffic isn't getting wasted
Or just go broad with great creative and skip the rebuild work. Both are valid in 2026. The lookalike-or-die era is over.
How many audiences should you test at once?
The honest answer most people don't want to hear: fewer than you're currently testing.
Testing 10 audiences at once on a $50/day budget is not testing. It's spreading $5 across ten ad sets, none of which exit learning, all of which feed you noisy data, and the only conclusion you can draw is "everything is bad."
Testing one audience is the opposite problem. You learn nothing about whether the audience or the creative was the variable. When it fatigues, you're caught flat.
The sweet spot for most accounts is 3 to 5 audiences per test, in clear phases.
Phase 1, weeks 1-2: Audience test. Run 3 to 5 audiences with the same creative and the same budget per ad set. You're isolating the audience variable. Give each at least $50-100 in spend or 3-5 days before you kill anything. Killing an ad set at $20 and two days is not a decision, it's a vibe.
Phase 2, weeks 3-4: Creative test. Take the top 1 or 2 audiences and run 3 to 5 different creatives inside them. Now you know which audience-creative pair is actually working.
Phase 3, ongoing: Scale and refresh. Once you have a winner, scale slowly. 20-30% budget increase at a time, not a doubling. Refresh creative every 3 to 4 weeks. Creative fatigue is the leading cause of "the campaign just stopped working" and almost nobody catches it early enough.
The whole point of this structure is that you can read the results. If you're running 10 audiences with 5 creatives each on a small budget, you have 50 cells of data, all of them too small to mean anything. You'll convince yourself of patterns that aren't there.
A simple framework for Meta ads targeting in 2026
If you want a working order of operations, here's what I actually do on accounts:
Check your pixel maturity first. If you have 50+ weekly conversions, start broad. If you don't, start with one or two interest-based audiences to give the algorithm a seed, and graduate to broad later.
Spend on creative before spending on audience strategy. In 2026 the audience is mostly decided by the algorithm. The creative is decided by you. Guess which one has more leverage.
Rebuild lookalikes only if you have a strong, fresh, large source. Otherwise let them go and run broad instead.
Test 3 to 5 audiences per cycle, not 10. Give each enough budget and time to actually report something.
Refresh creative on a calendar, not on a panic. Every 3 to 4 weeks, new variations into the winning audience.
Review weekly, scale monthly. Daily scaling is how you blow up a working campaign in 48 hours.
Final thoughts
The advertisers I see winning in 2026 are not the ones with the most clever audience setups. The clever-stack era ended around the time Advantage+ shipped. They're the ones treating audience targeting as something the algorithm mostly handles, and putting their actual brainpower into creative, offers, and structured testing.
That's the boring answer. It's also the one that's working.
If you've been pivoting between strategies every two weeks because results dipped, the problem usually isn't your targeting. It's that nothing has been allowed to run long enough to mean anything. Pick a framework, fund it properly, and let it actually generate data before you change course.
FAQ
Is interest targeting dead in 2026? No.
It's just no longer the main lever. Use it for new accounts without pixel data, genuinely niche products, and top-of-funnel awareness. For most everything else, broad with strong creative outperforms.
What's the minimum budget for broad targeting on Meta?
There's no fixed number, but you need enough spend to generate ~50 conversions a week per ad set to give the algorithm something to optimize toward. For most ecom accounts that's $50-100/day per ad set minimum.
How many conversions do I need before broad targeting works?
About 50 per week per ad set is the working threshold for Meta to exit learning and optimize reliably. Below that, broad is mostly guessing.
Should I use 1% or higher percentage lookalikes in 2026?
Test 2-5% before assuming 1% is best. With weaker post-iOS data, tight lookalikes often underperform wider ones. The old advice doesn't hold anymore.
How long should I run an audience test before killing it?
At least 3-5 days or $50-100 in spend per ad set, whichever comes later. Killing audiences faster than that is a guess, not a test.